idnits 2.17.1 draft-ietf-ipatm-ipmc-09.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** Cannot find the required boilerplate sections (Copyright, IPR, etc.) in this document. Expected boilerplate is as follows today (2024-04-26) according to https://trustee.ietf.org/license-info : IETF Trust Legal Provisions of 28-dec-2009, Section 6.a: This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. IETF Trust Legal Provisions of 28-dec-2009, Section 6.b(i), paragraph 2: Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved. IETF Trust Legal Provisions of 28-dec-2009, Section 6.b(i), paragraph 3: This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- ** Missing expiration date. The document expiration date should appear on the first and last page. ** The document seems to lack a 1id_guidelines paragraph about Internet-Drafts being working documents. ** The document seems to lack a 1id_guidelines paragraph about 6 months document validity. ** The document seems to lack a 1id_guidelines paragraph about the list of current Internet-Drafts. ** The document seems to lack a 1id_guidelines paragraph about the list of Shadow Directories. ** The document is more than 15 pages and seems to lack a Table of Contents. == No 'Intended status' indicated for this document; assuming Proposed Standard == The page length should not exceed 58 lines per page, but there was 1 longer page, the longest (page 1) being 61 lines Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) ** The document seems to lack separate sections for Informative/Normative References. All references will be assumed normative when checking for downward references. ** There are 3 instances of too long lines in the document, the longest one being 1 character in excess of 72. == There are 19 instances of lines with multicast IPv4 addresses in the document. If these are generic example addresses, they should be changed to use the 233.252.0.x range defined in RFC 5771 ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 245: '... administrators MUST ensure that each...' RFC 2119 keyword, line 249: '... restriction MAY only occur after fu...' RFC 2119 keyword, line 574: '...control messages MUST be LLC/SNAP enca...' RFC 2119 keyword, line 608: '... null addresses SHALL be encoded as z...' RFC 2119 keyword, line 672: '...the mar$pro.snap field MUST be zero on...' (151 more instances...) Miscellaneous warnings: ---------------------------------------------------------------------------- == Line 924 has weird spacing: '...qoctets sourc...' == Line 925 has weird spacing: '...roctets sourc...' == Line 926 has weird spacing: '...soctets sourc...' == Line 927 has weird spacing: '...zoctets targe...' == Line 928 has weird spacing: '...xoctets targe...' == (30 more instances...) == Using lowercase 'not' together with uppercase 'MUST', 'SHALL', 'SHOULD', or 'RECOMMENDED' is not an accepted usage according to RFC 2119. Please use uppercase 'NOT' together with RFC 2119 keywords (if that is what you mean). Found 'MUST not' in this paragraph: A host or endpoint interface that is using the same MARS to support multicasting needs of multiple protocols MUST not assume their CMI will be the same for each protocol. -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- Couldn't find a document date in the document -- date freshness check skipped. -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Looks like a reference, but probably isn't: '0xAA-AA-03' on line 2775 -- Looks like a reference, but probably isn't: '0x00-00-5E' on line 2775 -- Looks like a reference, but probably isn't: '0x00-03' on line 577 -- Looks like a reference, but probably isn't: 'Addresses' on line 2790 -- Looks like a reference, but probably isn't: '0x00-01' on line 2775 -- Looks like a reference, but probably isn't: '0x00-04' on line 1866 -- Looks like a reference, but probably isn't: '0x00-80' on line 1901 -- Looks like a reference, but probably isn't: '0x800' on line 1964 -- Looks like a reference, but probably isn't: '0x86DD' on line 2775 -- Looks like a reference, but probably isn't: '0x00-00' on line 2893 -- Looks like a reference, but probably isn't: '-' on line 3353 ** Obsolete normative reference: RFC 1483 (ref. '2') (Obsoleted by RFC 2684) ** Obsolete normative reference: RFC 1577 (ref. '3') (Obsoleted by RFC 2225) -- Possible downref: Non-RFC (?) normative reference: ref. '4' ** Downref: Normative reference to an Experimental RFC: RFC 1075 (ref. '5') ** Downref: Normative reference to an Informational RFC: RFC 1821 (ref. '7') -- Possible downref: Non-RFC (?) normative reference: ref. '8' Summary: 15 errors (**), 0 flaws (~~), 10 warnings (==), 16 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 1 Grenville Armitage gja@thumper.bellcore.com 2 Bellcore, 445 South St. http://gump.bellcore.com:8000/~gja/home.html 3 Morristown, NJ 07960 USA (voice) +1 201 829 2635 {.. 2504 (fax)} 5 Internet-Draft Grenville Armitage 6 Bellcore 7 November 17th, 1995 9 Support for Multicast over UNI 3.0/3.1 based ATM Networks. 10 12 Status of this Memo 14 This document was submitted to the IETF IP over ATM WG. Publication 15 of this document does not imply acceptance by the IP over ATM WG of 16 any ideas expressed within. Comments should be submitted to the ip- 17 atm@matmos.hpl.hp.com mailing list. 19 Distribution of this memo is unlimited. 21 This memo is an internet draft. Internet Drafts are working documents 22 of the Internet Engineering Task Force (IETF), its Areas, and its 23 Working Groups. Note that other groups may also distribute working 24 documents as Internet Drafts. 26 Internet Drafts are draft documents valid for a maximum of six 27 months. Internet Drafts may be updated, replaced, or obsoleted by 28 other documents at any time. It is not appropriate to use Internet 29 Drafts as reference material or to cite them other than as a "working 30 draft" or "work in progress". 32 Please check the lid-abstracts.txt listing contained in the 33 internet-drafts shadow directories on ds.internic.net (US East 34 Coast), nic.nordu.net (Europe), ftp.isi.edu (US West Coast), or 35 munnari.oz.au (Pacific Rim) to learn the current status of any 36 Internet Draft. 38 Abstract 40 Mapping the connectionless IP multicast service over the connection 41 oriented ATM services provided by UNI 3.0/3.1 is a non-trivial task. 42 This memo describes a mechanism to support the multicast needs of 43 Layer 3 protocols in general, and describes its application to IP 44 multicasting in particular. 46 ATM based IP hosts and routers use a Multicast Address Resolution 47 Server (MARS) to support RFC 1112 style Level 2 IP multicast over the 48 ATM Forum's UNI 3.0/3.1 point to multipoint connection service. 49 Clusters of endpoints share a MARS and use it to track and 50 disseminate information identifying the nodes listed as receivers for 51 given multicast groups. This allows endpoints to establish and manage 52 point to multipoint VCs when transmitting to the group. 54 The MARS behaviour allows Layer 3 multicasting to be supported using 55 either meshes of VCs or ATM level multicast servers. This choice may 56 be made on a per-group basis, and is transparent to the endpoints. 58 Contents. 60 1. Introduction. 61 1.1 The Multicast Address Resolution Server (MARS). 62 1.2 The ATM level multicast Cluster. 63 1.3 Document overview. 64 1.4 Conventions. 65 2. The IP multicast service model. 66 3. UNI 3.0/3.1 support for intra-cluster multicasting. 67 3.1 VC meshes. 68 3.2 Multicast Servers. 69 3.3 Tradeoffs. 70 3.4 Interaction with local UNI 3.0/3.1 signalling entity. 71 4. Overview of the MARS. 72 4.1 Architecture. 73 4.2 Control message format. 74 4.3 Fixed header fields in MARS control messages. 75 4.3.1 Hardware type. 76 4.3.2 Protocol type. 77 4.3.3 Checksum. 78 4.3.4 Extensions Offset. 79 4.3.5 Operation code. 80 4.3.6 Reserved. 81 5. Endpoint (MARS client) interface behaviour. 82 5.1 Transmit side behaviour. 83 5.1.1 Retrieving Group Membership from the MARS. 84 5.1.2 MARS_REQUEST, MARS_MULTI, and MARS_NAK messages. 85 5.1.3 Establishing the outgoing multipoint VC. 86 5.1.4 Monitoring updates on ClusterControlVC. 87 5.1.4.1 Updating the active VCs. 88 5.1.4.2 Tracking the Cluster Sequence Number. 89 5.1.5 Revalidating a VC's leaf nodes. 90 5.1.5.1 When leaf node drops itself. 91 5.1.5.2 When a jump is detected in the CSN. 92 5.1.6 'Migrating' the outgoing multipoint VC. 93 5.2. Receive side behaviour. 94 5.2.1 Format of the MARS_JOIN and MARS_LEAVE Messages. 95 5.2.1.1 Important IPv4 default values. 96 5.2.2 Retransmission of MARS_JOIN and MARS_LEAVE messages. 97 5.2.3 Cluster member registration and deregistration. 98 5.3 Support for Layer 3 group management. 99 5.4 Support for redundant/backup MARS entities. 100 5.4.1 First response to MARS problems. 101 5.4.2 Connecting to a backup MARS. 102 5.4.3 Dynamic backup lists, and soft redirects. 103 5.5 Data path LLC/SNAP encapsulations. 104 5.5.1 Type #1 encapsulation. 105 5.5.2 Type #2 encapsulation. 107 5.5.3 A Type #1 example. 108 6. The MARS in greater detail. 109 6.1 Basic interface to Cluster members. 110 6.1.1 Response to MARS_REQUEST. 111 6.1.2 Response to MARS_JOIN and MARS_LEAVE. 112 6.1.3 Generating MARS_REDIRECT_MAP. 113 6.1.4 Cluster Sequence Numbers. 114 6.2 MARS interface to Multicast Servers (MCSs). 115 6.2.1 MARS_REQUESTs for MCS supported groups. 116 6.2.2 MARS_MSERV and MARS_UNSERV messages. 117 6.2.3 Registering a Multicast Server (MCS). 118 6.2.4 Modified response to MARS_JOIN and MARS_LEAVE. 119 6.2.5 Sequence numbers for ServerControlVC traffic. 120 6.3 Why global sequence numbers? 121 6.4 Redundant/Backup MARS Architectures. 122 7. How an MCS utilises a MARS. 123 7.1 Association with a particular Layer 3 group. 124 7.2 Termination of incoming VCs. 125 7.3 Management of outgoing VC. 126 7.4 Use of a backup MARS. 127 8. Support for IP multicast routers. 128 8.1 Forwarding into a Cluster. 129 8.2 Joining in 'promiscuous' mode. 130 8.3 Forwarding across the cluster. 131 8.4 Joining in 'semi-promiscous' mode. 132 8.5 An alternative to IGMP Queries. 133 8.6 CMIs across multiple interfaces. 134 9. Multiprotocol applications of the MARS and MARS clients. 135 10. Supplementary parameter processing. 136 10.1 Interpreting the mar$extoff field. 137 10.2 The format of TLVs. 138 10.3 Processing MARS messages with TLVs. 139 10.4 Initial set of TLV elements. 140 11. Key Decisions and open issues. 141 Acknowledgments 142 References 143 Appendix A. Hole punching algorithms. 144 Appendix B. Minimising the impact of IGMP in IPv4 environments. 145 Appendix C. Further comments on 'Clusters'. 146 Appendix D. TLV list parsing algorithm. 147 Appendix E. Summary of timer values. 148 Appendix F. Pseudo code for MARS operation. 150 1. Introduction. 152 Multicasting is the process whereby a source host or protocol entity 153 sends a packet to multiple destinations simultaneously using a 154 single, local 'transmit' operation. The more familiar cases of 155 Unicasting and Broadcasting may be considered to be special cases of 156 Multicasting (with the packet delivered to one destination, or 'all' 157 destinations, respectively). 159 Most network layer models, like the one described in RFC 1112 [1] for 160 IP multicasting, assume sources may send their packets to an abstract 161 'multicast group addresses'. Link layer support for such an 162 abstraction is assumed to exist, and is provided by technologies such 163 as Ethernet. 165 ATM is being utilized as a new link layer technology to support a 166 variety of protocols, including IP. With RFC 1483 [2] the IETF 167 defined a multiprotocol mechanism for encapsulating and transmitting 168 packets using AAL5 over ATM Virtual Channels (VCs). However, the ATM 169 Forum's currently published signalling specifications (UNI 3.0 [8] 170 and UNI 3.1 [4]) does not provide the multicast address abstraction. 171 Unicast connections are supported by point to point, bidirectional 172 VCs. Multicasting is supported through point to multipoint 173 unidirectional VCs. The key limitation is that the sender must have 174 prior knowledge of each intended recipient, and explicitly establish 175 a VC with itself as the root node and the recipients as the leaf 176 nodes. 178 This document has two broad goals: 180 Define a group address registration and membership distribution 181 mechanism that allows UNI 3.0/3.1 based networks to support the 182 multicast service of protocols such as IP. 184 Define specific endpoint behaviours for managing point to 185 multipoint VCs to achieve multicasting of layer 3 packets. 187 As the IETF is currently in the forefront of using wide area 188 multicasting this document's descriptions will often focus on IP 189 service model of RFC 1112. A final chapter will note the 190 multiprotocol application of the architecture. 192 This document avoids discussion of one highly non-trivial aspect of 193 using ATM - the specification of QoS for VCs being established in 194 response to higher layer needs. Research in this area is still very 195 formative [7], and so it is assumed that future documents will 196 clarify the mapping of QoS requirements to VC establishment. The 197 default at this time is that VCs are established with a request for 198 Unspecified Bit Rate (UBR) service, as typified by the IETF's use of 199 VCs for unicast IP, described in RFC 1755 [6]. 201 1.1 The Multicast Address Resolution Server (MARS). 203 The Multicast Address Resolution Server (MARS) is an extended analog 204 of the ATM ARP Server introduced in RFC 1577 [3]. It acts as a 205 registry, associating layer 3 multicast group identifiers with the 206 ATM interfaces representing the group's members. MARS messages 207 support the distribution of multicast group membership information 208 between MARS and endpoints (hosts or routers). Endpoint address 209 resolution entities query the MARS when a layer 3 address needs to be 210 resolved to the set of ATM endpoints making up the group at any one 211 time. Endpoints keep the MARS informed when they need to join or 212 leave particular layer 3 groups. To provide for asynchronous 213 notification of group membership changes the MARS manages a point to 214 multipoint VC out to all endpoints desiring multicast support 216 Valid arguments can be made for two different approaches to ATM level 217 multicasting of layer 3 packets - through meshes of point to 218 multipoint VCs, or ATM level multicast servers (MCS). The MARS 219 architecture allows either VC meshes or MCSs to be used on a per- 220 group basis. 222 1.2 The ATM level multicast Cluster. 224 Each MARS manages a 'cluster' of ATM-attached endpoints. A Cluster is 225 defined as 227 The set of ATM interfaces chosing to participate in direct ATM 228 connections to achieve multicasting of AAL_SDUs between 229 themselves. 231 In practice, a Cluster is the set of endpoints that choose to use the 232 same MARS to register their memberships and receive their updates 233 from. 235 By implication of this definition, traffic between interfaces 236 belonging to different Clusters passes through an inter-cluster 237 device. (In the IP world an inter-cluster device would be an IP 238 multicast router with logical interfaces into each Cluster.) This 239 document explicitly avoids specifying the nature of inter-cluster 240 (layer 3) routing protocols. 242 The mapping of clusters to other constrained sets of endpoints (such 243 as unicast Logical IP Subnets) is left to each network administrator. 244 However, for the purposes of conformance with this document network 245 administrators MUST ensure that each Logical IP Subnet (LIS) is 246 served by a separate MARS, creating a one-to-one mapping between 247 cluster and unicast LIS. IP multicast routers then interconnect each 248 LIS as they do with conventional subnets. (Relaxation of this 249 restriction MAY only occur after future research on the interaction 250 between existing layer 3 multicast routing protocols and unicast 251 subnet boundaries.) 253 The term 'Cluster Member' will be used in this document to refer to 254 an endpoint that is currently using a MARS for multicast support. 255 Thus potential scope of a cluster may be the entire membership of a 256 LIS, while the actual scope of a cluster depends on which endpoints 257 are actually cluster members at any given time. 259 1.3 Document overview. 261 This document assumes an understanding of concepts explained in 262 greater detail in RFC 1112, RFC 1577, UNI 3.0/3.1, and RFC 1755 [6]. 264 Section 2 provides an overview of IP multicast and what RFC 1112 265 required from Ethernet. 267 Section 3 describes in more detail the multicast support services 268 offered by UNI 3.0/3.1, and outlines the differences between VC 269 meshes and multicast servers (MCSs) as mechanisms for distributing 270 packets to multiple destinations. 272 Section 4 provides an overview of the MARS and its relationship to 273 ATM endpoints. This section also discusses the encapsulation and 274 structure of MARS control messages. 276 Section 5 substantially defines the entire cluster member endpoint 277 behaviour, on both receive and transmit sides. This includes both 278 normal operation and error recovery. 280 Section 6 summarises the required behaviour of a MARS. 282 Section 7 looks at how a multicast server (MCS) interacts with a 283 MARS. 285 Section 8 discusses how IP multicast routers may make novel use of 286 promiscuous and semi-promiscuous group joins. Also discussed is a 287 mechanism designed to reduce the amount of IGMP traffic issued by 288 routers. 290 Section 9 discusses how this document applies in the more general 291 (non-IP) case. 293 Section 10 summarises the key proposals, and identifies areas for 294 future research that are generated by this MARS architecture. 296 The appendices provide discussion on issues that arise out the 297 implementation of this document. Appendix A discusses MARS and 298 endpoint algorithms for parsing MARS messages. Appendix B describes 299 the particular problems introduced by the current IGMP paradigms, and 300 possible interim work-arounds. Appendix C discusses the 'cluster' 301 concept in further detail, while Appendix D briefly outlines an 302 algorithm for parsing TLV lists. Appendix E summarises various timer 303 values used in this document, and Appendix F provides example 304 pseudo-code for a MARS entity. 306 1.4 Conventions. 308 In this document the following coding and packet representation rules 309 are used: 311 All multi-octet parameters are encoded in big-endian form (i.e. 312 the most significant octet comes first). 314 In all multi-bit parameters bit numbering begins at 0 for the 315 least significant bit when stored in memory (i.e. the n'th bit has 316 weight of 2^n). 318 A bit that is 'set', 'on', or 'one' holds the value 1. 320 A bit that is 'reset', 'off', 'clear', or 'zero' holds the value 321 0. 323 2. Summary of the IP multicast service model. 325 Under IP version 4 (IPv4), addresses in the range between 224.0.0.0 326 and 239.255.255.255 (224.0.0.0/4) are termed 'Class D' or 'multicast 327 group' addresses. These abstractly represent all the IP hosts in the 328 Internet (or some constrained subset of the Internet) who have 329 decided to 'join' the specified group. 331 RFC1112 requires that a multicast-capable IP interface must support 332 the transmission of IP packets to an IP multicast group address, 333 whether or not the node considers itself a 'member' of that group. 334 Consequently, group membership is effectively irrelevant to the 335 transmit side of the link layer interfaces. When Ethernet is used as 336 the link layer (the example used in RFC1112), no address resolution 337 is required to transmit packets. An algorithmic mapping from IP 338 multicast address to Ethernet multicast address is performed locally 339 before the packet is sent out the local interface in the same 'send 340 and forget' manner as a unicast IP packet. 342 Joining and Leaving an IP multicast group is more explicit on the 343 receive side - with the primitives JoinLocalGroup and LeaveLocalGroup 344 affecting what groups the local link layer interface should accept 345 packets from. When the IP layer wants to receive packets from a 346 group, it issues JoinLocalGroup. When it no longer wants to receive 347 packets, it issues LeaveLocalGroup. A key point to note is that 348 changing state is a local issue, it has no affect on other hosts 349 attached to the Ethernet. 351 IGMP is defined in RFC 1112 to support IP multicast routers attached 352 to a given subnet. Hosts issue IGMP Report messages when they perform 353 a JoinLocalGroup, or in response to an IP multicast router sending an 354 IGMP Query. By periodically transmitting queries IP multicast routers 355 are able to identify what IP multicast groups have non-zero 356 membership on a given subnet. 358 A specific IP multicast address, 224.0.0.1, is allocated for the 359 transmission of IGMP Query messages. Host IP layers issue a 360 JoinLocalGroup for 224.0.0.1 when they intend to participate in IP 361 multicasting, and issue a LeaveLocalGroup for 224.0.0.1 when they've 362 ceased participating in IP multicasting. 364 Each host keeps a list of IP multicast groups it has been 365 JoinLocalGroup'd to. When a router issues an IGMP Query on 224.0.0.1 366 each host begins to send IGMP Reports for each group it is a member 367 of. IGMP Reports are sent to the group address, not 224.0.0.1, "so 368 that other members of the same group on the same network can overhear 369 the Report" and not bother sending one of their own. IP multicast 370 routers conclude that a group has no members on the subnet when IGMP 371 Queries no longer elict associated replies. 373 3. UNI 3.0/3.1 support for intra-cluster multicasting. 375 For the purposes of the MARS protocol, both UNI 3.0 and UNI 3.1 376 provide equivalent support for multicasting. Differences between UNI 377 3.0 and UNI 3.1 in required signalling elements are covered in RFC 378 1755. 380 This document will describe its operation in terms of 'generic' 381 functions that should be available to clients of a UNI 3.0/3.1 382 signalling entity in a given ATM endpoint. The ATM model broadly 383 describes an 'AAL User' as any entity that establishes and manages 384 VCs and underlying AAL services to exchange data. An IP over ATM 385 interface is a form of 'AAL User' (although the default LLC/SNAP 386 encapsulation mode specified in RFC1755 really requires that an 'LLC 387 entity' is the AAL User, which in turn supports the IP/ATM 388 interface). 390 The most fundamental limitations of UNI 3.0/3.1's multicast support 391 are: 393 Only point to multipoint, unidirectional VCs may be established. 395 Only the root (source) node of a given VC may add or remove leaf 396 nodes. 398 Leaf nodes are identified by their unicast ATM addresses. UNI 399 3.0/3.1 defines two ATM address formats - native E.164 and NSAP 400 (although it must be stressed that the NSAP address is so called 401 because it uses the NSAP format - an ATM endpoint is NOT a Network 402 layer termination point). In UNI 3.0/3.1 an 'ATM Number' is the 403 primary identification of an ATM endpoint, and it may use either 404 format. Under some circumstances an ATM endpoint must be identified 405 by both a native E.164 address (identifying the attachment point of a 406 private network to a public network), and an NSAP address ('ATM 407 Subaddress') identifying the final endpoint within the private 408 network. For the rest of this document the term will be used to mean 409 either a single 'ATM Number' or an 'ATM Number' combined with an 'ATM 410 Subaddress'. 412 3.1 VC meshes. 414 The most fundamental approach to intra-cluster multicasting is the 415 multicast VC mesh. Each source establishes its own independent point 416 to multipoint VC (a single multicast tree) to the set of leaf nodes 417 (destinations) that it has been told are members of the group it 418 wishes to send packets to. 420 Interfaces that are both senders and group members (leaf nodes) to a 421 given group will originate one point to multipoint VC, and terminate 422 one VC for every other active sender to the group. This criss- 423 crossing of VCs across the ATM network gives rise to the name 'VC 424 mesh'. 426 3.2 Multicast Servers. 428 An alternative model has each source establish a VC to an 429 intermediate node - the multicast server (MCS). The multicast server 430 itself establishes and manages a point to multipoint VC out to the 431 actual desired destinations. 433 The MCS reassembles AAL_SDUs arriving on all the incoming VCs, and 434 then queues them for transmission on its single outgoing point to 435 multipoint VC. (Reassembly of incoming AAL_SDUs is required at the 436 multicast server as AAL5 does not support cell level multiplexing of 437 different AAL_SDUs on a single outgoing VC.) 438 The leaf nodes of the multicast server's point to multipoint VC must 439 be established prior to packet transmission, and the multicast server 440 requires an external mechanism to identify them. A side-effect of 441 this method is that ATM interfaces that are both sources and group 442 members will receive copies of their own packets back from the MCS 443 (An alternative method is for the multicast server to explicitly 444 retransmit packets on individual VCs between itself and group 445 members. A benefit of this second approach is that the multicast 446 server can ensure that sources do not receive copies of their own 447 packets.) 449 The simplest MCS pays no attention to the contents of each AAL_SDU. 450 It is purely an AAL/ATM level device. More complex MCS architectures 451 (where a single endpoint serves multiple layer 3 groups) are 452 possible, but are beyond the scope of this document. More detailed 453 discussion is provided in section 7. 455 3.3 Tradeoffs. 457 Arguments over the relative merits of VC meshes and multicast servers 458 have raged for some time. Ultimately the choice depends on the 459 relative trade-offs a system administrator must make between 460 throughput, latency, congestion, and resource consumption. Even 461 criteria such as latency can mean different things to different 462 people - is it end to end packet time, or the time it takes for a 463 group to settle after a membership change? The final choice depends 464 on the characteristics of the applications generating the multicast 465 traffic. 467 If we focussed on the data path we might prefer the VC mesh because 468 it lacks the obvious single congestion point of an MCS. Throughput 469 is likely to be higher, and end to end latency lower, because the 470 mesh lacks the intermediate AAL_SDU reassembly that must occur in 471 MCSs. The underlying ATM signalling system also has greater 472 opportunity to ensure optimal branching points at ATM switches along 473 the multicast trees originating on each source. 475 However, resource consumption will be higher. Every group member's 476 ATM interface must terminate a VC per sender (consuming on-board 477 memory for state information, instance of an AAL service, and 478 buffering in accordance with the vendors particular architecture). On 479 the contrary, with a multicast server only 2 VCs (one out, one in) 480 are required, independent of the number of senders. The allocation of 481 VC related resources is also lower within the ATM cloud when using a 482 multicast server. These points may be considered to have merit in 483 environments where VCs across the UNI or within the ATM cloud are 484 valuable (e.g. the ATM provider charges on a per VC basis), or AAL 485 contexts are limited in the ATM interfaces of endpoints. 487 If we focus on the signalling load then MCSs have the advantage when 488 faced with dynamic sets of receivers. Every time the membership of a 489 multicast group changes (a leaf node needs to be added or dropped), 490 only a single point to multipoint VC needs to be modified when using 491 an MCS. This generates a single signalling event across the MCS's 492 UNI. However, when membership change occurs in a VC mesh, signalling 493 events occur at the UNIs of every traffic source - the transient 494 signalling load scales with the number of sources. This has obvious 495 ramifications if you define latency as the time for a group's 496 connectivity to stabilise after change (especially as the number of 497 senders increases). 499 Finally, as noted above, MCSs introduce a 'reflected packet' problem, 500 which requires additional per-AAL_SDU information to be carried in 501 order for layer 3 sources to detect their own AAL_SDUs coming back. 503 The MARS architecture allows system administrators to utilize either 504 approach on a group by group basis. 506 3.4 Interaction with local UNI 3.0/3.1 signalling entity. 508 The following generic signalling functions are presumed to be 509 available to local AAL Users: 511 L_CALL_RQ - Establish a unicast VC to a specific endpoint. 512 L_MULTI_RQ - Establish multicast VC to a specific endpoint. 513 L_MULTI_ADD - Add new leaf node to previously established VC. 514 L_MULTI_DROP - Remove specific leaf node from established VC. 515 L_RELEASE - Release unicast VC, or all Leaves of a multicast VC. 517 The signalling exchanges and local information passed between AAL 518 User and UNI 3.0/3.1 signalling entity with these functions are 519 outside the scope of this document. 521 The following indications are assumed to be available to AAL Users, 522 generated by by the local UNI 3.0/3.1 signalling entity: 524 L_ACK - Succesful completion of a local request. 525 L_REMOTE_CALL - A new VC has been established to the AAL User. 526 ERR_L_RQFAILED - A remote ATM endpoint rejected an L_CALL_RQ, 527 L_MULTI_RQ, or L_MULTI_ADD. 528 ERR_L_DROP - A remote ATM endpoint dropped off an existing VC. 529 ERR_L_RELEASE - An existing VC was terminated. 531 The signalling exchanges and local information passed between AAL 532 User and UNI 3.0/3.1 signalling entity with these functions are 533 outside the scope of this document. 535 4. Overview of the MARS. 537 The MARS may reside within any ATM endpoint that is directly 538 addressable by the endpoints it is serving. Endpoints wishing to join 539 a multicast cluster must be configured with the ATM address of the 540 node on which the cluster's MARS resides. (Section 5.4 describes how 541 backup MARSs may be added to support the activities of a cluster. 542 References to 'the MARS' in following sections will be assumed to 543 mean the acting MARS for the cluster.) 545 4.1 Architecture. 547 Architecturally the MARS is an evolution of the RFC 1577 ARP Server. 548 Whilst the ARP Server keeps a table of {IP,ATM} address pairs for all 549 IP endpoints in an LIS, the MARS keeps extended tables of {layer 3 550 address, ATM.1, ATM.2, ..... ATM.n} mappings. It can either be 551 configured with certain mappings, or dynamically 'learn' mappings. 552 The format of the {layer 3 address} field is generally not 553 interpreted by the MARS. 555 A single ATM node may support multiple logical MARSs, each of which 556 support a separate cluster. The restriction is that each MARS has a 557 unique ATM address (e.g. a different SEL field in the NSAP address of 558 the node on which the multiple MARSs reside). By definition a single 559 instance of a MARS may not support more than one cluster. 561 The MARS distributes group membership update information to cluster 562 members over a point to multipoint VC known as the ClusterControlVC. 563 Additionally, when Multicast Servers (MCSs) are being used it also 564 establishes a separate point to multipoint VC out to registered MCSs, 565 known as the ServerControlVC. All cluster members are leaf nodes of 566 ClusterControlVC. All registered multicast servers are leaf nodes of 567 ServerControlVC (described further in section 6). 569 The MARS does NOT take part in the actual multicasting of layer 3 570 data packets. 572 4.2 Control message format. 574 By default all MARS control messages MUST be LLC/SNAP encapsulated 575 using the following codepoints: 577 [0xAA-AA-03][0x00-00-5E][0x00-03][MARS control message] 578 (LLC) (OUI) (PID) 580 (This is a PID from the IANA OUI.) 582 MARS control messages are made up of 4 major components: 584 [Fixed header][Mandatory fields][Addresses][Supplementary TLVs] 586 [Fixed header] contains fields indicating the operation being 587 performed and the layer 3 protocol being referred to (e.g IPv4, IPv6, 588 AppleTalk, etc). The fixed header also carries checksum information, 589 and hooks to allow this basic control message structure to be re-used 590 by other query/response protocols. 592 The [Mandatory fields] section carries fixed width parameters that 593 depend on the operation type indicated in [Fixed header]. 595 The following [Addresses] area carries variable length fields for 596 source and target addresses - both hardware (e.g. ATM) and layer 3 597 (e.g. IPv4). These provide the fundamental information that the 598 registrations, queries, and updates use and operate on. For the MARS 599 protocol fields in [Fixed header] indicate how to interpret the 600 contents of [Addresses]. 602 [Supplementary TLVs] represents an optional list of TLV (type, 603 length, value) encoded information elements that may be appended to 604 provide supplementary information. This feature is described in 605 further detail in section 10. 607 MARS messages contain variable length address fields. In all cases 608 null addresses SHALL be encoded as zero length, and have no space 609 allocated in the message. 611 (Unique LLC/SNAP encapsulation of MARS control messages means MARS 612 and ARP Server functionality may be implemented within a common 613 entity, and share a client-server VC, if the implementor so chooses. 614 Note that the LLC/SNAP codepoint for MARS is different to the 615 codepoint used for ATMARP.) 617 4.3 Fixed header fields in MARS control messages. 619 The [Fixed header] has the following format: 621 Data: 622 mar$hrd 16 bits Hardware type. 623 mar$pro 56 bits Protocol Identification. 624 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 625 mar$chksum 16 bits Checksum across entire MARS message. 626 mar$extoff 16 bits Extensions Offset. 627 mar$op 16 bits Operation code. 628 mar$shtl 8 bits Type & length of source ATM number. (r) 629 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 631 mar$shtl and mar$sstl provide information regarding the source's 632 hardware (ATM) address. In the MARS protocol these fields are always 633 present, as every MARS message carries a non-null source ATM address. 634 In all cases the source ATM address is the first variable length 635 field in the [Addresses] section. 637 The other fields in [Fixed header] are described in the following 638 subsections. 640 4.3.1 Hardware type. 642 mar$hrd defines the type of 'hardware' addresses being carried. When 643 mar$hrd = 0x13 the addresses are ATM addresses. Interpretation of the 644 ATM number and subaddress fields when mar$hrd != 0x13 is for future 645 definition. The remainder of this document assumes that mar$hrd = 646 0x13. 648 4.3.2 Protocol type. 650 The mar$pro field is made of up of two subfields: 652 mar$pro.type 16 bits Protocol type. 653 mar$pro.snap 40 bits Optional SNAP extension to protocol type. 655 The mar$pro.type field is a 16 bit unsigned integer representing the 656 following number space: 658 0x0000 to 0x00FF Protocols defined by the equivalent NLPIDs. 659 0x0100 to 0x03FF Reserved for future use by the IETF. 660 0x0400 to 0x04FF Allocated for use by the ATM Forum. 661 0x0500 to 0x05FF Experimental/Local use. 662 0x0600 to 0xFFFF Protocols defined by the equivalent Ethertypes. 664 (based on the observations that valid Ethertypes are never smaller 665 than 0x600, and NLPIDs never larger than 0xFF.) 667 The NLPID value of 0x80 is used to indicate a SNAP encoded extension 668 is being used to encode the protocol type. When mar$pro.type == 0x80 669 the SNAP extension is encoded in the mar$pro.snap field. This is 670 termed the 'long form' protocol ID. 672 If mar$pro.type != 0x80 then the mar$pro.snap field MUST be zero on 673 transmit and ignored on receive. The mar$pro.type field itself 674 identifies the protocol being referred to. This is termed the 'short 675 form' protocol ID. 677 In all cases, where a protocol has an assigned number in the 678 mar$pro.type space (excluding 0x80) the short form MUST be used when 679 transmitting MARS messages. Additionally, where a protocol has valid 680 short and long forms of identification, receivers MAY choose to 681 recognise the long form. 683 mar$pro.type values other than 0x80 MAY have 'long forms' defined in 684 future documents. 686 For the remainder of this document references to mar$pro SHALL be 687 interpreted to mean mar$pro.type, or mar$pro.type in combination with 688 mar$pro.snap as appropriate. 690 The use of different protocol types is described further in section 691 9. 693 4.3.3 Checksum. 695 The mar$chksum field carries a standard IP checksum calculated across 696 the entire MARS control message (excluding the LLC/SNAP header). The 697 field is set to zero before performing the checksum calculation. 699 As the entire LLC/SNAP encapsulated MARS message is protected by the 700 32 bit CRC of the AAL5 transport, implementors MAY choose to ignore 701 the checksum facility. If no checksum is calculated these bits MUST 702 be reset before transmission. If no checksum is performed on 703 reception, this field MUST be ignored. If a receiver is capable of 704 validating a checksum it MUST only perform the validation when the 705 received mar$chksum field is non-zero. Messages arriving with 706 mar$chksum of 0 are always considered valid. 708 4.3.4 Extensions Offset. 710 The mar$extoff field identifies the existence and location of an 711 optional supplementary parameters list. Its use is described in 712 section 10. 714 4.3.5 Operation code. 716 The mar$op field is further subdivided into two 8 bit fields - 717 mar$op.version (leading octet) and mar$op.type (trailing octet). 718 Together they indicate the nature of the control message, and the 719 context within which its [Mandatory fields], [Addresses], and 720 [Supplementary TLVs] should be interpreted. 722 mar$op.version 723 0 MARS protocol defined in this document. 724 0x01 - 0xEF Reserved for future use by the IETF. 725 0xF0 - 0xFE Allocated for use by the ATM Forum. 726 0xFF Experimental/Local use. 728 mar$op.type 729 Value indicates operation being performed, within context of 730 the control protocol version indicated by mar$op.version. 732 For the rest of this document references to the mar$op value SHALL be 733 taken to mean mar$op.type, with mar$op.version = 0x00. The values 734 used in this document are summarised in section 11. 736 (Note this number space is independent of the ATMARP operation code 737 number space.) 739 4.3.6 Reserved. 741 mar$hdrrsv may be subdivided and assigned specific meanings for other 742 control protocols indicated by mar$op.version != 0. 744 5. Endpoint (MARS client) interface behaviour. 746 An endpoint is best thought of as a 'shim' or 'convergence' layer, 747 sitting between a layer 3 protocol's link layer interface and the 748 underlying UNI 3.0/3.1 service. An endpoint in this context can exist 749 in a host or a router - any entity that requires a generic 'layer 3 750 over ATM' interface to support layer 3 multicast. It is broken into 751 two key subsections - one for the transmit side, and one for the 752 receive side. 754 Multiple logical ATM interfaces may be supported by a single physical 755 ATM interface (for example, using different SEL values in the NSAP 756 formatted address assigned to the physical ATM interface). Therefore 757 implementors MUST allow for multiple independent 'layer 3 over ATM' 758 interfaces too, each with its own configured MARS (or table of MARSs, 759 as discussed in section 5.4), and ability to be attached to the same 760 or different clusters. 762 The initial signalling path between a MARS client (managing an 763 endpoint) and its associated MARS is a transient point to point, 764 bidirectional VC. This VC is established by the MARS client, and is 765 used to send queries to, and receive replies from, the MARS. It has 766 an associated idle timer, and is dismantled if not used for a 767 configurable period of time. The minimum suggested value for this 768 time is 1 minute, and the RECOMMENDED default is 20 minutes. (Where 769 the MARS and ARP Server are co-resident, this VC may be used for both 770 ATM ARP traffic and MARS control traffic.) 772 The remaining signalling path is ClusterControlVC, to which the MARS 773 client is added as a leaf node when it registers (described in 774 section 5.2.3). 776 The majority of this document covers the distribution of information 777 allowing endpoints to establish and manage outgoing point to 778 multipoint VCs - the forwarding paths for multicast traffic to 779 particular multicast groups. The actual format of the AAL_SDUs sent 780 on these VCs is almost completely outside the scope of this 781 specification. However, endpoints are not expected to know whether 782 their forwarding path leads directly to a multicast group's members 783 or to an MCS (described in section 3). This requires additional per- 784 packet encapsulation (described in section 5.5) to aid in the the 785 detection of reflected AAL_SDUs. 787 5.1 Transmit side behaviour. 789 The following description will often be in terms of an IPv4/ATM 790 interface that is capable of transmitting packets to a Class D 791 address at any time, without prior warning. It should be trivial for 792 an implementor to generalise this behaviour to the requirements of 793 another layer 3 data protocol. 795 When a local Layer 3 entity passes down a packet for transmission, 796 the endpoint first ascertains whether an outbound path to the 797 destination multicast group already exists. If it does not, the MARS 798 is queried for a set of ATM endpoints that represent an appropriate 799 forwarding path. (The ATM endpoints may represent the actual group 800 members within the cluster, or a set of one or more MCSs. The 801 endpoint does not distinguish between either case. Section 6.2 802 describes the MARS behaviour that leads to MCSs being supplied as the 803 forwarding path for a multicast group.) 805 The query is executed by issuing a MARS_REQUEST. The reply from the 806 MARS may take one of two forms: 808 MARS_MULTI - Sequence of MARS_MULTI messages returning the set of 809 ATM endpoints that are to be leaf nodes of an 810 outgoing point to multipoint VC (the forwarding 811 path). 813 MARS_NAK - No mapping found, group is empty. 815 The formats of these messages are described in section 5.1.2. 817 Outgoing VCs are established with a request for Unspecified Bit Rate 818 (UBR) service, as typified by the IETF's use of VCs for unicast IP, 819 described in RFC 1755 [6]. Future documents may vary this approach 820 and allow the specification of different ATM traffic parameters from 821 locally configured information or parameters obtained through some 822 external means. 824 5.1.1 Retrieving Group Membership from the MARS. 826 If the MARS had no mapping for the desired Class D address a MARS_NAK 827 will be returned. In this case the IP packet MUST be discarded 828 silently. If a match is found in the MARS's tables it proceeds to 829 return addresses ATM.1 through ATM.n in a sequence of one or more 830 MARS_MULTIs. A simple mechanism is used to detect and recover from 831 loss of MARS_MULTI messages. 833 (If the client learns that there is no other group member in the 834 cluster - the MARS returns a MARS_NAK or returns a MARS_MULTI with 835 the client as the only member - it MUST delay sending out a new 836 MARS_REQUEST for that group for a period no less than 5 seconds and 837 no more than 10 seconds.) 839 Each MARS_MULTI carries a boolean field x, and a 15 bit integer field 840 y - expressed as MARS_MULTI(x,y). Field y acts as a sequence number, 841 starting at 1 and incrementing for each MARS_MULTI sent. Field x 842 acts as an 'end of reply' marker. When x == 1 the MARS response is 843 considered complete. 845 In addition, each MARS_MULTI may carry multiple ATM addresses from 846 the set {ATM.1, ATM.2, .... ATM.n}. A MARS MUST minimise the number 847 of MARS_MULTIs transmitted by placing as many group member's 848 addresses in a single MARS_MULTI as possible. The limit on the length 849 of an individual MARS_MULTI message MUST be the MTU of the underlying 850 VC. 852 For example, assume n ATM addresses must be returned, each MARS_MULTI 853 is limited to only p ATM addresses, and p << n. This would require a 854 sequence of k MARS_MULTI messages (where k = (n/p)+1, using integer 855 arithmetic), transmitted as follows: 857 MARS_MULTI(0,1) carries back {ATM.1 ... ATM.p} 858 MARS_MULTI(0,2) carries back {ATM.(p+1) ... ATM.(2p)} 859 [.......] 860 MARS_MULTI(1,k) carries back { ... ATM.n} 862 If k == 1 then only MARS_MULTI(1,1) is sent. 864 Typical failure mode will be losing one or more of MARS_MULTI(0,1) 865 through MARS_MULTI(0,k-1). This is detected when y jumps by more than 866 one between consecutive MARS_MULTI's. An alternative failure mode is 867 losing MARS_MULTI(1,k). A timer MUST be implemented to flag the 868 failure of the last MARS_MULTI to arrive. A default value of 10 869 seconds is RECOMMENDED. 871 If a 'sequence jump' is detected, the host MUST wait for the 872 MARS_MULTI(1,k), discard all results, and repeat the MARS_REQUEST. 874 If a timeout occurs, the host MUST discard all results, and repeat 875 the MARS_REQUEST. 877 A final failure mode involves the MARS Sequence Number (described in 878 section 5.1.4.2 and carried in each part of a multi-part MARS_MULTI). 879 If its value changes during the reception of a multi-part MARS_MULTI 880 the host MUST wait for the MARS_MULTI(1,k), discard all results, and 881 repeat the MARS_REQUEST. 883 (Corruption of cell contents will lead to loss of a MARS_MULTI 884 through AAL5 CPCS_PDU reassembly failure, which will be detected 885 through the mechanisms described above.) 887 If the MARS is managing a cluster of endpoints spread across 888 different but directly accessible ATM networks it will not be able to 889 return all the group members in a single MARS_MULTI. The MARS_MULTI 890 message format allows for either E.164, ISO NSAP, or (E.164 + NSAP) 891 to be returned as ATM addresses. However, each MARS_MULTI message may 892 only return ATM addresses of the same type and length. The returned 893 addresses MUST be grouped according to type (E.164, ISO NSAP, or 894 both) and returned in a sequence of separate MARS_MULTI parts. 896 5.1.2 MARS_REQUEST, MARS_MULTI, and MARS_NAK messages. 898 MARS_REQUEST is shown below. It is indicated by an 'operation type 899 value' (mar$op) of 1. 901 The multicast address being resolved is placed into the the target 902 protocol address field (mar$tpa), and the target hardware address is 903 set to null (mar$thtl and mar$tstl both zero). 905 In IPv4 environments the protocol type (mar$pro) is 0x800 and the 906 target protocol address length (mar$tpln) MUST be set to 4. The 907 source fields MUST contain the ATM number and subaddress of the 908 client issuing the MARS_REQUEST (the subaddress MAY be null). 910 Data: 911 mar$hrd 16 bits Hardware type. 912 mar$pro 56 bits Protocol Identification. 913 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 914 mar$chksum 16 bits Checksum across entire MARS message. 915 mar$extoff 16 bits Extensions Offset. 916 mar$op 16 bits Operation code (MARS_REQUEST = 1) 917 mar$shtl 8 bits Type & length of source ATM number. (r) 918 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 919 mar$spln 8 bits Length of source protocol address (s) 920 mar$thtl 8 bits Type & length of target ATM number (x) 921 mar$tstl 8 bits Type & length of target ATM subaddress (y) 922 mar$tpln 8 bits Length of target group address (z) 923 mar$pad 64 bits Padding (aligns mar$sha with MARS_MULTI). 924 mar$sha qoctets source ATM number 925 mar$ssa roctets source ATM subaddress 926 mar$spa soctets source protocol address 927 mar$tpa zoctets target multicast group address 928 mar$tha xoctets target ATM number 929 mar$tsa yoctets target ATM subaddress 931 Following the RFC1577 approach, the mar$shtl, mar$sstl, mar$thtl and 932 mar$tstl fields are coded as follows: 934 7 6 5 4 3 2 1 0 935 +-+-+-+-+-+-+-+-+ 936 |0|x| length | 937 +-+-+-+-+-+-+-+-+ 939 The most significant bit is reserved and MUST be set to zero. The 940 second most significant bit (x) is a flag indicating whether the ATM 941 address being referred to is in: 943 - ATM Forum NSAPA format (x = 0). 944 - Native E.164 format (x = 1). 946 The bottom 6 bits is an unsigned integer value indicating the length 947 of the associated ATM address in octets. If this value is zero the 948 flag x is ignored. 950 The mar$spln and mar$tpln fields are unsigned 8 bit integers, giving 951 the length in octets of the source and target protocol address fields 952 respectively. 954 MARS packets use true variable length fields. A null (non-existant) 955 address MUST be coded as zero length, and no space allocated for it 956 in the message body. 958 MARS_NAK is the MARS_REQUEST returned with operation type value of 6. 959 All other fields are left unchanged from the MARS_REQUEST (e.g. do 960 not transpose the source and target information. In all cases MARS 961 clients use the source address fields to identify their own messages 962 coming back). 964 The MARS_MULTI message is identified by an mar$op value of 2. The 965 message format is: 967 Data: 969 mar$hrd 16 bits Hardware type. 970 mar$pro 56 bits Protocol Identification. 971 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 972 mar$chksum 16 bits Checksum across entire MARS message. 973 mar$extoff 16 bits Extensions Offset. 974 mar$op 16 bits Operation code (MARS_MULTI = 2). 975 mar$shtl 8 bits Type & length of source ATM number. (r) 976 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 977 mar$spln 8 bits Length of source protocol address (s) 978 mar$thtl 8 bits Type & length of target ATM number (x) 979 mar$tstl 8 bits Type & length of target ATM subaddress (y) 980 mar$tpln 8 bits Length of target group address (z) 981 mar$tnum 16 bits Number of target ATM addresses returned (N) 982 mar$seqxy 16 bits Boolean flag x and sequence number y. 983 mar$msn 32 bits MARS Sequence Number. 984 mar$sha qoctets source ATM number 985 mar$ssa roctets source ATM subaddress 986 mar$spa soctets source protocol address 987 mar$tpa zoctets target multicast group address 988 mar$tha.1 xoctets target ATM number 1 989 mar$tsa.1 yoctets target ATM subaddress 1 990 mar$tha.2 xoctets target ATM number 2 991 mar$tsa.2 yoctets target ATM subaddress 2 992 [.......] 993 mar$tha.N xoctets target ATM number N 994 mar$tsa.N yoctets target ATM subaddress N 996 The source protocol and ATM address fields are copied directly from 997 the MARS_REQUEST that this MARS_MULTI is in response to (not the MARS 998 itself). 1000 mar$seqxy is coded with flag x in the leading bit, and sequence 1001 number y coded as an unsigned integer in the remaining 15 bits. 1003 | 1st octet | 2nd octet | 1004 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 1005 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1006 |x| y | 1007 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1009 mar$tnum is an unsigned integer indicating how many pairs of 1010 {mar$tha,mar$tsa} (i.e. how many group member's ATM addresses) are 1011 present in the message. mar$msn is an unsigned 32 bit number filled 1012 in by the MARS before transmitting each MARS_MULTI. Its use is 1013 described further in section 5.1.4. 1015 As an example, assume we have a multicast cluster using 4 byte 1016 protocol addresses, 20 byte ATM numbers, and 0 byte ATM subaddresses. 1018 For n group members in a single MARS_MULTI we require a (48 + 20n) 1019 byte message. If we assume the default MTU of 9180 bytes, we can 1020 return a maximum of 456 group member's addresses in a single 1021 MARS_MULTI. 1023 5.1.3 Establishing the outgoing multipoint VC. 1025 Following the completion of the MARS_MULTI reply the endpoint may 1026 establish a new point to multipoint VC, or reuse an existing one. 1028 If establishing a new VC, an L_MULTI_RQ is issued for ATM.1, followed 1029 by an L_MULTI_ADD for every member of the set {ATM.2, ....ATM.n} 1030 (assuming the set is non-null). The packet is then transmitted over 1031 the newly created VC just as it would be for a unicast VC. 1033 After transmitting the packet, the local interface holds the VC open 1034 and marks it as the active path out of the host for any subsequent IP 1035 packets being sent to that Class D address. 1037 When establishing a new multicast VC it is possible that one or more 1038 L_MULTI_RQ or L_MULTI_ADD may fail. The UNI 3.0/3.1 failure cause 1039 must be returned in the ERR_L_RQFAILED signal from the local 1040 signalling entity to the AAL User. If the failure cause is not 49 1041 (Quality of Service unavailable), 51 (user cell rate not available - 1042 UNI 3.0), 37 (user cell rate not available - UNI 3.1), or 41 1043 (Temporary failure), the endpoint's ATM address is dropped from the 1044 set {ATM.1, ATM.2, ..., ATM.n} returned by the MARS. Otherwise, the 1045 L_MULTI_RQ or L_MULTI_ADD should be reissued after a random delay of 1046 5 to 10 seconds. If the request fails again, another request should 1047 be issued after twice the previous delay has elapsed. This process 1048 should be continued until the call succeeds or the multipoint VC gets 1049 released. 1051 If the initial L_MULTI_RQ fails for ATM.1, and n is greater than 1 1052 (i.e. the returned set of ATM addresses contains 2 or more addresses) 1053 a new L_MULTI_RQ should be immediately issued for the next ATM 1054 address in the set. This procedure is repeated until an L_MULTI_RQ 1055 succeeds, as no L_MULTI_ADDs may be issued until an initial outgoing 1056 VC is established. 1058 Each ATM address for which an L_MULTI_RQ failed with cause 49, 51, 1059 37, or 41 MUST be tagged rather than deleted. An L_MULTI_ADD is 1060 issued for these tagged addresses using the random delay procedure 1061 outlined above. 1063 The VC MAY be considered 'up' before failed L_MULTI_ADDs have been 1064 successfully re-issued. An endpoint MAY implement a concurrent 1065 mechanism that allows data to start flowing out the new VC even while 1066 failed L_MULTI_ADDs are being re-tried. (The alternative of waiting 1067 for each leaf node to accept the connection could lead to significant 1068 delays in transmitting the first packet.) 1070 Each VC MUST have a configurable inactivity timer associated with it. 1071 If the timer expires, an L_RELEASE is issued for that VC, and the 1072 Class D address is no longer considered to have an active path out of 1073 the local host. The timer SHOULD be no less than 1 minute, and a 1074 default of 20 minutes is RECOMMENDED. Choice of specific timer 1075 periods is beyond the scope of this document. 1077 VC consumption may also be reduced by endpoints noting when a new 1078 group's set of {ATM.1, ....ATM.n} matches that of a pre-existing VC 1079 out to another group. With careful local management, and assuming the 1080 QoS of the existing VC is sufficient for both groups, a new pt to mpt 1081 VC may not be necessary. Under certain circumstances endpoints may 1082 decide that it is sufficient to re-use an existing VC whose set of 1083 leaf nodes is a superset of the new group's membership (in which case 1084 some endpoints will receive multicast traffic for a layer 3 group 1085 they haven't joined, and must filter them above the ATM interface). 1086 Algorithms for performing this type of optimization are not discussed 1087 here, and are not required for conformance with this document. 1089 5.1.4 Tracking subsequent group updates. 1091 Once a new VC has been established, the transmit side of the cluster 1092 member's interface needs to monitor subsequent group changes - adding 1093 or dropping leaf nodes as appropriate. This is achieved by watching 1094 for MARS_JOIN and MARS_LEAVE messages from the MARS itself. These 1095 messages are described in detail in section 5.2 - at this point it is 1096 sufficient to note that they carry: 1098 - The ATM address of a node joining or leaving a group. 1099 - The layer 3 address of the group(s) being joined or left. 1100 - A Cluster Sequence Number (CSN) from the MARS. 1102 MARS_JOIN and MARS_LEAVE messages arrive at each cluster member 1103 across ClusterControlVC. MARS_JOIN or MARS_LEAVE messages that simply 1104 confirm information already held by the cluster member are used to 1105 track the Cluster Sequence Number, but are otherwise ignored. 1107 5.1.4.1 Updating the active VCs. 1109 If a MARS_JOIN is seen that refers to (or encompasses) a group for 1110 which the transmit side already has a VC open, the new member's ATM 1111 address is extracted and an L_MULTI_ADD issued locally. This ensures 1112 that endpoints already sending to a given group will immediately add 1113 the new member to their list of recipients. 1115 If a MARS_LEAVE is seen that refers to (or encompasses) a group for 1116 which the transmit side already has a VC open, the old member's ATM 1117 address is extracted and an L_MULTI_DROP issued locally. This ensures 1118 that endpoints already sending to a given group will immediately drop 1119 the old member from their list of recipients. When the last leaf of a 1120 VC is dropped, the VC is closed completely and the affected group no 1121 longer has a path out of the local endpoint (the next outbound packet 1122 to that group's address will trigger the creation of a new VC, as 1123 described in sections 5.1.1 to 5.1.3). 1125 The transmit side of the interface MUST NOT shut down an active VC to 1126 a group for which the receive side has just executed a 1127 LeaveLocalGroup. (This behaviour is consistent with the model of 1128 hosts transmitting to groups regardless of their own membership 1129 status.) 1131 If a MARS_JOIN or MARS_LEAVE arrives with mar$pnum == 0 it carries no 1132 pairs, and is only used for tracking the CSN. 1134 5.1.4.2 Tracking the Cluster Sequence Number. 1136 It is important that endpoints do not miss group membership updates 1137 issued by the MARS over ClusterControlVC. However, this will happen 1138 from time to time. The Cluster Sequence Number is carried as an 1139 unsigned 32 bit value in the mar$msn field of many MARS messages 1140 (except for MARS_REQUEST and MARS_NAK). It increments once for every 1141 transmission the MARS makes on ClusterControlVC, regardless of 1142 whether the transmission represents a change in the MARS database or 1143 not. By tracking this counter, cluster members can determine whether 1144 they have missed a previous message on ClusterControlVC, and possibly 1145 a membership change. This is then used to trigger revalidation 1146 (described in section 5.1.5). 1148 The current CSN is copied into the mar$msn field of MARS messages 1149 being sent to cluster members, whether out ClusterControlVC or on an 1150 point to point VC. 1152 Calculations on the sequence numbers MUST be performed as unsigned 32 1153 bit arithmetic. 1155 Every cluster member keeps its own 32 bit Host Sequence Number (HSN) 1156 to track the MARS's sequence number. Whenever a message is received 1157 that carries an mar$msn field the following processing is performed: 1159 Seq.diff = mar$msn - HSN 1161 mar$msn -> HSN 1162 {...process MARS message as appropriate...} 1163 if ((Seq.diff != 1) && (Seq.diff != 0)) 1164 then {...revalidate group membership information...} 1166 The basic result is that the cluster member attempts to keep locked 1167 in step with membership changes noted by the MARS. If it ever detects 1168 that a membership change occurred (in any group) without it noticing, 1169 it re-validates the membership of all groups it currently has 1170 multicast VCs open to. 1172 The mar$msn value in an individual MARS_MULTI is not used to update 1173 the HSN until all parts of the MARS_MULTI (if more than 1) have 1174 arrived. (If the mar$msn changes the MARS_MULTI is discarded, as 1175 described in section 5.1.1.) 1177 The MARS is free to choose an initial value of CSN. When a new 1178 cluster member starts up it should initialise HSN to zero. When the 1179 cluster member sends the MARS_JOIN to register (described later), the 1180 HSN will be correctly updated to the current CSN value when the 1181 endpoint receives the copy of its MARS_JOIN back from the MARS. 1183 5.1.5 Revalidating a VC's leaf nodes. 1185 Certain events may inform a cluster member that it has incorrect 1186 information about the sets of leaf nodes it should be sending to. If 1187 an error occurs on a VC associated with a particular group, the 1188 cluster member initiates revalidation procedures for that specific 1189 group. If a jump is detected in the Cluster Sequence Number, this 1190 initiates revalidation of all groups to which the cluster member 1191 currently has open point to multipoint VCs. 1193 Each open and active multipoint VC has a flag associated with it 1194 called 'VC_revalidate'. This flag is checked everytime a packet is 1195 queued for transmission on that VC. If the flag is false, the packet 1196 is transmitted and no further action is required. 1198 However, if the VC_revalidate flag is true then the packet is 1199 transmitted and a new sequence of events is started locally. 1201 Revalidation begins with re-issuing a MARS_REQUEST for the group 1202 being revalidated. The returned set of members {NewATM.1, NewATM.2, 1203 .... NewATM.n} is compared with the set already held locally. 1204 L_MULTI_DROPs are issued on the group's VC for each node that appears 1205 in the original set of members but not in the revalidated set of 1206 members. L_MULTI_ADDs are issued on the group's VC for each node that 1207 appears in the revalidated set of members but not in the original set 1208 of members. The VC_revalidate flag is reset when revalidation 1209 concludes for the given group. Implementation specific mechanisms 1210 will be needed to flag the 'revalidation in progress' state. 1212 The key difference between constructing a VC (section 5.1.3) and 1213 revalidating a VC is that packet transmission continues on the open 1214 VC while it is being revalidated. This minimises the disruption to 1215 existing traffic. 1217 The algorithm for initiating revalidation is: 1219 - When a packet arrives for transmission on a given group, 1220 the groups membership is revalidated if VC_revalidate == TRUE. 1221 Revalidation resets VC_revalidate. 1222 - When an event occurs that demands revalidation, every 1223 group has its VC_revalidate flag set TRUE at a random time 1224 between 1 and 10 seconds. 1226 Benefit: Revalidation of active groups occurs quickly, and 1227 essentially idle groups are revalidated as needed. Randomly 1228 distributed setting of VC_revalidate flag improves chances of 1229 staggered revalidation requests from senders when a sequence number 1230 jump is detected. 1232 5.1.5.1 When leaf node drops itself. 1234 During the life of a multipoint VC an ERR_L_DROP may be received 1235 indicating that a leaf node has terminated its participation at the 1236 ATM level. The ATM endpoint associated with the ERR_L_DROP MUST be 1237 removed from the locally held set {ATM.1, ATM.2, .... ATM.n} 1238 associated with the VC. 1240 After a random period of time between 1 and 10 seconds the 1241 VC_revalidate flag associated with that VC MUST be set true. 1243 If an ERR_L_RELEASE is received then the entire set {ATM.1, ATM.2, 1244 .... ATM.n} is cleared and the VC is considered to be completely shut 1245 down. Further packet transmission to the group served by this VC will 1246 result in a new VC being established as described in section 5.1.3. 1248 5.1.5.2 When a jump is detected in the CSN. 1250 Section 5.1.4.2 describes how a CSN jump is detected. If a CSN jump 1251 is detected upon receipt of a MARS_JOIN or a MARS_LEAVE then every 1252 outgoing multicast VC MUST have its VC_revalidate flag set true at 1253 some random interval between 1 and 10 seconds from when the CSN jump 1254 was detected. 1256 The only exception to this rule is if a sequence number jump is 1257 detected during the establishment of a new group's VC (i.e. a 1258 MARS_MULTI reply was correctly received, but its mar$msn indicated 1259 that some previous MARS traffic had been missed on ClusterControlVC). 1261 In this case every open VC, EXCEPT the one just established, MUST 1262 have its VC_revalidate flag set true at some random interval between 1263 1 and 10 seconds from when the CSN jump was detected. (The VC being 1264 established at the time is considered already validated.) 1266 5.1.6 'Migrating' the outgoing multipoint VC 1268 In addition to the group tracking described in section 5.1.4, the 1269 transmit side of a cluster member must respond to 'migration' 1270 requests by the MARS. This is triggered by the reception of a 1271 MARS_MIGRATE message from ClusterControlVC. The MARS_MIGRATE message 1272 is shown below, with an mar$op code of 13. 1274 Data: 1275 mar$hrd 16 bits Hardware type. 1276 mar$pro 56 bits Protocol Identification. 1277 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 1278 mar$chksum 16 bits Checksum across entire MARS message. 1279 mar$extoff 16 bits Extensions Offset. 1280 mar$op 16 bits Operation code (MARS_MIGRATE = 13). 1281 mar$shtl 8 bits Type & length of source ATM number. (r) 1282 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 1283 mar$spln 8 bits Length of source protocol address (s) 1284 mar$thtl 8 bits Type & length of target ATM number (x) 1285 mar$tstl 8 bits Type & length of target ATM subaddress (y) 1286 mar$tpln 8 bits Length of target group address (z) 1287 mar$tnum 16 bits Number of target ATM addresses returned (N) 1288 mar$resv 16 bits Reserved. 1289 mar$msn 32 bits MARS Sequence Number. 1290 mar$sha qoctets source ATM number 1291 mar$ssa roctets source ATM subaddress 1292 mar$spa soctets source protocol address 1293 mar$tpa zoctets target multicast group address 1294 mar$tha.1 xoctets target ATM number 1 1295 mar$tsa.1 yoctets target ATM subaddress 1 1296 mar$tha.2 xoctets target ATM number 2 1297 mar$tsa.2 yoctets target ATM subaddress 2 1298 [.......] 1299 mar$tha.N xoctets target ATM number N 1300 mar$tsa.N yoctets target ATM subaddress N 1302 A migration is requested when the MARS determines that it no longer 1303 wants cluster members forwarding their packets directly to the ATM 1304 addresses it had previously specified (through MARS_REQUESTs or 1305 MARS_JOINs). When a MARS_MIGRATE is received each cluster member MUST 1306 perform the following steps: 1308 Close down any existing outgoing VC associated with the group 1309 carried in the mar$tpa field (L_RELEASE), or dissociate the group 1310 from any outgoing VC it may have been sharing (as described in 1311 section 5.1.3). 1313 Establish a new outgoing VC for the specified group, using the 1314 algorithm described in section 5.1.3 and taking the set of ATM 1315 addresses supplied in the MARS_MIGRATE as the group's new set of 1316 members {ATM.1, .... ATM.n}. 1318 The MARS_MIGRATE carries the new set of members {ATM.1, .... ATM.n} 1319 in a single message, in similar manner to a single part MARS_MULTI. 1320 As with other messages from the MARS, the Cluster Sequence Number 1321 carried in mar$msn is checked as described in section 5.1.4.2. 1323 5.2. Receive side behaviour. 1325 A cluster member is a 'group member' (in the sense that it receives 1326 packets directed at a given multicast group) when its ATM address 1327 appears in the MARS's table entry for the group's multicast address. 1328 A key function within each cluster is the distribution of group 1329 membership information from the MARS to cluster members. 1331 An endpoint may wish to 'join a group' in response to a local, higher 1332 level request for membership of a group, or because the endpoint 1333 supports a layer 3 multicast forwarding engine that requires the 1334 ability to 'see' intra-cluster traffic in order to forward it. 1336 Two messages support these requirements - MARS_JOIN and MARS_LEAVE. 1337 These are sent to the MARS by endpoints when the local layer 3/ATM 1338 interface is requested to join or leave a multicast group. The MARS 1339 propagates these messages back out over ClusterControlVC, to ensure 1340 the knowledge of the group's membership change is distributed in a 1341 timely fashion to other cluster members. 1343 Certain models of layer 3 endpoints (e.g. IP multicast routers) 1344 expect to be able to receive packet traffic 'promiscuously' across 1345 all groups. This functionality may be emulated by allowing routers 1346 to request that the MARS returns them as 'wild card' members of all 1347 Class D addresses. However, a problem inherent in the current ATM 1348 model is that a completely promiscuous router may exhaust the local 1349 reassembly resources in its ATM interface. MARS_JOIN supports a 1350 generalisation to the notion of 'wild card' entries, enabling routers 1351 to limit themselves to 'blocks' of the Class D address space. Use of 1352 this facility is described in greater detail in Section 8. 1354 A block can be as small as 1 (a single group) or as large as the 1355 entire multicast address space (e.g. default IPv4 'promiscuous' 1356 behaviour). A block is defined as all addresses between, and 1357 inclusive of, a address pair. A MARS_JOIN or MARS_LEAVE may 1358 carry multiple pairs. 1360 Cluster members MUST provide ONLY a single pair in each 1361 JOIN/LEAVE message they issue. However, they MUST be able to process 1362 multiple pairs in JOIN/LEAVE messages when performing VC 1363 management as described in section 5.1.4 (the interpretation being 1364 that the join/leave operation applies to all addresses in range from 1365 to inclusive, for every pair). 1367 In RFC1112 environments a MARS_JOIN for a single group is triggered 1368 by a JoinLocalGroup signal from the IP layer. A MARS_LEAVE for a 1369 single group is triggered by a LeaveLocalGroup signal from the IP 1370 layer. 1372 Cluster members with special requirements (e.g. multicast routers) 1373 may issue MARS_JOINs and MARS_LEAVEs specifying a block of multicast 1374 group addresses. 1376 An endpoint MUST register with a MARS in order to become a member of 1377 a cluster and be added as a leaf to ClusterControlVC. Registration 1378 is covered in section 5.2.3. 1380 Finally, the endpoint MUST be capable of terminating unidirectional 1381 VCs (i.e. act as a leaf node of a UNI 3.0/3.1 point to multipoint VC, 1382 with zero bandwidth assigned on the return path). RFC 1755 describes 1383 the signalling information required to terminate VCs carrying 1384 LLC/SNAP encapsulated traffic (discussed further in section 5.5). 1386 5.2.1 Format of the MARS_JOIN and MARS_LEAVE Messages. 1388 The MARS_JOIN message is indicated by an operation type value of 4. 1389 MARS_LEAVE has the same format and operation type value of 5. The 1390 message format is: 1392 Data: 1393 mar$hrd 16 bits Hardware type. 1394 mar$pro 56 bits Protocol Identification. 1395 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 1396 mar$chksum 16 bits Checksum across entire MARS message. 1397 mar$extoff 16 bits Extensions Offset. 1398 mar$op 16 bits Operation code (MARS_JOIN or MARS_LEAVE). 1399 mar$shtl 8 bits Type & length of source ATM number. (r) 1400 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 1401 mar$spln 8 bits Length of source protocol address (s) 1402 mar$tpln 8 bits Length of group address (z) 1403 mar$pnum 16 bits Number of group address pairs (N) 1404 mar$flags 16 bits layer3grp, copy, and register flags. 1405 mar$cmi 16 bits Cluster Member ID 1406 mar$msn 32 bits MARS Sequence Number. 1407 mar$sha qoctets source ATM number. 1408 mar$ssa roctets source ATM subaddress. 1409 mar$spa soctets source protocol address 1410 mar$min.1 zoctets Minimum multicast group address - pair.1 1411 mar$max.1 zoctets Maximum multicast group address - pair.1 1412 [.......] 1413 mar$min.N zoctets Minimum multicast group address - pair.N 1414 mar$max.N zoctets Maximum multicast group address - pair.N 1416 mar$spln indicates the number of bytes in the source endpoint's 1417 protocol address, and is interpreted in the context of the protocol 1418 indicated by the mar$pro field. (e.g. in IPv4 environments mar$pro 1419 will be 0x800, mar$spln is 4, and mar$tpln is 4.) 1421 The mar$flags field contains three flags: 1423 Bit 15 - mar$flags.layer3grp. 1424 Bit 14 - mar$flags.copy. 1425 Bit 13 - mar$flags.register. 1426 Bit 12 - mar$flags.punched. 1427 Bit 0-7 - mar$flags.sequence. 1429 Bits 8 to 11 are reserved and MUST be zero. 1431 mar$flags.sequence is set by cluster members, and MUST always be 1432 passed on unmodified by the MARS when retransmitting MARS_JOIN or 1433 MARS_LEAVE messages. It is source specific, and MUST be ignored by 1434 other cluster members. Its use is described in section 5.2.2. 1436 mar$flags.punched MUST be zero when the MARS_JOIN or MARS_LEAVE is 1437 transmitted to the MARS. Its use is described in section 5.2.2 and 1438 section 6.2.4. 1440 mar$flags.copy MUST be set to 0 when the message is being sent from a 1441 MARS client, and MUST be set to 1 when the message is being sent from 1442 a MARS. (This flag is intended to support integrating the MARS 1443 function with one of the MARS clients in your cluster. The 1444 destination of an incoming MARS_JOIN can be determined from its 1445 value.) 1447 mar$flags.layer3grp allows the MARS to provide the group membership 1448 information described further in section 5.3. The rules for its use 1449 are: 1451 mar$flags.layer3grp MUST be set when the cluster member is issuing 1452 the MARS_JOIN as the result of a layer 3 multicast group being 1453 explicitly joined. (e.g. as a result of a JoinHostGroup operation 1454 in an RFC1112 compliant host). 1456 mar$flags.layer3grp MUST be reset in each MARS_JOIN if the 1457 MARS_JOIN is simply the local ip/atm interface registering to 1458 receive traffic on that group for its own reasons. 1460 mar$flags.layer3grp is ignored and MUST be treated as reset by the 1461 MARS for any MARS_JOIN that specifies a block covering more than a 1462 single group (e.g. a block join from a router ensuring their 1463 forwarding engines 'see' all traffic). 1465 mar$flags.register indicates whether the MARS_JOIN or MARS_LEAVE is 1466 being used to register or deregister a cluster member (described in 1467 section 5.2.3). When used to join or leave specific groups the 1468 mar$register flag MUST be zero. 1470 mar$pnum indicates how many pairs are included in the 1471 message. This field MUST be 1 when the message is sent from a cluster 1472 member. A MARS MAY return a MARS_JOIN or MARS_LEAVE with any mar$pnum 1473 value, including zero. This will be explained futher in section 1474 6.2.4. 1476 The mar$cmi field MUST be zeroed by cluster members, and is used by 1477 the MARS during cluster member registration, described in section 1478 5.2.3. 1480 mar$msn MUST be zero when transmitted by an endpoint. It is set to 1481 the current value of the Cluster Sequence Number by the MARS when the 1482 MARS_JOIN or MARS_LEAVE is retransmitted. Its use has been described 1483 in section 5.1.4. 1485 To simplify construction and parsing of MARS_JOIN and MARS_LEAVE 1486 messages, the following restrictions are imposed on the 1487 pairs: 1489 Assume max(N) is the field from the Nth pair. 1490 Assume min(N) is the field from the Nth pair. 1491 Assume a join/leave message arrives with K pairs. 1492 The following must hold: 1493 max(N) < min(N+1) for 1 <= N < K 1494 max(N) >= min(N) for 1 <= N <= K 1496 In plain language, the set must specify an ascending sequence of 1497 address blocks. The definition of "greater" or "less than" may be 1498 protocol specific. In IPv4 environments the addresses are treated as 1499 32 bit, unsigned binary values (most significant byte first). 1501 5.2.1.1 Important IPv4 default values. 1503 The JoinLocalGroup and LeaveLocalGroup operations are only valid for 1504 a single group. For any arbitrary group address X the associated 1505 MARS_JOIN or MARS_LEAVE MUST specify a single pair . 1506 mar$flags.layer3grp MUST be set under these circumstances. 1508 A router choosing to behave strictly in accordance with RFC1112 MUST 1509 specify the entire Class D space. The associated MARS_JOIN or 1510 MARS_LEAVE MUST specify a single pair <224.0.0.0, 239.255.255.255>. 1511 Whenever a router issues a MARS_JOIN only in order to forward IP 1512 traffic it MUST reset mar$flags.layer3grp. 1514 The use of alternative values by multicast routers is 1515 discussed in Section 8. 1517 5.2.2 Retransmission of MARS_JOIN and MARS_LEAVE messages. 1519 Transient problems may result in the loss of messages between the 1520 MARS and cluster members 1522 A simple algorithm is used to solve this problem. Cluster members 1523 retransmit each MARS_JOIN and MARS_LEAVE message at regular intervals 1524 until they receive a copy back again, either on ClusterControlVC or 1525 the VC on which they are sending the message. At this point the 1526 local endpoint can be certain that the MARS received and processed 1527 it. 1529 The interval should be no shorter than 5 seconds, and a default value 1530 of 10 seconds is recommended. After 5 retransmissions the attempt 1531 should be flagged locally as a failure. This MUST be considered as a 1532 MARS failure, and triggers the MARS reconnection described in section 1533 5.4. 1535 A 'copy' is defined as a received message with the following fields 1536 matching a previously transmitted MARS_JOIN/LEAVE: 1538 - mar$op 1539 - mar$flags.register 1540 - mar$flags.sequence 1541 - mar$pnum 1542 - Source ATM address 1543 - First pair 1545 In addition, a valid copy MUST have the following field values: 1547 - mar$flags.punched = 0 1548 - mar$flags.copy = 1 1550 The mar$flags.sequence field is never modified or checked by a MARS. 1551 Implementors MAY choose to utilize locally significant sequence 1552 number schemes, which MAY differ from one cluster member to the next. 1553 In the absence of such schemes the default value for 1554 mar$flags.sequence MUST be zero. 1556 Careful implementations MAY have more than one outstanding 1557 (unacknowledged) MARS_JOIN/LEAVE at a time. 1559 5.2.3 Cluster member registration and deregistration. 1561 To become a cluster member an endpoint must register with the MARS. 1562 This achieves two things - the endpoint is added as a leaf node of 1563 ClusterControlVC, and the endpoint is assigned a 16 bit Cluster 1564 Member Identifier (CMI). The CMI uniquely identifies each endpoint 1565 that is attached to the cluster. 1567 Registration with the MARS occurs when an endpoint issues a MARS_JOIN 1568 with the mar$flags.register flag set to one (bit 13 of the mar$flags 1569 field). 1571 The cluster member MUST include its source ATM address, and MAY 1572 choose to specify a null source protocol address when registering. 1574 No protocol specific group addresses are included in a registration 1575 MARS_JOIN. 1577 The cluster member retransmits this MARS_JOIN in accordance with 1578 section 5.2.2 until it confirms that the MARS has received it. 1580 When the registration MARS_JOIN is returned it contains a non-zero 1581 value in mar$cmi. This value MUST be noted by the cluster member, and 1582 used whenever circumstances require the cluster member's CMI. 1584 An endpoint may also choose to de-register, using a MARS_LEAVE with 1585 mar$flags.register set. This would result in the MARS dropping the 1586 endpoint from ClusterControlVC, removing all references to the member 1587 in the mapping database, and freeing up its CMI. 1589 As for registration, a deregistration request MUST include the 1590 correct source ATM address for the cluster member, but MAY choose to 1591 specify a null source protocol address. 1593 The cluster member retransmits this MARS_LEAVE in accordance with 1594 section 5.2.2 until it confirms that the MARS has received it. 1596 5.3 Support for Layer 3 group management. 1598 Whilst the intention of this specification is to be independent of 1599 layer 3 issues, an attempt is being made to assist the operation of 1600 layer 3 multicast routing protocols that need to ascertain if any 1601 groups have members within a cluster. 1603 One example is IP, where IGMP is used (as described in section 2) 1604 simply to determine whether any other cluster members are listening 1605 to a group because they have higher layer applications that want to 1606 receive a group's traffic. 1608 Routers may choose to query the MARS for this information, rather 1609 than multicasting IGMP queries to 224.0.0.1 and incurring the 1610 associated cost of setting up a VC to all systems in the cluster. 1612 The query is issued by sending a MARS_GROUPLIST_REQUEST to the MARS. 1613 MARS_GROUPLIST_REQUEST is built from a MARS_JOIN, but it has an 1614 operation code of 10. The first pair will be used by the 1615 MARS to identify the range of groups in which the querying cluster 1616 member is interested. Any additional pairs will be ignored. 1617 A request with mar$pnum = 0 will be ignored. 1619 The response from the MARS is a MARS_GROUPLIST_REPLY, carrying a list 1620 of the multicast groups within the specified block that 1621 have Layer 3 members. A group is noted in this list if one or more 1622 of the MARS_JOINs that generated its mapping entry in the MARS 1623 contained a set mar$flags.layer3grp flag. 1625 MARS_GROUPLIST_REPLYs are transmitted back to the querying cluster 1626 member on the VC used to send the MARS_GROUPLIST_REQUEST. 1628 MARS_GROUPLIST_REPLY is derived from the MARS_MULTI but with mar$op = 1629 11. It may have multiple parts if needed, and is received in a 1630 similar manner to a MARS_MULTI. 1632 Data: 1633 mar$hrd 16 bits Hardware type. 1634 mar$pro 56 bits Protocol Identification. 1635 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 1636 mar$chksum 16 bits Checksum across entire MARS message. 1637 mar$extoff 16 bits Extensions Offset. 1638 mar$op 16 bits Operation code (MARS_GROUPLIST_REPLY). 1639 mar$shtl 8 bits Type & length of source ATM number. (r) 1640 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 1641 mar$spln 8 bits Length of source protocol address (s) 1642 mar$thtl 8 bits Unused - set to zero. 1643 mar$tstl 8 bits Unused - set to zero. 1645 mar$tpln 8 bits Length of target group address (z) 1646 mar$tnum 16 bits Number of group addresses returned (N). 1647 mar$seqxy 16 bits Boolean flag x and sequence number y. 1648 mar$msn 32 bits MARS Sequence Number. 1649 mar$sha qoctets source ATM number. 1650 mar$ssa roctets source ATM subaddress. 1651 mar$spa soctets source protocol address 1652 mar$mgrp.1 zoctets Group address 1 1653 [.......] 1654 mar$mgrp.N zoctets Group address N 1656 mar$seqxy is coded as for the MARS_MULTI - multiple 1657 MARS_GROUPLIST_REPLY components are transmitted and received using 1658 the same algorithm as described in section 5.1.1 for MARS_MULTI. The 1659 only difference is that protocol address are being returned rather 1660 than ATM addresses. 1662 As for MARS_MULTIs, if an error occurs in the reception of a multi 1663 part MARS_GROUPLIST_REPLY the whole thing MUST be discarded and the 1664 MARS_GROUPLIST_REQUEST re-issued. (This includes the mar$msn value 1665 being constant.) 1667 Note that the ability to generate MARS_GROUPLIST_REQUEST messages, 1668 and receive MARS_GROUPLIST_REPLY messages, is not required for 1669 general host interface implementations. It is optional for interfaces 1670 being implemented to support layer 3 multicast forwarding engines. 1671 However, this functionality MUST be supported by the MARS. 1673 5.4 Support for redundant/backup MARS entities. 1675 Endpoints are assumed to have been configured with the ATM address of 1676 at least one MARS. Endpoints MAY choose to maintain a table of ATM 1677 addresses, representing alternative MARSs that will be contacted in 1678 the event that normal operation with the original MARS is deemed to 1679 have failed. It is assumed that this table orders the ATM addresses 1680 in descending order of preference. 1682 An endpoint will typically decide there are problems with the MARS 1683 when: 1685 - It fails to establish a point to point VC to the MARS. 1686 - MARS_REQUESTs fail (section 5.1.1). 1687 - MARS_JOIN/MARS_LEAVEs fail (section 5.2.2). 1688 - It has not received a MARS_REDIRECT_MAP in the last 4 minutes 1689 (section 5.4.3). 1691 (If it is able to discern which connection represents 1692 ClusterControlVC, it may also use connection failures on this VC to 1693 indicate problems with the MARS). 1695 5.4.1 First response to MARS problems. 1697 The first response is to assume a transient problem with the MARS 1698 being used at the time. The cluster member should wait a random 1699 period of time between 1 and 10 seconds before attempting to re- 1700 connect and re-register with the MARS. If the registration MARS_JOIN 1701 is successful then: 1703 The cluster member MUST then proceed to rejoin every group that 1704 its local higher layer protocol(s) have joined. It is recommended 1705 that a random delay between 1 and 10 seconds be inserted before 1706 attempting each MARS_JOIN. 1708 The cluster member MUST initiate the revalidation of every 1709 multicast group it was sending to (as though a sequence number 1710 jump had been detected, section 5.1.5). 1712 The rejoin and revalidation procedure must not disrupt the cluster 1713 member's use of multipoint VCs that were already open at the time 1714 of the MARS failure. 1716 If re-registration with the current MARS fails, and there are no 1717 backup MARS addresses configured, the cluster member MUST wait for at 1718 least 1 minute before repeating the re-registration procedure. It is 1719 RECOMMENDED that the cluster member signals an error condition in 1720 some locally significant fashion. 1722 This procedure may repeat until network administrators manually 1723 intervene or the current MARS returns to normal operation. 1725 5.4.2 Connecting to a backup MARS. 1727 If the re-registration with the current MARS fails, and other MARS 1728 addresses have been configured, the next MARS address on the list is 1729 chosen to be the current MARS, and the cluster member immediately 1730 restarts the re-registration procedure described in section 5.4.1. If 1731 this is succesful the cluster member will resume normal operation 1732 using the new MARS. It is RECOMMENDED that the cluster member signals 1733 a warning of this condition in some locally significant fashion. 1735 If the attempt at re-registration with the new MARS fails, the 1736 cluster member MUST wait for at least 1 minute before chosing the 1737 next MARS address in the table and repeating the procedure. If the 1738 end of the table has been reached, the cluster member starts again at 1739 the top of the table (which should be the original MARS that the 1740 cluster member started with). 1742 In the worst case scenario this will result in cluster members 1743 looping through their table of possible MARS addresses until network 1744 administrators manually intervene. 1746 5.4.3 Dynamic backup lists, and soft redirects. 1748 To support some level of autoconfiguration, a MARS message is defined 1749 that allows the current MARS to broadcast on ClusterControlVC a table 1750 of backup MARS addresses. When this message is received, cluster 1751 members that maintain a list of backup MARS addresses MUST insert 1752 this information at the top of their locally held list (i.e. the 1753 information provided by the MARS has a higher preference than 1754 addresses that may have been manually configured into the cluster 1755 member). 1757 The message is MARS_REDIRECT_MAP. It is based on the MARS_MULTI 1758 message, with the following changes: 1760 - mar$tpln field replaced by mar$redirf. 1761 - mar$spln field reserved. 1762 - mar$tpa and mar$spa eliminated. 1764 MARS_REDIRECT_MAP has an operation type code of 12 decimal. 1766 Data: 1767 mar$hrd 16 bits Hardware type. 1768 mar$pro 56 bits Protocol Identification. 1769 mar$hdrrsv 24 bits Reserved. Unused by MARS control protocol. 1770 mar$chksum 16 bits Checksum across entire MARS message. 1771 mar$extoff 16 bits Extensions Offset. 1772 mar$op 16 bits Operation code (MARS_REDIRECT_MAP). 1773 mar$shtl 8 bits Type & length of source ATM number. (r) 1774 mar$sstl 8 bits Type & length of source ATM subaddress. (q) 1775 mar$spln 8 bits Length of source protocol address (s) 1776 mar$thtl 8 bits Type & length of target ATM number (x) 1777 mar$tstl 8 bits Type & length of target ATM subaddress (y) 1778 mar$redirf 8 bits Flag controlling client redirect behaviour. 1779 mar$tnum 16 bits Number of MARS addresses returned (N). 1780 mar$seqxy 16 bits Boolean flag x and sequence number y. 1781 mar$msn 32 bits MARS Sequence Number. 1782 mar$sha qoctets source ATM number 1783 mar$ssa roctets source ATM subaddress 1784 mar$tha.1 xoctets ATM number for MARS 1 1785 mar$tsa.1 yoctets ATM subaddress for MARS 1 1786 mar$tha.2 xoctets ATM number for MARS 2 1787 mar$tsa.2 yoctets ATM subaddress for MARS 2 1789 [.......] 1790 mar$tha.N xoctets ATM number for MARS N 1791 mar$tsa.N yoctets ATM subaddress for MARS N 1793 The source ATM address field(s) MUST identify the originating MARS. 1794 A multi-part MARS_REDIRECT_MAP may be transmitted and reassembled 1795 using the mar$seqxy field in the same manner as a multi-part 1796 MARS_MULTI (section 5.1.1). If a failure occurs during the reassembly 1797 of a multi-part MARS_REDIRECT_MAP (a part lost, reassembly timeout, 1798 or illegal MARS Sequence Number jump) the entire message MUST be 1799 discarded. 1801 This message is transmitted regularly by the MARS (it MUST be 1802 transmitted at least every 2 minutes, it is RECOMMENDED that it is 1803 transmitted every 1 minute). 1805 The MARS_REDIRECT_MAP is also used to force cluster members to shift 1806 from one MARS to another. If the ATM address of the first MARS 1807 contained in a MARS_REDIRECT_MAP table is not the address of cluster 1808 member's current MARS the client MUST 'redirect' to the new MARS. The 1809 mar$redirf field controls how the redirection occurs. 1811 mar$redirf has the following format: 1813 7 6 5 4 3 2 1 0 1814 +-+-+-+-+-+-+-+-+ 1815 |x| | 1816 +-+-+-+-+-+-+-+-+ 1818 If Bit 7 (the most significant bit) of mar$redirf is 1 then the 1819 cluster member MUST perform a 'hard' redirect. Having installed the 1820 new table of MARS addresses carried by the MARS_REDIRECT_MAP, the 1821 cluster member re-registers with the MARS now at the top of the table 1822 using the mechanism described in sections 5.4.1 and 5.4.2. 1824 If Bit 7 of mar$redirf is 0 then the cluster member MUST perform a 1825 'soft' redirect, beginning with the following actions: 1827 - open a point to point VC to the first ATM address. 1828 - attempt a registration (section 5.2.3). 1830 If the registration succeeds, the cluster member shuts down its point 1831 to point VC to the current MARS (if it had one open), and then 1832 proceeds to use the newly opened point to point VC as its connection 1833 to the 'current MARS'. The cluster member does NOT attempt to rejoin 1834 the groups it is a member of, or revalidate groups it is currently 1835 sending to. 1837 This is termed a 'soft redirect' because it avoids the extra 1838 rejoining and revalidation processing that occurs when a MARS failure 1839 is being recovered from. It assumes some external synchronisation 1840 mechanisms exist between the old and new MARS - mechanisms that are 1841 outside the scope of this specification. 1843 Some level of trust is required before initiating a soft redirect. A 1844 cluster member MUST check that the calling party at the other end of 1845 the VC on which the MARS_REDIRECT_MAP arrived (supposedly 1846 ClusterControlVC) is in fact the node it trusts as the current MARS. 1848 Additional applications of this function are for further study. 1850 5.5 Data path LLC/SNAP encapsulations. 1852 An extended encapsulation scheme is required to support the filtering 1853 of possible reflected packets (section 3.3). 1855 Two LLC/SNAP codepoints are allocated from the IANA OUI space. These 1856 support two different mechanisms for detecting reflected packets. 1857 They are called Type #1 and Type #2 multicast encapsulations. 1859 Type #1 1861 [0xAA-AA-03][0x00-00-5E][0x00-01][Type #1 Extended Layer 3 packet] 1862 LLC OUI PID 1864 Type #2 1866 [0xAA-AA-03][0x00-00-5E][0x00-04][Type #2 Extended Layer 3 packet] 1867 LLC OUI PID 1869 For conformance with this document MARS clients: 1871 MUST transmit data using Type #1 encapsulation. 1873 MUST be able to correctly receive traffic using Type #1 OR Type #2 1874 encapsulation. 1876 MUST NOT transmit using Type #2 encapsulation. 1878 5.5.1 Type #1 encapsulation. 1880 The Type #1 Extended layer 3 packet carries within it a copy of the 1881 source's Cluster Member ID (CMI) and either the 'short form' or 'long 1882 form' of the protocol type as appropriate (section 4.3). 1884 When carrying packets belonging to protocols with valid short form 1885 representations the [Type #1 Extended Layer 3 packet] is encoded as: 1887 [pkt$cmi][pkt$pro][Original Layer 3 packet] 1888 2octet 2octet N octet 1890 The first 2 octets (pkt$cmi) carry the CMI assigned when an endpoint 1891 registers with the MARS (section 5.2.3). The second 2 octets 1892 (pkt$pro) indicate the protocol type of the packet carried in the 1893 remainder of the payload. This is copied from the mar$pro field used 1894 in the MARS control messages. 1896 When carrying packets belonging to protocols that only have a long 1897 form representation (pkt$pro = 0x80) the overhead SHALL be further 1898 extended to carry the 5 byte mar$pro.snap field (with padding for 32 1899 bit alignment). The encoded form SHALL be: 1901 [pkt$cmi][0x00-80][mar$pro.snap][padding][Original Layer 3 packet] 1902 2octet 2octet 5 octets 3 octets N octet 1904 The CMI is copied into the pkt$cmi field of every outgoing Type #1 1905 packet. When an endpoint interface receives an AAL_SDU with the 1906 LLC/SNAP codepoint indicating Type #1 encapsulation it compares the 1907 CMI field with its own Cluster Member ID for the indicated protocol. 1908 The packet is discarded silently if they match. Otherwise the packet 1909 is accepted for processing by the local protocol entity identified by 1910 the pkt$pro (and possibly SNAP) field(s). 1912 Where a protocol has valid short and long forms of identification, 1913 receivers MAY choose to additionally recognise the long form. 1915 5.5.2 Type #2 encapsulation. 1917 Future developments may enable direct multicasting of AAL_SDUs beyond 1918 cluster boundaries. Expanding the set of possible sources in this way 1919 may cause the CMI to become an inadequate parameter with which to 1920 detect reflected packets. A larger source identification field may 1921 be required. 1923 The Type #2 Extended layer 3 packet carries within it an 8 octet 1924 source ID field and either the 'short form' or 'long form' of the 1925 protocol type as appropriate (section 4.3). The form and content of 1926 the source ID field is currently unspecified, and is not relevant to 1927 any MARS client built in conformance with this document. Received 1928 Type #2 encapsulated packets MUST always be accepted and passed up to 1929 the higher layer indicated by the protocol identifier. 1931 When carrying packets belonging to protocols with valid short form 1932 representations the [Type #2 Extended Layer 3 packet] is encoded as: 1934 [8 octet sourceID][mar$pro.type][Null pad][Original Layer 3 1935 packet] 1936 2octets 2octets 1938 When carrying packets belonging to protocols that only have a long 1939 form representation (pkt$pro = 0x80) the overhead SHALL be further 1940 extended to carry the 5 byte mar$pro.snap field (with padding for 32 1941 bit alignment). The encoded form SHALL be: 1943 [8 octet sourceID][mar$pro.type][mar$pro.snap][Null pad][Layer 3 1944 packet] 1945 2octets 5octets 1octet 1947 (Note that in this case the padding after the SNAP field is 1 octet 1948 rather than the 3 octets used in Type #1.) 1950 Where a protocol has valid short and long forms of identification, 1951 receivers MAY choose to additionally recognise the long form. 1953 (Future documents may specify the contents of the source ID field. 1954 This will only be relevant to implementations sending Type #2 1955 encapsulated packets, as they are the only entities that need to be 1956 concerned about detecting reflected Type #2 packets.) 1958 5.5.3 A Type #1 example. 1960 An IPv4 packet (fully identified by an Ethertype of 0x800, therefore 1961 requiring 'short form' protocol type encoding) would be transmitted 1962 as: 1964 [0xAA-AA-03][0x00-00-5E][0x00-01][pkt$cmi][0x800][IPv4 packet] 1966 The different LLC/SNAP codepoints for unicast and multicast packet 1967 transmission allows a single IPv4/ATM interface to support both by 1968 demuxing on the LLC/SNAP header. 1970 6. The MARS in greater detail. 1972 Section 5 implies a lot about the MARS's basic behaviour as observed 1973 by cluster members. This section summarises the behaviour of the MARS 1974 for groups that are VC mesh based, and describes how a MARSs 1975 behaviour changes when an MCS is registered to support a group. 1977 The MARS is intended to be a multiprotocol entity - all its mapping 1978 tables, CMIs, and control VCs MUST be managed within the context of 1979 the mar$pro field in incoming MARS messages. For example, a MARS 1980 supports completely separate ClusterControlVCs for each layer 3 1981 protocol that it is registering members for. If a MARS receives 1982 messages with an mar$pro that it does not support, the message is 1983 dropped. 1985 In general the MARS treats protocol addresses as arbitrary byte 1986 strings. For example, the MARS will not apply IPv4 specific 'class' 1987 checks to addresses supplied under mar$pro = 0x800. It is sufficient 1988 for the MARS to simply assume that endpoints know how to interpret 1989 the protocol addresses that they are establishing and releasing 1990 mappings for. 1992 The MARS requires control messages to carry the originator's identity 1993 in the source ATM address field(s). Messages that arrive with an 1994 empty ATM Number field are silently discarded prior to any other 1995 processing by the MARS. (Only the ATM Number field needs to be 1996 checked. An empty ATM Number field combined with a non-empty ATM 1997 Subaddress field does not represent a valid ATM address.) 1999 (Some example pseudo-code for a MARS can be found in Appendix F.) 2001 6.1 Basic interface to Cluster members. 2003 The following MARS messages are used or required by cluster members: 2005 1 MARS_REQUEST 2006 2 MARS_MULTI 2007 4 MARS_JOIN 2008 5 MARS_LEAVE 2009 6 MARS_NAK 2010 10 MARS_GROUPLIST_REQUEST 2011 11 MARS_GROUPLIST_REPLY 2012 12 MARS_REDIRECT_MAP 2014 6.1.1 Response to MARS_REQUEST. 2016 Except as described in section 6.2, if a MARS_REQUEST arrives whose 2017 source ATM address does not match that of any registered Cluster 2018 member the message MUST be dropped and ignored. 2020 6.1.2 Response to MARS_JOIN and MARS_LEAVE. 2022 When a registration MARS_JOIN arrives (described in section 5.2.3) 2023 the MARS performs the following actions: 2025 - Adds the node to ClusterControlVC. 2026 - Allocates a new Cluster Member ID (CMI). 2028 - Inserts the new CMI into the mar$cmi field of the MARS_JOIN. 2029 - Retransmits the MARS_JOIN back privately. 2031 If the node is already a registered member of the cluster associated 2032 with the specified protocol type then its existing CMI is simply 2033 copied into the MARS_JOIN, and the MARS_JOIN retransmitted back to 2034 the node. A single node may register multiple times if it supports 2035 multiple layer 3 protocols. The CMIs allocated by the MARS for each 2036 such registration may or may not be the same. 2038 The retransmitted registration MARS_JOIN must NOT be sent on 2039 ClusterControlVC. If a cluster member issues a deregistration 2040 MARS_LEAVE it too is retransmitted privately. 2042 Non-registration MARS_JOIN and MARS_LEAVE messages are ignored if 2043 they arrive from a node that is not registered as a cluster member. 2045 Except as described in section 6.2.4, after performing any required 2046 database updates non-registration MARS_JOIN and MARS_LEAVE messages 2047 are retransmitted on ClusterControlVC. The following fields are 2048 modified just prior to retransmission: 2050 mar$flags.copy is set to 1. 2052 mar$msn is set to the current Cluster Sequence Number for 2053 ClusterControlVC (Section 5.1.4.2). 2055 The MARS retransmits MARS_JOIN and MARS_LEAVE messages even if they 2056 resulted in no change to the database. 2058 MARS_JOIN or MARS_LEAVE messages MUST arrive at the MARS with 2059 mar$flags.copy set to 0, otherwise the message is silently ignored. 2060 All outgoing MARS_JOIN or MARS_LEAVE messages have mar$flags.copy set 2061 to 1. 2063 mar$flags.layer3grp (section 5.3) MUST be ignored (and treated as 2064 reset) for MARS_JOINs specifying more than a single group. If a 2065 MARS_JOIN/LEAVE is received that contains more than one 2066 pair, the MARS MUST silently drop the message. 2068 If the MARS receives a deregistration MARS_LEAVE (described in 2069 section 5.2.3) that member's ATM address MUST be removed from all 2070 groups for which it may have joined, dropped from ClusterControlVC, 2071 and the CMI released. 2073 If the MARS receives an ERR_L_RELEASE on ClusterControlVC indicating 2074 that a cluster member has disconnected, that member's ATM address 2075 MUST be removed from all groups for which it may have joined, and the 2076 CMI released. 2078 6.1.3 Generating MARS_REDIRECT_MAP. 2080 A MARS_REDIRECT_MAP message (described in section 5.4.3) MUST be 2081 regularly transmitted on ClusterControlVC. It is RECOMMENDED that 2082 this occur every 1 minute, and it MUST occur at least every 2 2083 minutes. If the MARS has no knowledge of other backup MARSs serving 2084 the cluster, it MUST include its own address as the only entry in the 2085 MARS_REDIRECT_MAP message (in addition to filling in the source 2086 address fields). 2088 The design and use of backup MARS entities is beyond the scope of 2089 this document, and will be covered in future work. 2091 6.1.4 Cluster Sequence Numbers. 2093 The Cluster Sequence Number (CSN) is described in section 5.1.4, and 2094 is carried in the mar$msn field of MARS messages being sent to 2095 cluster members (either out ClusterControlVC or on an individual VC). 2096 The MARS increments the CSN after every transmission of a message on 2097 ClusterControlVC. The current CSN is copied into the mar$msn field 2098 of MARS messages being sent to cluster members, whether out 2099 ClusterControlVC or on a private VC. 2101 A MARS should be carefully designed to minimise the possibility of 2102 the CSN jumping unecessarily. Under normal operation only cluster 2103 members affected by transient link problems will miss CSN updates and 2104 be forced to revalidate. If the MARS itself glitches, it will be 2105 innundated with requests for a period as every cluster member 2106 attempts to revalidate. 2108 Calculations on the CSN MUST be performed as unsigned 32 bit 2109 arithmetic. 2111 One implication of this mechanism is that the MARS should serialize 2112 its processing of 'simultaneous' MARS_REQUEST, MARS_JOIN and 2113 MARS_LEAVE messages. Join and Leave operations should be queued 2114 within the MARS along with MARS_REQUESTS, and not processed until all 2115 the reply packets of a preceeding MARS_REQUEST have been transmitted. 2116 The transmission of MARS_REDIRECT_MAP should also be similarly 2117 queued. 2119 (The regular transmission of MARS_REDIRECT_MAP serves a secondary 2120 purpose of allowing cluster members to track the CSN, even if they 2121 miss an earlier MARS_JOIN or MARS_LEAVE.) 2123 6.2 MARS interface to Multicast Servers (MCS). 2125 When the MARS returns the actual addresses of group members, the 2126 endpoint behaviour described in section 5 results in all groups being 2127 supported by meshes of point to multipoint VCs. However, when MCSs 2128 register to support particular layer 3 multicast groups the MARS 2129 modifies its use of various MARS messages to fool endpoints into 2130 using the MCS instead. 2132 The following MARS messages are associated with interaction between 2133 the MARS and MCSs. 2135 3 MARS_MSERV 2136 7 MARS_UNSERV 2137 8 MARS_SJOIN 2138 9 MARS_SLEAVE 2140 The following MARS messages are treated in a slightly different 2141 manner when MCSs have registered to support certain group addresses: 2143 1 MARS_REQUEST 2144 4 MARS_JOIN 2145 5 MARS_LEAVE 2147 A MARS must keep two sets of mappings for each layer 3 group using 2148 MCS support. The original {layer 3 address, ATM.1, ATM.2, ... ATM.n} 2149 mapping (now termed the 'host map', although it includes routers) is 2150 augmented by a parallel {layer 3 address, server.1, server.2, .... 2151 server.K} mapping (the 'server map'). It is assumed that no ATM 2152 addresses appear in both the server and host maps for the same 2153 multicast group. Typically K will be 1, but it will be larger if 2154 multiple MCSs are configured to support a given group. 2156 The MARS also maintains a point to multipoint VC out to any MCSs 2157 registered with it, called ServerControlVC (section 6.2.3). This 2158 serves an analogous role to ClusterControlVC, allowing the MARS to 2159 update the MCSs with group membership changes as they occur. A MARS 2160 MUST also send its regular MARS_REDIRECT_MAP transmissions on both 2161 ServerControlVC and ClusterControlVC. 2163 6.2.1 Response to a MARS_REQUEST if MCS is registered. 2165 When the MARS receives a MARS_REQUEST for an address that has both 2166 host and server maps it generates a response based on the identity of 2167 the request's source. If the requestor is a member of the server map 2168 for the requested group then the MARS returns the contents of the 2169 host map in a sequence of one or more MARS_MULTIs. Otherwise, if the 2170 source is a valid cluster member, the MARS returns the contents of 2171 the server map in a sequence of one or more MARS_MULTIs. If the 2172 source is neither a cluster member, nor a member of the server map 2173 for the group, the request is dropped and ignored. 2175 Servers use the host map to establish a basic distribution VC for the 2176 group. Cluster members will establish outgoing multipoint VCs to 2177 members of the group's server map, without being aware that their 2178 packets will not be going directly to the multicast group's members. 2180 6.2.2 MARS_MSERV and MARS_UNSERV messages. 2182 MARS_MSERV and MARS_UNSERV are identical to the MARS_JOIN message. 2183 An MCS uses a MARS_MSERV with a pair of to specify 2184 the multicast group X that it is willing to support. A single group 2185 MARS_UNSERV indicates the group that the MCS is no longer willing to 2186 support. The operation code for MARS_MSERV is 3 (decimal), and 2187 MARS_UNSERV is 7 (decimal). 2189 Both of these messages are sent to the MARS over a point to point VC 2190 (between MCS and MARS). After processing, they are retransmitted on 2191 ServerControlVC to allow other MCSs to note the new node. 2193 When registering or deregistering support for specific groups the 2194 mar$flags.register flag MUST be zero. (This flag is only one when the 2195 MCS is registering as a member of ServerControlVC, as described in 2196 section 6.2.3.) 2198 When an MCS issues a MARS_MSERV for a specific group the message MUST 2199 be dropped and ignored if the source has not already registered with 2200 the MARS as a multicast server (section 6.2.3). Otherwise, the MARS 2201 adds the new ATM address to the server map for the specified group, 2202 possibly constructing a new server map if this is the first MCS for 2203 the group. 2205 If a MARS_MSERV represents the first MCS to register for a particular 2206 group, and there exists a non null host map serving that particular 2207 group, the MARS issues a MARS_MIGRATE (section 5.1.6) on 2208 ClusterControlVC. The MARS's own identity is placed in the source 2209 protocol and hardware address fields of the MARS_MIGRATE. The ATM 2210 address of the MCS is placed as the first and only target ATM 2211 address. The address of the affected group is placed in the target 2212 multicast group address field. 2214 If a MARS_MSERV is not the first MCS to register for a particular 2215 group the MARS simply changes its operation code to MARS_JOIN, and 2216 sends a copy of the message on ClusterControlVC. This fools the 2217 cluster members into thinking a new leaf node as been added to the 2218 group specified. In the retransmitted MARS_JOIN mar$flags.layer3grp 2219 MUST be zero, mar$flags.copy MUST be one, and mar$flags.register MUST 2220 be zero. 2222 When an MCS issues a MARS_UNSERV the MARS removes its ATM address 2223 from the server maps for each specified group, deleting any server 2224 maps that end up being null after the operation. 2226 The operation code is then changed to MARS_LEAVE and sends a copy of 2227 the message on ClusterControlVC. This fools the cluster members into 2228 thinking a leaf node as been dropped from the group specified. In the 2229 retransmitted MARS_LEAVE mar$flags.layer3grp MUST be zero, 2230 mar$flags.copy MUST be one, and mar$flags.register MUST be zero. 2232 The MARS retransmits redundant MARS_MSERV and MARS_UNSERV messages 2233 onto ServerControlVC, generates the appropriate MARS_JOIN or or 2234 MARS_LEAVE messages on ClusterControlVC, but takes no further action. 2235 MARS_MIGRATE messages are never repeated in response to redundant 2236 MARS_MSERVs. 2238 The last or only MCS for a group MAY choose to issue a MARS_UNSERV 2239 while the group still has members. When the MARS_UNSERV is processed 2240 by the MARS the 'server map' will be deleted. When the associated 2241 MARS_LEAVE is issued on ClusterControlVC, all cluster members with a 2242 VC open to the MCS for that group will close down the VC (in 2243 accordance with section 5.1.4, since the MCS was their only leaf 2244 node). When cluster members subsequently find they need to transmit 2245 packets to the group, they will begin again with the 2246 MARS_REQUEST/MARS_MULTI sequence to establish a new VC. Since the 2247 MARS will have deleted the server map, this will result in the host 2248 map being return, and the group reverts to being supported by a VC 2249 mesh. 2251 The reverse process is achieved through the MARS_MIGRATE message when 2252 the first MCS registers to support a group. This ensures that 2253 cluster members explicitly dismantle any VC mesh they may have had 2254 up, and re-establish their multicast forwarding path with the MCS as 2255 its termination point. 2257 6.2.3 Registering a Multicast Server (MCS). 2259 Section 5.2.3 describes how endpoints register as cluster members, 2260 and hence get added as leaf nodes to ClusterControlVC. The same 2261 approach is used to register endpoints that intend to provide MCS 2262 support. 2264 Registration with the MARS occurs when an endpoint issues a 2265 MARS_MSERV with mar$flags.register set to one. Upon registration the 2266 endpoint is added as a leaf node to ServerControlVC, and the 2267 MARS_MSERV is returned to the MCS privately. 2269 The MCS retransmits this MARS_MSERV until it confirms that the MARS 2270 has received it (by receiving a copy back, in an analogous way to the 2271 mechanism described in section 5.2.2 for reliably transmitting 2272 MARS_JOINs). 2274 The mar$cmi field in MARS_MSERVs MUST be set to zero by both MCS and 2275 MARS. 2277 An MCS may also choose to de-register, using a MARS_UNSERV with 2278 mar$flags.register set to one. When this occurs the MARS MUST remove 2279 all references to that MCS in all servermaps associated with the 2280 protocol (mar$pro) specified in the MARS_UNSERV, and drop the MCS 2281 from ServerControlVC. 2283 Note that multiple logical MCSs may share the same physical ATM 2284 interface, provided that each MCS uses a separate ATM address (e.g. a 2285 different SEL field in the NSAP format address). In fact, an MCS may 2286 share the ATM interface of a node that is also a cluster member 2287 (either host or router), provided each logical entity has a different 2288 ATM address. 2290 A MARS MUST be capable of handling a multi-entry servermap. However, 2291 the possible use of multiple MCSs registering to support the same 2292 group is a subject for further study. In the absence of an MCS 2293 synchronisation protocol a system administrator MUST NOT allow more 2294 than one logical MCS to register for a given group. 2296 6.2.4 Modified response to MARS_JOIN and MARS_LEAVE. 2298 The existence of MCSs supporting some groups but not others requires 2299 the MARS to modify its distribution of single and block join/leave 2300 updates to cluster members. The MARS also adds two new messages - 2301 MARS_SJOIN and MARS_SLEAVE - for communicating group changes to MCSs 2302 over ServerControlVC. 2304 The MARS_SJOIN and MARS_SLEAVE messages are identical to MARS_JOIN, 2305 with operation codes 18 and 19 (decimal) respectively. 2307 When a cluster member issues MARS_JOIN or MARS_LEAVE for a single 2308 group, the MARS checks to see if the group has an associated server 2309 map. If the specified group does not have a server map the MARS 2310 simply retransmits the MARS_JOIN or MARS_LEAVE on ClusterControlVC. 2312 However, if a server map exists for the group a new set of actions 2313 are taken. 2315 A copy of the MARS_JOIN/LEAVE is made with type MARS_SJOIN or 2316 MARS_SLEAVE as appropriate, and transmitted on ServerControlVC. 2317 This allows the MCS(s) supporting the group to note the new member 2318 and update their data VCs. 2320 The original message is transmitted back to the source cluster 2321 member unchanged, using the VC it arrived on rather than 2322 ClusterControlVC. The mar$flags.punched field MUST be reset to 0 2323 in this message. 2325 (Section 5.2.2 requires cluster members have a mechanism to confirm 2326 the reception of their message by the MARS. For mesh supported 2327 groups, using ClusterControlVC serves dual purpose of providing this 2328 confirmation and distributing group update information. When a group 2329 is MCS supported, there is no reason for all cluster members to 2330 process null join/leave messages on ClusterControlVC, so they are 2331 sent back on the private VC between cluster member and MARS.) 2333 Receipt of a block MARS_JOIN (e.g. from a router coming on-line) or 2334 MARS_LEAVE requires a more complex response. The single 2335 block may simultaneously cover mesh supported and MCS supported 2336 groups. However, cluster members only need to be informed of the 2337 mesh supported groups that the endpoint has joined. Only the MCSs 2338 need to know if the endpoint is joining any MCS supported groups. 2340 The solution is to modify the MARS_JOIN or MARS_LEAVE that is 2341 retransmitted on ClusterControlVC. The following action is taken: 2343 A copy of the MARS_JOIN/LEAVE is made with type MARS_SJOIN or 2344 MARS_SLEAVE as appropriate, and transmitted on ServerControlVC. 2345 This allows the MCS(s) supporting the group to note the membership 2346 change and update their outgoing point to multipoint VCs. 2348 Before transmission on the ClusterControlVC, the original 2349 MARS_JOIN/LEAVE then has its block replaced with a 'hole 2350 punched' set of zero or more pairs. The 'hole punched' 2351 set of pairs covers the entire address range specified 2352 by the original pair, but excludes those 2353 addresses/groups supported by MCSs. 2355 If no 'holes' were punched in the specified block, the original 2356 MARS_JOIN/LEAVE is re-transmitted out on ClusterControlVC 2357 unchanged. Otherwise the following occurs: 2359 The original MARS_JOIN/LEAVE is transmitted back to the source 2360 cluster member unchanged, using the VC it arrived on. The 2361 mar$flags.punched field MUST be reset to 0 in this message. 2363 If the hole-punched set contains 1 or more pair, a 2364 copy of the original MARS_JOIN/LEAVE is transmitted on 2365 ClusterControlVC, carrying the new list. The 2366 mar$flags.punched field MUST be set to 1 in this message. 2368 The mar$flags.punched field is set to ensure the hole-punched copy 2369 is ignored by the message's source when trying to match received 2370 MARS_JOIN/LEAVE messages with ones previously sent (section 2371 5.2.2). 2373 (Appendix A discusses some algorithms for 'hole punching'.) 2375 It is assumed that MCSs use the MARS_SJOINs and MARS_SLEAVEs to 2376 update their own VCs out to the actual group's members. 2378 mar$flags.layer3grp is copied over into the messages transmitted by 2379 the MARS. mar$flags.copy MUST be set to one. 2381 6.2.5 Sequence numbers for ServerControlVC traffic. 2383 In an analogous fashion to the Cluster Sequence Number, the MARS 2384 keeps a Server Sequence Number (SSN) that is incremented after every 2385 transmission on ServerControlVC. The current value of the SSN is 2386 inserted into the mar$msn field of every message the MARS issues that 2387 it believes is destined for an MCS. This includes MARS_MULTIs that 2388 are being returned in response to a MARS_REQUEST from an MCS, and 2389 MARS_REDIRECT_MAP being sent on ServerControlVC. The MARS must check 2390 the MARS_REQUESTs source, and if it is a registered MCS the SSN is 2391 copied into the mar$msn field, otherwise the CSN is copied into the 2392 mar$msn field. 2394 MCSs are expected to track and use the SSNs in an analogous manner to 2395 the way endpoints use the CSN in section 5.1 (to trigger revalidation 2396 of group membership information). 2398 A MARS should be carefully designed to minimise the possibility of 2399 the SSN jumping unecessarily. Under normal operation only MCSs that 2400 are affected by transient link problems will miss mar$msn updates and 2401 be forced to revalidate. If the MARS itself glitches it will be 2402 innundated with requests for a period as every MCS attempts to 2403 revalidate. 2405 6.3 Why global sequence numbers? 2407 The CSN and SSN are global within the context of a given protocol 2408 (e.g. IPv4, mar$pro = 0x800). They count ClusterControlVC and 2409 ServerControlVC activity without reference to the multicast group(s) 2410 involved. This may be perceived as a limitation, because there is no 2411 way for cluster members or multicast servers to isolate exactly which 2412 multicast group they may have missed an update for. An alternative 2413 was to try and provide a per-group sequence number. 2415 Unfortunately per-group sequence numbers are not practical. The 2416 current mechanism allows sequence information to be piggy-backed onto 2417 MARS messages already in transit for other reasons. The ability to 2418 specify blocks of multicast addresses with a single MARS_JOIN or 2419 MARS_LEAVE means that a single message can refer to membership change 2420 for multiple groups simultaneously. A single mar$msn field cannot 2421 provide meaningful information about each group's sequence. Multiple 2422 mar$msn fields would have been unwieldy. 2424 Any MARS or cluster member that supports different protocols MUST 2425 keep separate mapping tables and sequence numbers for each protocol. 2427 6.4 Redundant/Backup MARS Architectures. 2429 If backup MARSs exist for a given cluster then mechanisms are needed 2430 to ensure consistency between their mapping tables and those of the 2431 active, current MARS. 2433 (Cluster members will consider backup MARSs to exist if they have 2434 been configured with a table of MARS addresses, or the regular 2435 MARS_REDIRECT_MAP messages contain a list of 2 or more addresses.) 2437 The definition of an MARS-synchronization protocol is beyond the 2438 current scope of this document, and is expected to be the subject of 2439 further research work. However, the following observations may be 2440 made: 2442 MARS_REDIRECT_MAP messages exist, enabling one MARS to force 2443 endpoints to move to another MARS (e.g. in the aftermath of a MARS 2444 failure, the chosen backup MARS will eventually wish to hand 2445 control of the cluster over to the main MARS when it is 2446 functioning properly again). 2448 Cluster members and MCSs do not need to start up with knowledge of 2449 more than one MARS, provided that MARS correctly issues 2450 MARS_REDIRECT_MAP messages with the full list of MARSs for that 2451 cluster. 2453 Any mechanism for synchronising backup MARSs (and coping with the 2454 aftermath of MARS failures) should be compatible with the cluster 2455 member behaviour described in this document. 2457 7. How an MCS utilises a MARS. 2459 When an MCS supports a multicast group it acts as a proxy cluster 2460 endpoint for the senders to the group. It also behaves in an 2461 analogous manner to a sender, managing a single outgoing point to 2462 multipoint VC to the real group members. 2464 Detailed description of possible MCS architectures are beyond the 2465 scope of this document. This section will outline the main issues. 2467 7.1 Association with a particular Layer 3 group. 2469 When an MCS issues a MARS_MSERV it forces all senders to the 2470 specified layer 3 group to terminate their VCs on the supplied source 2471 ATM address. 2473 The simplest MCS architecture involves taking incoming AAL_SDUs and 2474 simply flipping them back out a single point to multipoint VC. Such 2475 an MCS cannot support more than one group at once, as it has no way 2476 to differentiate between traffic destined for different groups. 2477 Using this architecture, a physical node would provide MCS support 2478 for multiple groups by creating multiple logical instances of the 2479 MCS, each with different ATM Addresses (e.g. a different SEL value in 2480 the node's NSAPA). 2482 A slightly more complex approach would be to add minimal layer 3 2483 specific processing into the MCS. This would look inside the received 2484 AAL_SDUs and determine which layer 3 group they are destined for. A 2485 single instance of such an MCS might register its ATM Address with 2486 the MARS for multiple layer 3 groups, and manage multiple independent 2487 outgoing point to multipoint VCs (one for each group). 2489 When an MCS starts up it MUST register with the MARS as described in 2490 section 6.2.3, identifying the protocol it supports with the mar$pro 2491 field of the MARS_MSERV. This also applies to logical MCSs, even if 2492 they share the same physical ATM interface. This is important so that 2493 the MARS can react to the loss of an MCS when it drops off 2494 ServControlVC. (One consequence is that 'simple' MCS architectures 2495 end up with one ServerControlVC member per group. MCSs with layer 3 2496 specific processing may support multiple groups while still only 2497 registering as one member of ServerControlVC.) 2499 An MCS MUST NOT share the same ATM address as a cluster member, 2500 although it may share the same physical ATM interface. 2502 7.2 Termination of incoming VCs. 2504 An MCS MUST terminate unidirectional VCs in the same manner as a 2505 cluster member. (e.g. terminate on an LLC entity when LLC/SNAP 2506 encapsulation is used, as described in RFC 1755 for unicast 2507 endpoints.) 2509 7.3 Management of outgoing VC. 2511 An MCS MUST establish and manage its outgoing point to multipoint VC 2512 as a cluster member does (section 5.1). 2514 MARS_REQUEST is used by the MCS to establish the initial leaf nodes 2515 for the MCS's outgoing point to multipoint VC. After the VC is 2516 established, the MCS reacts to MARS_SJOINs and MARS_SLEAVEs in the 2517 same way a cluster member reacts to MARS_JOINs and MARS_LEAVEs. 2519 The MCS tracks the Server Sequence Number from the mar$msn fields of 2520 messages from the MARS, and revalidates its outgoing point to 2521 multipoint VC(s) when a sequence number jump occurs. 2523 7.4 Use of a backup MARS. 2525 The MCS uses the same approach to backup MARSs as a cluster member 2526 (section 5.4), tracking MARS_REDIRECT_MAP messages on 2527 ServerControlVC. 2529 8. Support for IP multicast routers. 2531 Multicast routers are required for the propagation of multicast 2532 traffic beyond the constraints of a single cluster (inter-cluster 2533 traffic). (There is a sense in which they are multicast servers 2534 acting at the next higher layer, with clusters, rather than 2535 individual endpoints, as their abstract sources and destinations.) 2537 Multicast routers typically participate in higher layer multicast 2538 routing algorithms and policies that are beyond the scope of this 2539 memo (e.g. DVMRP [5] in the IPv4 environment). 2541 It is assumed that the multicast routers will be implemented over the 2542 same sort of IP/ATM interface that a multicast host would use. Their 2543 IP/ATM interfaces will will register with the MARS as a cluster 2544 members, joining and leaving multicast groups as necessary. As noted 2545 in section 5, multiple logical 'endpoints' may be implemented over a 2546 single physical ATM interface. Routers use this approach to provide 2547 interfaces into each clusters they will be routing between. 2549 The rest of this section will assume a simple IPv4 scenario where the 2550 scope of a cluster has been limited to a particular LIS that is part 2551 of an overlaid IP network. Not all members of the LIS are necessarily 2552 registered cluster members (you may have unicast-only hosts in the 2553 LIS). 2555 8.1 Forwarding into a Cluster. 2557 If the multicast router needs to transmit a packet to a group within 2558 the cluster its IP/ATM interface opens a VC in the same manner as a 2559 normal host would. Once a VC is open, the router watches for 2560 MARS_JOIN and MARS_LEAVE messages and responds to them as a normal 2561 host would. 2563 The multicast router's transmit side MUST implement inactivity timers 2564 to shut down idle outgoing VCs, as for normal hosts. 2566 As with normal host, the multicast router does not need to be a 2567 member of a group it is sending to. 2569 8.2 Joining in 'promiscuous' mode. 2571 Once registered and initialised, the simplest model of IPv4 multicast 2572 router operation is for it to issue a MARS_JOIN encompassing the 2573 entire Class D address space. In effect it becomes 'promiscuous', as 2574 it will be a leaf node to all present and future multipoint VCs 2575 established to IPv4 groups on the cluster. 2577 How a router chooses which groups to propagate outside the cluster is 2578 beyond the scope of this document. 2580 Consistent with RFC 1112, IP multicast routers may retain the use of 2581 IGMP Query and IGMP Report messages to ascertain group membership. 2582 However, certain optimisations are possible, and are described in 2583 section 8.5. 2585 8.3 Forwarding across the cluster. 2587 Under some circumstances the cluster may simply be another hop 2588 between IP subnets that have participants in a multicast group. 2590 [LAN.1] ----- IPmcR.1 -- [cluster/LIS] -- IPmcR.2 ----- [LAN.2] 2592 LAN.1 and LAN.2 are subnets (such as Ethernet) with attached hosts 2593 that are members of group X. 2595 IPmcR.1 and IPmcR.2 are multicast routers with interfaces to the LIS. 2597 A traditional solution would be to treat the LIS as a unicast subnet, 2598 and use tunneling routers. However, this would not allow hosts on the 2599 LIS to participate in the cross-LIS traffic. 2601 Assume IPmcR.1 is receiving packets promiscuously on its LAN.1 2602 interface. Assume further it is configured to propagate multicast 2603 traffic to all attached interfaces. In this case that means the LIS. 2605 When a packet for group X arrives on its LAN.1 interface, IPmcR.1 2606 simply sends the packet to group X on the LIS interface as a normal 2607 host would (Issuing MARS_REQUEST for group X, creating the VC, 2608 sending the packet). 2610 Assuming IPmcR.2 initialised itself with the MARS as a member of the 2611 entire Class D space, it will have been returned as a member of X 2612 even if no other nodes on the LIS were members. All packets for group 2613 X received on IPmcR.2's LIS interface may be retransmitted on LAN.2. 2615 If IPmcR.1 is similarly initialised the reverse process will apply 2616 for multicast traffic from LAN.2 to LAN.1, for any multicast group. 2617 The benefit of this scenario is that cluster members within the LIS 2618 may also join and leave group X at anytime. 2620 8.4 Joining in 'semi-promiscuous' mode. 2622 Both unicast and multicast IP routers have a common problem - 2623 limitations on the number of AAL contexts available at their ATM 2624 interfaces. Being 'promiscuous' in the RFC 1112 sense means that for 2625 every M hosts sending to N groups, a multicast router's ATM interface 2626 will have M*N incoming reassembly engines tied up. 2628 It is not hard to envisage situations where a number of multicast 2629 groups are active within the LIS but are not required to be 2630 propagated beyond the LIS itself. An example might be a distributed 2631 simulation system specifically designed to use the high speed IP/ATM 2632 environment. There may be no practical way its traffic could be 2633 utilised on 'the other side' of the multicast router, yet under the 2634 conventional scheme the router would have to be a leaf to each 2635 participating host anyway. 2637 As this problem occurs below the IP layer, it is worth noting that 2638 'scoping' mechanisms at the IP multicast routing level do not provide 2639 a solution. An IP level scope would still result in the router's ATM 2640 interface receiving traffic on the scoped groups, only to drop it. 2642 In this situation the network administrator might configure their 2643 multicast routers to exclude sections of the Class D address space 2644 when issuing MARS_JOIN(s). Multicast groups that will never be 2645 propagated beyond the cluster will not have the router listed as a 2646 member, and the router will never have to receive (and simply ignore) 2647 traffic from those groups. 2649 Another scenario involves the product M*N exceeding the capacity of a 2650 single router's interface (especially if the same interface must also 2651 support a unicast IP router service). 2653 A network administrator may choose to add a second node, to function 2654 as a parallel IP multicast router. Each router would be configured to 2655 be 'promiscuous' over separate parts of the Class D address space, 2656 thus exposing themselves to only part of the VC load. This sharing 2657 would be completely transparent to IP hosts within the LIS. 2659 Restricted promiscuous mode does not break RFC 1112's use of IGMP 2660 Report messages. If the router is configured to serve a given block 2661 of Class D addresses, it will receive the IGMP Report. If the router 2662 is not configured to support a given block, then the existence of an 2663 IGMP Report for a group in that block is irrelevant to the router. 2664 All routers are able to track membership changes through the 2665 MARS_JOIN and MARS_LEAVE traffic anyway. (Section 8.5 discusses a 2666 better alternative to IGMP within a cluster.) 2668 Mechanisms and reasons for establishing these modes of operation are 2669 beyond the scope of this document. 2671 8.5 An alternative to IGMP Queries. 2673 An unfortunate aspect of IGMP is that it assumes multicasting of IP 2674 packets is a cheap and trivial event at the link layer. As a 2675 consequence, regular IGMP Queries are multicasted by routers to group 2676 224.0.0.1. These queries are intended to trigger IGMP Replies by 2677 cluster members that have layer 3 members of particular groups. 2679 The MARS_GROUPLIST_REQUEST and MARS_GROUPLIST_REPLY messages were 2680 designed to allow routers to avoid actually transmitting IGMP Queries 2681 out into a cluster. 2683 Whenever the router's forwarding engine wishes to transmit an IGMP 2684 query, a MARS_GROUPLIST_REQUEST can be sent to the MARS instead. The 2685 resulting MARS_GROUPLIST_REPLY(s) (described in section 5.3) from the 2686 MARS carry all the information that the router would have ascertained 2687 from IGMP replies. 2689 It is RECOMMENDED that multicast routers utilise this MARS service to 2690 minimise IGMP traffic within the cluster. 2692 By default a MARS_GROUPLIST_REQUEST SHOULD specify the entire address 2693 space (e.g. <224.0.0.0, 239.255.255.255> in an IPv4 environment). 2694 However, routers serving part of the address space (as described in 2695 section 8.4) MAY choose to issue MARS_GROUPLIST_REQUESTs that specify 2696 only the subset of the address space they are serving. 2698 (On the surface it would also seem useful for multicast routers to 2699 track MARS_JOINs and MARS_LEAVEs that arrive with mar$flags.layer3grp 2700 set. These might be used in lieu of IGMP Reports, to provide the 2701 router with timely indication that a new layer 3 group member exists 2702 within the cluster. However, this only works on VC mesh supported 2703 groups, and is therefore NOT recommended). 2705 Appendix B discusses less elegant mechanisms for reducing the impact 2706 of IGMP traffic within a cluster, on the assumption that the IP/ATM 2707 interfaces to the cluster are being used by un-optimised IP 2708 multicasting code. 2710 8.6 CMIs across multiple interfaces. 2712 The Cluster Member ID is only unique within the Cluster managed by a 2713 given MARS. On the surface this might appear to leave us with a 2714 problem when a multicast router is routing between two or more 2715 Clusters using a single physical ATM interface. The router will 2716 register with two or more MARSs, and thereby acquire two or more 2717 independent CMI's. Given that each MARS has no reason to synchronise 2718 their CMI allocations, it is possible for a host in one cluster to 2719 have the same CMI has the router's interface to another Cluster. How 2720 does the router distinguish between its own reflected packets, and 2721 packets from that other host? 2723 The answer lies in the fact that routers (and hosts) actually 2724 implement logical IP/ATM interfaces over a single physical ATM 2725 interface. Each logical interface will have a unique ATM Address (eg. 2726 an NSAP with different SELector fields, one for each logical 2727 interface). 2729 Each logical IP/ATM interface is configured with the address of a 2730 single MARS, attaches to only one cluster, and so had only one CMI to 2731 worry about. Each of the MARSs that the router is registered with 2732 will have been given a different ATM Address (corresponding to the 2733 different logical IP/ATM interfaces) in each registration MARS_JOIN. 2735 When hosts in a cluster add the router as a leaf node, they'll 2736 specify the ATM Address of the appropriate logical IP/ATM interface 2737 on the router in the L_MULTI_ADD message. Thus, each logical IP/ATM 2738 interface will only have to check and filter on CMIs assigned by its 2739 own MARS. 2741 In essence the cluster differentiation is achieved by ensuring that 2742 logical IP/ATM interfaces are assigned different ATM Addresses. 2744 9. Multiprotocol applications of the MARS and MARS clients. 2746 A deliberate attempt has been made to describe the MARS and 2747 associated mechanisms in a manner independent of a specific higher 2748 layer protocol being run over the ATM cloud. The immediate 2749 application of this document will be in an IPv4 environment, and this 2750 is reflected by the focus of key examples. However, the mar$pro.type 2751 and mar$pro.snap fields in every MARS control message allow any 2752 higher layer protocol that has a 'short form' or 'long form' of 2753 protocol identification (section 4.3) to be supported by a MARS. 2755 Every MARS MUST implement entirely separate logical mapping tables 2756 and support. Every cluster member must interpret messages from the 2757 MARS in the context of the protocol type that the MARS message refers 2758 to. 2760 Every MARS and MARS client MUST treat Cluster Member IDs in the 2761 context of the protocol type carried in the MARS message or data 2762 packet containing the CMI. 2764 For example, IPv6 has been allocated an Ethertype of 0x86DD. This 2765 means the 'short form' of protocol identification must be used in the 2766 MARS control messages and the data path encapsulation (section 5.5). 2767 An IPv6 multicasting client sets the mar$pro.type field of every MARS 2768 message to 0x86DD. When carrying IPv6 addresses the mar$spln and 2769 mar$tpln fields are either 0 (for null or non-existent information) 2770 or 16 (for the full IPv6 address). 2772 Following the rules in section 5.5, an IPv6 data packet is 2773 encapsulated as: 2775 [0xAA-AA-03][0x00-00-5E][0x00-01][pkt$cmi][0x86DD][IPv6 packet] 2777 A host or endpoint interface that is using the same MARS to support 2778 multicasting needs of multiple protocols MUST not assume their CMI 2779 will be the same for each protocol. 2781 10. Supplementary parameter processing. 2783 The mar$extoff field in the [Fixed header] indicates whether 2784 supplementary parameters are being carried by a MARS control message. 2785 This mechanism is intended to enable the addition of new 2786 functionality to the MARS protocol in later documents. 2788 Supplementary parameters are conveyed as a list of TLV (type, length, 2789 value) encoded information elements. The TLV(s) begin on the first 2790 32 bit boundary following the [Addresses] field in the MARS control 2791 message (e.g. after mar$tsa.N in a MARS_MULTI, after mar$max.N in a 2792 MARS_JOIN, etc). 2794 10.1 Interpreting the mar$extoff field. 2796 If the mar$extoff field is non-zero it indicates that a list of one 2797 or more TLVs have been appended to the MARS message. The first TLV 2798 is found by treating mar$extoff as an unsigned integer representing 2799 an offset (in octets) from the beginning of the MARS message (the MSB 2800 of the mar$hrd field). 2802 As TLVs are 32 bit aligned the bottom 2 bits of mar$extoff are also 2803 reserved. A receiver MUST mask off these two bits before calculating 2804 the octet offset to the TLV list. A sender MUST set these two bits 2805 to zero. 2807 If mar$extoff is zero no TLVs have been appended. 2809 10.2 The format of TLVs. 2811 When they exist, TLVs begin on 32 bit boundaries, are multiples of 32 2812 bits in length, and form a sequential list terminated by a NULL TLV. 2814 The TLV structure is: 2816 [Type - 2 octets][Length - 2 octets][Value - n*4 octets] 2818 The Type subfield indicates how the contents of the Value subfield 2819 are to be interpreted. 2821 The Length subfield indicates the number of VALID octets in the Value 2822 subfield. Valid octets in the Value subfield start immediately after 2823 the Length subfield. The offset (in octets) from the start of this 2824 TLV to the start of the next TLV in the list is given by the 2825 following formula: 2827 offset = (length + 4 + ((4-(length & 3)) % 4)) 2829 (where % is the modulus operator) 2831 The Value subfield is padded with 0, 1, 2, or 3 octets to ensure the 2832 next TLV is 32 bit aligned. The padded locations MUST be set to zero. 2834 (For example, a TLV that needed only 5 valid octets of information 2835 would be 12 octets long. The Length subfield would hold the value 5, 2836 and the Value subfield would be padded out to 8 bytes. The 5 valid 2837 octets of information begin at the first octet of the Value 2838 subfield.) 2840 The Type subfield is formatted in the following way: 2842 | 1st octet | 2nd octet | 2843 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 2844 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2845 | x | y | 2846 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 2848 The most significant 2 bits (Type.x) determine how a recipient should 2849 behave when it doesn't recognise the TLV type indicated by the lower 2850 14 bits (Type.y). The required behaviours are: 2852 Type.x = 0 Skip the TLV, continue processing the list. 2853 Type.x = 1 Stop processing, silently drop the MARS message. 2854 Type.x = 2 Stop processing, drop message, give error indication. 2855 Type.x = 3 Reserved. (currently treat as x = 0) 2857 (The error indication generated when Type.x = 2 SHOULD be logged in 2858 some locally significant fashion. Consequential MARS message activity 2859 in response to such an error condition will be defined in future 2860 documents.) 2862 The TLV type space (Type.y) is further subdivided to encourage use 2863 outside the IETF. 2865 0 Null TLV. 2866 0x0001 - 0x0FFF Reserved for the IETF. 2867 0x1000 - 0x11FF Allocated to the ATM Forum. 2868 0x1200 - 0x37FF Reserved for the IETF. 2869 0x3800 - 0x3FFF Experimental use. 2871 10.3 Processing MARS messages with TLVs. 2873 Supplementary parameters act as modifiers to the basic behaviour 2874 specified by the mar$op field of any given MARS message. 2876 If a MARS message arrives with a non-zero mar$extoff field its TLV 2877 list MUST be parsed before handling the MARS message in accordance 2878 with the mar$op value. Unrecognised TLVs MUST be handled as required 2879 by their Type.x value. 2881 How TLVs modify basic MARS operations will be mar$op and TLV 2882 specific. 2884 10.4 Initial set of TLV elements. 2886 Conformance with this document only REQUIRES the recognition of one 2887 TLV, the Null TLV. This terminates a list of TLVs, and MUST be 2888 present if mar$extoff is non-zero in a MARS message. It MAY be the 2889 only TLV present. 2891 The Null TLV is coded as: 2893 [0x00-00][0x00-00] 2895 Future documents will describe the formats, contents, and 2896 interpretations of additional TLVs. The minimal parsing requirements 2897 imposed by this document are intended to allow conformant MARS and 2898 MARS client implementations to deal gracefully and predictably with 2899 future TLV developments. 2901 11. Key Decisions and open issues. 2903 The key decisions this document proposes: 2905 A Multicast Address Resolution Server (MARS) is proposed to co- 2906 ordinate and distribute mappings of ATM endpoint addresses to 2907 arbitrary higher layer 'multicast group addresses'. The specific 2908 case of IPv4 multicast is used as the example. 2910 The concept of 'clusters' is introduced to define the scope of a 2911 MARS's responsibility, and the set of ATM endpoints willing to 2912 participate in link level multicasting. 2914 A MARS is described with the functionality required to support 2915 intra-cluster multicasting using either VC meshes or ATM level 2916 multicast servers (MCSs). 2918 LLC/SNAP encapsulation of MARS control messages allows MARS and 2919 ATMARP traffic to share VCs, and allows partially co-resident MARS 2920 and ATMARP entities. 2922 New message types: 2924 MARS_JOIN, MARS_LEAVE, MARS_REQUEST. Allow endpoints to join, 2925 leave, and request the current membership list of multicast 2926 groups. 2928 MARS_MULTI. Allows multiple ATM addresses to be returned by the 2929 MARS in response to a MARS_REQUEST. 2931 MARS_MSERV, MARS_UNSERV. Allow multicast servers to register 2932 and deregister themselves with the MARS. 2934 MARS_SJOIN, MARS_SLEAVE. Allow MARS to pass on group membership 2935 changes to multicast servers. 2937 MARS_GROUPLIST_REQUEST, MARS_GROUPLIST_REPLY. Allow MARS to 2938 indicate which groups have actual layer 3 members. May be used 2939 to support IGMP in IPv4 environments, and similar functions in 2940 other environments. 2942 MARS_REDIRECT_MAP. Allow MARS to specify a set of backup MARS 2943 addresses. 2945 MARS_MIGRATE. Allows MARS to force cluster members to shift 2946 from VC mesh to MCS based forwarding tree in single operation. 2948 'wild card' MARS mapping table entries are possible, where a 2949 single ATM address is simultaneously associated with blocks of 2950 multicast group addresses. 2952 For the MARS protocol mar$op.version = 0. The complete set of MARS 2953 control messages and mar$op.type values is: 2955 1 MARS_REQUEST 2956 2 MARS_MULTI 2957 3 MARS_MSERV 2958 4 MARS_JOIN 2959 5 MARS_LEAVE 2960 6 MARS_NAK 2961 7 MARS_UNSERV 2962 8 MARS_SJOIN 2963 9 MARS_SLEAVE 2964 10 MARS_GROUPLIST_REQUEST 2965 11 MARS_GROUPLIST_REPLY 2966 12 MARS_REDIRECT_MAP 2967 13 MARS_MIGRATE 2969 A number of issues are left open at this stage, and are likely to be 2970 the subject of on-going research and additional documents that build 2971 upon this one. 2973 The specified endpoint behaviour allows the use of 2974 redundant/backup MARSs within a cluster. However, no 2975 specifications yet exist on how these MARSs co-ordinate amongst 2976 themselves. (The default is to only have one MARS per cluster.) 2978 The specified endpoint behaviour and MARS service allows the use 2979 of multiple MCSs per group. However, no specifications yet exist 2980 on how this may be used, or how these MCSs co-ordinate amongst 2981 themselves. Until futher work is done on MCS co-ordination 2982 protocols the default is to only have one MCS per group. 2984 The MARS relies on the cluster member dropping off 2985 ClusterControlVC if the cluster member dies. It is not clear if 2986 additional mechanisms are needed to detect and delete 'dead' 2987 cluster members. 2989 Supporting layer 3 'broadcast' as a special case of multicasting 2990 (where the 'group' encompasses all cluster members) has not been 2991 explicitly discussed. 2993 Supporting layer 3 'unicast' as a special case of multicasting 2994 (where the 'group' is a single cluster member, identified by the 2995 cluster member's unicast protocol address) has not been explicitly 2996 discussed. 2998 The future development of ATM Group Addresses and Leaf Initiated 2999 Join to ATM Forum's UNI specification has not been addressed. 3000 (However, the problems identified in this document with respect to 3001 VC scarcity and impact on AAL contexts will not be fixed by such 3002 developments in the signalling protocol.) 3004 Possible modifications to the interpretation of the mar$hrdrsv and 3005 mar$hrd fields in the Fixed header, based on different values for 3006 mar$op.version, are for further study. 3008 Security Consideration 3010 Security consideration are not addressed in this document. 3012 Acknowledgments 3014 The discussions within the IP over ATM Working Group have helped 3015 clarify the ideas expressed in this document. John Moy (Cascade 3016 Communications Corp.) initially suggested the idea of wild-card 3017 entries in the ARP Server. Drew Perkins (Fore Systems) provided 3018 rigorous and useful critique of early proposed mechanisms for 3019 distributing and validating group membership information. Susan 3020 Symington (and co-workers at MITRE Corp., Don Chirieleison, Rich 3021 Verjinski, and Bill Barns) clearly articulated the need for multicast 3022 server support, proposed a solution, and challenged earlier block 3023 join/leave mechanisms. John Shirron (Fore Systems) provided useful 3024 improvements on my original revalidation procedures. 3026 Susan Symington and Bryan Gleeson (Adaptec) independently championed 3027 the need for the service provided by MARS_GROUPLIST_REQUEST/REPLY. 3028 The new encapsulation scheme arose from WG discussions, captured by 3029 Bryan Gleeson in an interim Internet Draft (with Keith McCloghrie 3030 (Cisco), Andy Malis (Ascom Nexion), and Andrew Smith (Bay Networks) 3031 as key contributors). James Watt (Newbridge) and Joel Halpern 3032 (Newbridge) motivated the development of a more multiprotocol MARS 3033 control message format, evolving it away from its original ATMARP 3034 roots. They also motivated the development of Type #1 and Type #2 3035 data path encapsulations. At the last minute Rajesh Talpade (Georgia 3036 Tech) clarified the need for the MARS_MIGRATE function. 3038 Finally, Jim Rubas (IBM) supplied the MARS pseudo-code in Appendix F 3039 and provided detailed proof-reading in the latter stages of the 3040 documents development. 3042 Author's Address 3044 Grenville Armitage 3045 Bellcore, 445 South Street 3046 Morristown, NJ, 07960 3047 USA 3049 Email: gja@thumper.bellcore.com 3050 Ph. +1 201 829 2635 3052 References 3053 [1] S. Deering, "Host Extensions for IP Multicasting", RFC 1112, 3054 Stanford University, August 1989. 3056 [2] Heinanen, J., "Multiprotocol Encapsulation over ATM Adaption 3057 Layer 5", RFC 1483, USC/Information Science Institute, July 1993. 3059 [3] Laubach, M., "Classical IP and ARP over ATM", RFC1577, Hewlett- 3060 Packard Laboratories, December 1993 3062 [4] ATM Forum, "ATM User Network Interface (UNI) Specification 3063 Version 3.1", ISBN 0-13-393828-X, Prentice Hall, Englewood Cliffs, 3064 NJ, June 1995. 3066 [5] D. Waitzman, C. Partridge, S. Deering, "Distance Vector Multicast 3067 Routing Protocol", RFC 1075, November 1988. 3069 [6] M. Perez, F. Liaw, D. Grossman, A. Mankin, E. Hoffman, A. Malis, 3070 "ATM Signaling Support for IP over ATM", RFC 1755, February 1995. 3072 [7] M. Borden, E. Crawley, B. Davie, S. Batsell, "Integration of 3073 Real-time Services in an IP-ATM Network Architecture.", RFC 1821, 3074 August 1995. 3076 [8] ATM Forum, "ATM User-Network Interface Specification Version 3077 3.0", Englewood Cliffs, NJ: Prentice Hall, September 1993 3079 Appendix A. Hole punching algorithms. 3081 Implementations are entirely free to comply with the body of this 3082 memo in any way they see fit. This appendix is purely for 3083 clarification. 3085 A MARS implementation might pre-construct a set of pairs 3086 (P) that reflects the entire Class D space, excluding any addresses 3087 currently supported by multicast servers. The field of the 3088 first pair MUST be 224.0.0.0, and the field of the last pair 3089 must be 239.255.255.255. The first and last pair may be the same. 3090 This set is updated whenever a multicast server registers or 3091 deregisters. 3093 When the MARS must perform 'hole punching' it might consider the 3094 following algorithm: 3096 Assume the MARS_JOIN/LEAVE received by the MARS from the cluster 3097 member specified the block . 3099 Assume Pmin(N) and Pmax(N) are the and fields from the 3100 Nth pair in the MARS's current set P. 3102 Assume set P has K pairs. Pmin(1) MUST equal 224.0.0.0, and 3103 Pmax(M) MUST equal 239.255.255.255. (If K == 1 then no hole 3104 punching is required). 3106 Execute pseudo-code: 3108 create copy of set P, call it set C. 3110 index1 = 1; 3111 while (Pmax(index1) <= Emin) 3112 index1++; 3114 index2 = K; 3115 while (Pmin(index2) >= Emax) 3116 index2--; 3118 if (index1 > index2) 3119 Exit, as the hole-punched set is null. 3121 if (Pmin(index1) < Emin) 3122 Cmin(index1) = Emin; 3124 if (Pmax(index2) > Emax) 3125 Cmax(index2) = Emax; 3127 Set C is the required 'hole punched' set of address blocks. 3129 The resulting set C retains all the MARS's pre-constructed 'holes' 3130 covering the multicast servers, but will have been pruned to cover 3131 the section of the Class D space specified by the originating host's 3132 values. 3134 The host end should keep a table, H, of open VCs in ascending order 3135 of Class D address. 3137 Assume H(x).addr is the Class address associated with VC.x. 3138 Assume H(x).addr < H(x+1).addr. 3140 The pseudo code for updating VCs based on an incoming JOIN/LEAVE 3141 might be: 3143 x = 1; 3144 N = 1; 3146 while (x < no.of VCs open) 3147 { 3148 while (H(x).addr > max(N)) 3149 { 3150 N++; 3151 if (N > no. of pairs in JOIN/LEAVE) 3152 return(0); 3153 } 3155 if ((H(x).addr <= max(N) && 3156 ((H(x).addr >= min(N)) 3157 perform_VC_update(); 3158 x++; 3159 } 3161 Appendix B. Minimising the impact of IGMP in IPv4 environments. 3163 Implementing any part of this appendix is not required for 3164 conformance with this document. It is provided solely to document 3165 issues that have been identified. 3167 The intent of section 5.1 is for cluster members to only have 3168 outgoing point to multipoint VCs when they are actually sending data 3169 to a particular multicast groups. However, in most IPv4 environments 3170 the multicast routers attached to a cluster will periodically issue 3171 IGMP Queries to ascertain if particular groups have members. The 3172 current IGMP specification attempts to avoid having every group 3173 member respond by insisting that each group member wait a random 3174 period, and responding if no other member has responded before them. 3175 The IGMP reply is sent to the multicast address of the group being 3176 queried. 3178 Unfortunately, as it stands the IGMP algorithm will be a nuisance for 3179 cluster members that are essentially passive receivers within a given 3180 multicast group. It is just as likely that a passive member, with no 3181 outgoing VC already established to the group, will decide to send an 3182 IGMP reply - causing a VC to be established where there was no need 3183 for one. This is not a fatal problem for small clusters, but will 3184 seriously impact on the ability of a cluster to scale. 3186 The most obvious solution is for routers to use the 3187 MARS_GROUPLIST_REQUEST and MARS_GROUPLIST_REPLY messages, as 3188 described in section 8.5. This would remove the regular IGMP Queries, 3189 resulting in cluster members only sending an IGMP Report when they 3190 first join a group. 3192 Alternative solutions do exist. One would be to modify the IGMP reply 3193 algorithm, for example: 3195 If the group member has VC open to the group proceed as per RFC 3196 1112 (picking a random reply delay between 0 and 10 seconds). 3198 If the group member does not have VC already open to the group, 3199 pick random reply delay between 10 and 20 seconds instead, and 3200 then proceed as per RFC 1112. 3202 If even one group member is sending to the group at the time the IGMP 3203 Query is issued then all the passive receivers will find the IGMP 3204 Reply has been transmitted before their delay expires, so no new VC 3205 is required. If all group members are passive at the time of the IGMP 3206 Query then a response will eventually arrive, but 10 seconds later 3207 than under conventional circumstances. 3209 The preceeding solution requires re-writing existing IGMP code, and 3210 implies the ability of the IGMP entity to ascertain the status of VCs 3211 on the underlying ATM interface. This is not likely to be available 3212 in the short term. 3214 One short term solution is to provide something like the preceeding 3215 functionality with a 'hack' at the IP/ATM driver level within cluster 3216 members. Arrange for the IP/ATM driver to snoop inside IP packets 3217 looking for IGMP traffic. If an IGMP packet is accepted for 3218 transmission, the IP/ATM driver can buffer it locally if there is no 3219 VC already active to that group. A 10 second timer is started, and if 3220 an IGMP Reply for that group is received from elsewhere on the 3221 cluster the timer is reset. If the timer expires, the IP/ATM driver 3222 then establishes a VC to the group as it would for a normal IP 3223 multicast packet. 3225 Some network implementors may find it advantageous to configure a 3226 multicast server to support the group 224.0.0.1, rather than rely on 3227 a mesh. Given that IP multicast routers regularly send IGMP queries 3228 to this address, a mesh will mean that each router will permanently 3229 consume an AAL context within each cluster member. In clusters served 3230 by multiple routers the VC load within switches in the underlying ATM 3231 network will become a scaling problem. 3233 Finally, if a multicast server is used to support 224.0.0.1, another 3234 ATM driver level hack becomes a possible solution to IGMP Reply 3235 traffic. The ATM driver may choose to grab all outgoing IGMP packets 3236 and send them out on the VC established for sending to 224.0.0.1, 3237 regardless of the Class D address the IGMP message was actually for. 3238 Given that all hosts and routers must be members of 224.0.0.1, the 3239 intended recipients will still receive the IGMP Replies. The negative 3240 impact is that all cluster members will receive the IGMP Replies. 3242 Appendix C. Further comments on 'Clusters'. 3244 The cluster concept was introduced in section 1 for two reasons. The 3245 more well known term of Logical IP Subnet is both very IP specific, 3246 and constrained to unicast routing boundaries. As the architecture 3247 described in this document may be re-used in non-IP environments a 3248 more neutral term was needed. As the needs of multicasting are not 3249 always bound by the same scopes as unicasting, it was not immediately 3250 obvious that apriori limiting ourselves to LISs was beneficial in the 3251 long term. 3253 It must be stressed that Clusters are purely an administrative being. 3254 You choose their size (i.e. the number of endpoints that register 3255 with the same MARS) based on your multicasting needs, and the 3256 resource consumption you are willing to put up with. The larger the 3257 number of ATM attached hosts you require multicast support for, the 3258 more individual clusters you might choose to establish (along with 3259 multicast routers to provide inter-cluster traffic paths). 3261 Given that not all the hosts in any given LIS may require multicast 3262 support, it becomes conceivable that you might assign a single MARS 3263 to support hosts from across multiple LISs. In effect you have a 3264 cluster covering multiple LISs, and have achieved 'cut through' 3265 routing for multicast traffic. Under these circumstances increasing 3266 the geographical size of a cluster might be considered a good thing. 3268 However, practical considerations limit the size of clusters. Having 3269 a cluster span multiple LISs may not always be a particular 'win' 3270 situation. As the number of multicast capable hosts in your LISs 3271 increases it becomes more likely that you'll want to constrain a 3272 cluster's size and force multicast traffic to aggregate at multicast 3273 routers scattered across your ATM cloud. 3275 Finally, multi-LIS clusters require a degree of care when deploying 3276 IP multicast routers. Under the Classical IP model you need unicast 3277 routers on the edges of LISs. Under the MARS architecture you only 3278 need multicast routers at the edges of clusters. If your cluster 3279 spans multiple LISs, then the multicast routers will perceive 3280 themselves to have a single interface that is simultaneously attached 3281 to multiple unicast subnets. Whether this situation will work depends 3282 on the inter-domain multicast routing protocols you use, and your 3283 multicast router's ability to understand the new relationship between 3284 unicast and multicast topologies. 3286 In the absence of futher research in this area, networks deployed in 3287 conformance to this document MUST make their IP cluster and IP LIS 3288 coincide, so as to avoid these complications. 3290 Appendix D. TLV list parsing algorithm. 3292 The following pseudo-code represents how the TLV list format 3293 described in section 10 could be handled by a MARS or MARS client. 3295 list = (mar$extoff & 0xFFFC); 3297 if (list == 0) exit; 3299 list = list + message_base; 3301 while (list->Type.y != 0) 3302 { 3303 switch (list->Type.y) 3304 { 3305 default: 3306 { 3307 if (list->Type.x == 0) break; 3309 if (list->Type.x == 1) exit; 3311 if (list->Type.x == 2) log-error-and-exit; 3312 } 3314 [...other handling goes here..] 3316 } 3318 list += (list->Length + 4 + ((4-(list->Length & 3)) % 3319 4)); 3321 } 3323 return; 3325 Appendix E. Summary of timer values. 3327 This appendix summarises of various timers or limits mentioned in the 3328 main body of the document. Values are specified in the following 3329 format: [x, y, z] indicating a minimum value of x, a recommended 3330 value of y, and a maximum value of z. A '-' will indicate that a 3331 category has no value specified. Values in minutes are followed by 3332 'min', values in seconds are followed by 'sec'. 3334 Idle time for MARS - MARS client pt to pt VC: 3335 [1 min, 20 min, -] 3337 Idle time for multipoint VCs from client. 3338 [1 min, 20 min, -] 3340 Allowed time between MARS_MULTI components. 3341 [-, -, 10 sec] 3343 Initial random L_MULTI_RQ/ADD retransmit timer range. 3344 [5 sec, -, 10 sec] 3346 Random time to set VC_revalidate flag. 3347 [1 sec, -, 10 sec] 3349 MARS_JOIN/LEAVE retransmit interval. 3350 [5 sec, 10 sec, -] 3352 MARS_JOIN/LEAVE retransmit limit. 3353 [-, -, 5] 3355 Random time to re-register with MARS. 3356 [1 sec, -, 10 sec] 3358 Force wait if MARS re-registration is looping. 3359 [1 min, -, -] 3361 Transmission interval for MARS_REDIRECT_MAP. 3362 [1 min, 1 min, 2 min] 3364 Limit for client to miss MARS_REDIRECT_MAPs. 3365 [-, -, 4 min] 3367 Appendix F. Pseudo code for MARS operation. 3369 Implementations are entirely free to comply with the body of this 3370 memo in any way they see fit. This appendix is purely for possible 3371 clarification. 3373 A MARS implementation might be built along the lines suggested in 3374 this pseudo-code. 3376 1. Main 3378 1.1 Initilization 3380 Define a server list as the list of leaf nodes 3381 on ServerControlVC. 3382 Define a cluster list as the list of leaf nodes 3383 on ClusterControlVC. 3384 Define a host map as the list of hosts that are 3385 members of a group. 3386 Define a server map as the list of hosts (MCSs) 3387 that are serving a group. 3388 Read config file. 3389 Allocate message queues. 3390 Allocate internal tables. 3391 Set up passive open VC connection. 3392 Set up redirect_map timer. 3393 Establish logging. 3395 1.2 Message Processing 3397 Forever { 3398 If the message has a TLV then { 3399 If TLV is unsupported then { 3400 process as defined in TLV type field. 3401 } /* unknown TLV */ 3402 } /* TLV present */ 3403 Place incoming message in the queue. 3404 For (all messages in the queue) { 3405 If the message is not a JOIN/LEAVE/MSERV/UNSERV with 3406 mar$flags.register == 1 then { 3407 If the message source is (not a member of server list) && 3408 (not a member of cluster list) then { 3409 Drop the message silently. 3410 } 3411 } 3412 If (mar$pro.type is not supported) or 3413 (the ATM source address is missing) then { 3414 Continue. 3416 } 3417 Determine type of message. 3418 If an ERR_L_RELEASE arrives on ClusterControlVC then { 3419 Remove the endpoints ATM address from all groups 3420 for which it has joined. 3421 Release the CMI. 3422 Continue. 3423 } /* error on CCVC */ 3424 Call specific message handling routine. 3425 If redirect_map timer pops { 3426 Call MARS_REDIRECT_MAP message handling routine. 3427 } /* redirect timer pop */ 3428 } /* all msgs in the queue */ 3429 } /* forever loop */ 3431 2. Message Handler 3433 2.1 Messages: 3435 - MARS_REQUEST 3437 Indicate no MARS_MULTI support of TLV. 3438 If the supported TLV is not NULL then { 3439 Indicate MARS_MULTI support of TLV. 3440 Process as required. 3441 } else { /* TLV NULL */ 3442 Indicate message to be sent on Private VC. 3443 If the message source is a member of server list then { 3444 If the group has a non-null host map then { 3445 Call MARS_MULTI with the host map for the group. 3446 } else { /* no group */ 3447 Call MARS_NAK message routine. 3448 } /* no group */ 3449 } else { /* source is cluster list */ 3450 If the group has a non-null server map then { 3451 Call MARS_MULTI with the server map for the group. 3452 } else { /* cluster member but no server map */ 3453 If the group has a non-null host map then { 3454 Call MARS_MULTI with the host map for the group. 3455 } else { /* no group */ 3456 Call MARS_NAK message routine. 3457 } /* no group */ 3458 } /* cluster member but no server map */ 3459 } /* source is a cluster list */ 3460 } /* TLV NULL */ 3461 If a message exists then { 3462 Send message as indicated. 3463 } 3464 Return. 3466 - MARS_MULTI 3468 Construct a MARS_MULTI for the specified map. 3469 If the param indicates TLV support then { 3470 Process the TLV as required. 3471 } 3472 Return. 3474 - MARS_JOIN 3476 If (mar$flags.copy != 0) silently ignore the message. 3477 If more than a single pair is specified then 3478 ignore the 2nd and subsequent pairs. 3479 Indicate message to be sent on private VC. 3480 If (mar$flags.register == 1) then { 3481 If the node is already a registered member of the cluster 3482 associated with protocol type then { /*previous register*/ 3483 Copy the existing CMI into the MARS_JOIN. 3484 } else { /* new register */ 3485 Add the node to ClusterControlVC. 3486 Add the node to cluster list. 3487 mar$cmi = obtain CMI. 3488 } /* new register */ 3489 } else { /* not a register */ 3490 If the message source is in server map then { 3491 Drop the message silently. 3492 Indicate no message to be sent. 3493 } else { 3494 If the first encompasses any group with 3495 a server map then { 3496 Call the Modified JOIN/LEAVE Processing routine. 3497 Indicate no message to be sent. 3498 } else { /* server map does not exist */ 3499 Update internal tables. 3500 Indicate message to be sent on ClusterControlVC. 3501 } /* server map does not exist */ 3502 } 3503 } /* not a register */ 3504 mar$flags.copy = 1. 3505 Send message as indicated. 3506 Return. 3508 - MARS_LEAVE 3510 If (mar$flags.copy != 0) silently ignore the message. 3511 Indicate message to be sent on ClusterControlVC. 3513 If (mar$flags.register == 1) then { /* deregistration */ 3514 Update internal tables to remove the member's ATM addr 3515 from all groups it has joined. 3516 Drop the endpoint from ClusterControlVC. 3517 Drop the endpoint from cluster list. 3518 Release the CMI. 3519 Indicate message to be sent on Private VC. 3520 } else { /* not a deregistration */ 3521 If the first encompasses any group with 3522 a server map then { 3523 Call the Modified JOIN/LEAVE Processing routine. 3524 Indicate no message to be sent. 3525 } else { /* server map does not exist */ 3526 Update internal tables. 3527 Indicate message to be sent on ClusterControlVC. 3528 } 3529 } /* not a deregistration */ 3530 If a message exists then { 3531 mar$flags.copy = 1. 3532 Send message as indicated. 3533 } 3534 Return. 3536 - MARS_MSERV 3538 If (mar$flags.register == 1) then { /* server register */ 3539 Add the endpoint as a leaf node to ServerControlVC. 3540 Add the endpoint to the server list. 3541 Indicate the message to be sent on Private VC. 3542 mar$cmi = 0. 3543 } else { /* not a register */ 3544 If the source has not registered then { 3545 Drop and ignore the message. 3546 Indicate no message to be sent. 3547 } else { /* source is registered */ 3548 Add the server ATM addr to server map for group. 3549 Indicate the message to be sent on ServerControlVC. 3550 Send message as indicated. 3551 Make a copy of the message. 3552 Indicate the message to be sent on ClusterControlVC. 3553 If new server map was just created { 3554 Construct MARS_MIGRATE, with MCS as target. 3555 } else { 3556 Change the op code to MARS_JOIN. 3557 mar$flags.layer3grp = 0. 3558 mar$flags.copy = 1. 3559 } /* new server map */ 3561 } /* source is registered */ 3563 } /* not a register */ 3565 If a message exists then { 3566 Send message as indicated. 3567 } 3568 Return. 3570 - MARS_UNSERV 3572 If (mar$flags.register == 1) then { /* deregister */ 3573 Remove the ATM addr of the MCS from all server maps. 3574 If a server map becomes null then delete it. 3575 Remove the endpoint as a leaf of ServerControlVC. 3576 Remove the endpoint from server list. 3577 Indicate the message to be sent on Private VC. 3578 } else { /* not a deregister */ 3579 If the source is not a member of server list then { 3580 Drop and ignore the message. 3581 Indicate no message to be sent. 3582 } else { /* source is registered */ 3583 Remove ATM addr of the MCS from each server map indicated. 3584 If a server map is null then delete it. 3585 Indicate the message to be sent on ServerControlVC. 3586 Send message as indicated. 3587 Make a copy of the message. 3588 Change the op code to MARS_LEAVE. 3589 Indicate the message (copy) to be sent on ClusterControlVC. 3590 mar$flags.layer3grp = 0; 3591 mar$flags.copy = 1. 3592 } /* source is registered */ 3593 } /* not a deregister */ 3594 If a message exists then { 3595 Send message as indicated. 3596 } 3597 Return. 3599 - MARS_NAK 3601 Build command. 3602 Return. 3604 - MARS_GROUPLIST_REQUEST 3606 If (mar$pnum != 1) then Return. 3607 Call MARS_GROUPLIST_REPLY with the range and output VC. 3609 Return. 3611 - MARS_GROUPLIST_REPLY 3613 Build command for specified range. 3614 Indicate message to be sent on specified VC. 3615 Send message as indicated. 3616 Return. 3618 - MARS_REDIRECT_MAP 3620 Include the MARSs own address in the message. 3621 If there are backup MARSs then include their addresses. 3622 Indicate MARS_REDIRECT_MAP is to be sent on ClusterControlVC. 3623 Send message back as indicated. 3624 Return. 3626 3. Send Message Handler 3628 If the message is going out ClusterControlVC then { 3629 mar$msn = obtain a CSN 3630 } 3631 If the message is going out ServerControlVC then { 3632 mar$msn = obtain a SSN 3633 } 3634 Return. 3636 4. Number Generator 3638 4.1 Cluster Sequence Number 3640 Generate the next sequence number. 3641 Return. 3643 4.2 Server Sequence Number 3645 Generate the next sequence number. 3646 Return. 3648 4.3 CMI 3650 CMIs are allocated uniquely per registered cluster member 3651 within the context of a particular layer 3 protocol type. 3652 A single node may register multiple times if it supports 3653 multiple layer 3 protocols. 3654 The CMIs allocated for each such registration may or may 3655 not be the same. 3656 Generate a CMI for this protocol. 3658 Return. 3660 5. Modified JOIN/LEAVE Processing 3662 This routine processes JOIN/LEAVE when a server map exists. 3664 Make a copy of the message. 3665 Change the type of the copy to MARS_SJOIN. 3666 If the message is a MARS_LEAVE then { 3667 Change the type of the copy to MARS_SLEAVE. 3668 } 3669 mar$flags.copy = 1 (copy). 3670 Indicate the message to be sent on ServerControlVC. 3671 Send message (copy) as indicated. 3672 mar$flags.punched = 0 in the original message. 3673 Indicate the message to be sent on Private VC. 3674 Send message (original) as indicated. 3675 Hole punch the group by excluding 3676 from the range those groups that are served by MCSs. 3677 Indicate the (original) message to be sent on ClusterControlVC. 3678 If (number of holes punched > 0) then { /* punched holes */ 3679 In original message do { 3680 mar$flags.punched = 1. 3681 old punched list <- new punched list. 3682 } 3683 } /* punched holes */ 3684 mar$flags.copy = 1. 3685 Send message as indicated. 3686 Return.