idnits 2.17.1 draft-ietf-mboned-dc-deploy-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- == It seems as if not all pages are separated by form feeds - found 0 form feeds but 11 pages Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- == There are 3 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 Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (August 23, 2013) is 3898 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: '224-239' is mentioned on line 412, but not defined -- Obsolete informational reference (is this intentional?): RFC 4601 (Obsoleted by RFC 7761) Summary: 0 errors (**), 0 flaws (~~), 4 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 1 Internet Engineering Task Force M. McBride 2 Internet-Draft Huawei Technologies 3 Intended status: Informational August 23, 2013 4 Expires: February 24, 2014 6 Multicast in the Data Center Overview 7 draft-ietf-mboned-dc-deploy-01 9 Abstract 11 There has been much interest in issues surrounding massive amounts of 12 hosts in the data center. These issues include the prevalent use of 13 IP Multicast within the Data Center. Its important to understand how 14 IP Multicast is being deployed in the Data Center to be able to 15 understand the surrounding issues with doing so. This document 16 provides a quick survey of uses of multicast in the data center and 17 should serve as an aid to further discussion of issues related to 18 large amounts of multicast in the data center. 20 Status of this Memo 22 This Internet-Draft is submitted in full conformance with the 23 provisions of BCP 78 and BCP 79. 25 Internet-Drafts are working documents of the Internet Engineering 26 Task Force (IETF). Note that other groups may also distribute 27 working documents as Internet-Drafts. The list of current Internet- 28 Drafts is at http://datatracker.ietf.org/drafts/current/. 30 Internet-Drafts are draft documents valid for a maximum of six months 31 and may be updated, replaced, or obsoleted by other documents at any 32 time. It is inappropriate to use Internet-Drafts as reference 33 material or to cite them other than as "work in progress." 35 This Internet-Draft will expire on February 24, 2014. 37 Copyright Notice 39 Copyright (c) 2013 IETF Trust and the persons identified as the 40 document authors. All rights reserved. 42 This document is subject to BCP 78 and the IETF Trust's Legal 43 Provisions Relating to IETF Documents 44 (http://trustee.ietf.org/license-info) in effect on the date of 45 publication of this document. Please review these documents 46 carefully, as they describe your rights and restrictions with respect 47 to this document. Code Components extracted from this document must 48 include Simplified BSD License text as described in Section 4.e of 49 the Trust Legal Provisions and are provided without warranty as 50 described in the Simplified BSD License. 52 Table of Contents 54 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 55 2. Multicast Applications in the Data Center . . . . . . . . . . 3 56 2.1. Client-Server Applications . . . . . . . . . . . . . . . . 3 57 2.2. Non Client-Server Multicast Applications . . . . . . . . . 4 58 3. L2 Multicast Protocols in the Data Center . . . . . . . . . . 6 59 4. L3 Multicast Protocols in the Data Center . . . . . . . . . . 7 60 5. Challenges of using multicast in the Data Center . . . . . . . 7 61 6. Layer 3 / Layer 2 Topological Variations . . . . . . . . . . . 9 62 7. Address Resolution . . . . . . . . . . . . . . . . . . . . . . 9 63 7.1. Solicited-node Multicast Addresses for IPv6 address 64 resolution . . . . . . . . . . . . . . . . . . . . . . . . 9 65 7.2. Direct Mapping for Multicast address resolution . . . . . 9 66 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 10 67 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10 68 10. Security Considerations . . . . . . . . . . . . . . . . . . . 10 69 11. Informative References . . . . . . . . . . . . . . . . . . . . 10 70 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 11 72 1. Introduction 74 Data center servers often use IP Multicast to send data to clients or 75 other application servers. IP Multicast is expected to help conserve 76 bandwidth in the data center and reduce the load on servers. IP 77 Multicast is also a key component in several data center overlay 78 solutions. Increased reliance on multicast, in next generation data 79 centers, requires higher performance and capacity especially from the 80 switches. If multicast is to continue to be used in the data center, 81 it must scale well within and between datacenters. There has been 82 much interest in issues surrounding massive amounts of hosts in the 83 data center. There was a lengthy discussion, in the now closed ARMD 84 WG, involving the issues with address resolution for non ARP/ND 85 multicast traffic in data centers. This document provides a quick 86 survey of multicast in the data center and should serve as an aid to 87 further discussion of issues related to multicast in the data center. 89 ARP/ND issues are not addressed in this document except to explain 90 how address resolution occurs with multicast. 92 2. Multicast Applications in the Data Center 94 There are many data center operators who do not deploy Multicast in 95 their networks for scalability and stability reasons. There are also 96 many operators for whom multicast is a critical protocol within their 97 network and is enabled on their data center switches and routers. 98 For this latter group, there are several uses of multicast in their 99 data centers. An understanding of the uses of that multicast is 100 important in order to properly support these applications in the ever 101 evolving data centers. If, for instance, the majority of the 102 applications are discovering/signaling each other, using multicast, 103 there may be better ways to support them then using multicast. If, 104 however, the multicasting of data is occurring in large volumes, 105 there is a need for good data center overlay multicast support. The 106 applications either fall into the category of those that leverage L2 107 multicast for discovery or of those that require L3 support and 108 likely span multiple subnets. 110 2.1. Client-Server Applications 112 IPTV servers use multicast to deliver content from the data center to 113 end users. IPTV is typically a one to many application where the 114 hosts are configured for IGMPv3, the switches are configured with 115 IGMP snooping, and the routers are running PIM-SSM mode. Often 116 redundant servers are sending multicast streams into the network and 117 the network is forwarding the data across diverse paths. 119 Windows Media servers send multicast streaming to clients. Windows 120 Media Services streams to an IP multicast address and all clients 121 subscribe to the IP address to receive the same stream. This allows 122 a single stream to be played simultaneously by multiple clients and 123 thus reducing bandwidth utilization. 125 Market data relies extensively on IP multicast to deliver stock 126 quotes from the data center to a financial services provider and then 127 to the stock analysts. The most critical requirement of a multicast 128 trading floor is that it be highly available. The network must be 129 designed with no single point of failure and in a way the network can 130 respond in a deterministic manner to any failure. Typically 131 redundant servers (in a primary/backup or live live mode) are sending 132 multicast streams into the network and the network is forwarding the 133 data across diverse paths (when duplicate data is sent by multiple 134 servers). 136 With publish and subscribe servers, a separate message is sent to 137 each subscriber of a publication. With multicast publish/subscribe, 138 only one message is sent, regardless of the number of subscribers. 139 In a publish/subscribe system, client applications, some of which are 140 publishers and some of which are subscribers, are connected to a 141 network of message brokers that receive publications on a number of 142 topics, and send the publications on to the subscribers for those 143 topics. The more subscribers there are in the publish/subscribe 144 system, the greater the improvement to network utilization there 145 might be with multicast. 147 2.2. Non Client-Server Multicast Applications 149 Routers, running Virtual Routing Redundancy Protocol (VRRP), 150 communicate with one another using a multicast address. VRRP packets 151 are sent, encapsulated in IP packets, to 224.0.0.18. A failure to 152 receive a multicast packet from the master router for a period longer 153 than three times the advertisement timer causes the backup routers to 154 assume that the master router is dead. The virtual router then 155 transitions into an unsteady state and an election process is 156 initiated to select the next master router from the backup routers. 157 This is fulfilled through the use of multicast packets. Backup 158 router(s) are only to send multicast packets during an election 159 process. 161 Overlays may use IP multicast to virtualize L2 multicasts. IP 162 multicast is used to reduce the scope of the L2-over-UDP flooding to 163 only those hosts that have expressed explicit interest in the 164 frames.VXLAN, for instance, is an encapsulation scheme to carry L2 165 frames over L3 networks. The VXLAN Tunnel End Point (VTEP) 166 encapsulates frames inside an L3 tunnel. VXLANs are identified by a 167 24 bit VXLAN Network Identifier (VNI). The VTEP maintains a table of 168 known destination MAC addresses, and stores the IP address of the 169 tunnel to the remote VTEP to use for each. Unicast frames, between 170 VMs, are sent directly to the unicast L3 address of the remote VTEP. 171 Multicast frames are sent to a multicast IP group associated with the 172 VNI. Underlying IP Multicast protocols (PIM-SM/SSM/BIDIR) are used 173 to forward multicast data across the overlay. 175 The Ganglia application relies upon multicast for distributed 176 discovery and monitoring of computing systems such as clusters and 177 grids. It has been used to link clusters across university campuses 178 and can scale to handle clusters with 2000 nodes 180 Windows Server, cluster node exchange, relies upon the use of 181 multicast heartbeats between servers. Only the other interfaces in 182 the same multicast group use the data. Unlike broadcast, multicast 183 traffic does not need to be flooded throughout the network, reducing 184 the chance that unnecessary CPU cycles are expended filtering traffic 185 on nodes outside the cluster. As the number of nodes increases, the 186 ability to replace several unicast messages with a single multicast 187 message improves node performance and decreases network bandwidth 188 consumption. Multicast messages replace unicast messages in two 189 components of clustering: 191 o Heartbeats: The clustering failure detection engine is based on a 192 scheme whereby nodes send heartbeat messages to other nodes. 193 Specifically, for each network interface, a node sends a heartbeat 194 message to all other nodes with interfaces on that network. 195 Heartbeat messages are sent every 1.2 seconds. In the common case 196 where each node has an interface on each cluster network, there 197 are N * (N - 1) unicast heartbeats sent per network every 1.2 198 seconds in an N-node cluster. With multicast heartbeats, the 199 message count drops to N multicast heartbeats per network every 200 1.2 seconds, because each node sends 1 message instead of N - 1. 201 This represents a reduction in processing cycles on the sending 202 node and a reduction in network bandwidth consumed. 204 o Regroup: The clustering membership engine executes a regroup 205 protocol during a membership view change. The regroup protocol 206 algorithm assumes the ability to broadcast messages to all cluster 207 nodes. To avoid unnecessary network flooding and to properly 208 authenticate messages, the broadcast primitive is implemented by a 209 sequence of unicast messages. Converting the unicast messages to 210 a single multicast message conserves processing power on the 211 sending node and reduces network bandwidth consumption. 213 Multicast addresses in the 224.0.0.x range are considered link local 214 multicast addresses. They are used for protocol discovery and are 215 flooded to every port. For example, OSPF uses 224.0.0.5 and 216 224.0.0.6 for neighbor and DR discovery. These addresses are 217 reserved and will not be constrained by IGMP snooping. These 218 addresses are not to be used by any application. 220 3. L2 Multicast Protocols in the Data Center 222 The switches, in between the servers and the routers, rely upon igmp 223 snooping to bound the multicast to the ports leading to interested 224 hosts and to L3 routers. A switch will, by default, flood multicast 225 traffic to all the ports in a broadcast domain (VLAN). IGMP snooping 226 is designed to prevent hosts on a local network from receiving 227 traffic for a multicast group they have not explicitly joined. It 228 provides switches with a mechanism to prune multicast traffic from 229 links that do not contain a multicast listener (an IGMP client). 230 IGMP snooping is a L2 optimization for L3 IGMP. 232 IGMP snooping, with proxy reporting or report suppression, actively 233 filters IGMP packets in order to reduce load on the multicast router. 234 Joins and leaves heading upstream to the router are filtered so that 235 only the minimal quantity of information is sent. The switch is 236 trying to ensure the router only has a single entry for the group, 237 regardless of how many active listeners there are. If there are two 238 active listeners in a group and the first one leaves, then the switch 239 determines that the router does not need this information since it 240 does not affect the status of the group from the router's point of 241 view. However the next time there is a routine query from the router 242 the switch will forward the reply from the remaining host, to prevent 243 the router from believing there are no active listeners. It follows 244 that in active IGMP snooping, the router will generally only know 245 about the most recently joined member of the group. 247 In order for IGMP, and thus IGMP snooping, to function, a multicast 248 router must exist on the network and generate IGMP queries. The 249 tables (holding the member ports for each multicast group) created 250 for snooping are associated with the querier. Without a querier the 251 tables are not created and snooping will not work. Furthermore IGMP 252 general queries must be unconditionally forwarded by all switches 253 involved in IGMP snooping. Some IGMP snooping implementations 254 include full querier capability. Others are able to proxy and 255 retransmit queries from the multicast router. 257 In source-only networks, however, which presumably describes most 258 data center networks, there are no IGMP hosts on switch ports to 259 generate IGMP packets. Switch ports are connected to multicast 260 source ports and multicast router ports. The switch typically learns 261 about multicast groups from the multicast data stream by using a type 262 of source only learning (when only receiving multicast data on the 263 port, no IGMP packets). The switch forwards traffic only to the 264 multicast router ports. When the switch receives traffic for new IP 265 multicast groups, it will typically flood the packets to all ports in 266 the same VLAN. This unnecessary flooding can impact switch 267 performance. 269 4. L3 Multicast Protocols in the Data Center 271 There are three flavors of PIM used for Multicast Routing in the Data 272 Center: PIM-SM [RFC4601], PIM-SSM [RFC4607], and PIM-BIDIR [RFC5015]. 273 SSM provides the most efficient forwarding between sources and 274 receivers and is most suitable for one to many types of multicast 275 applications. State is built for each S,G channel therefore the more 276 sources and groups there are, the more state there is in the network. 277 BIDIR is the most efficient shared tree solution as one tree is built 278 for all S,G's, therefore saving state. But it is not the most 279 efficient in forwarding path between sources and receivers. SSM and 280 BIDIR are optimizations of PIM-SM. PIM-SM is still the most widely 281 deployed multicast routing protocol. PIM-SM can also be the most 282 complex. PIM-SM relies upon a RP (Rendezvous Point) to set up the 283 multicast tree and then will either switch to the SPT (shortest path 284 tree), similar to SSM, or stay on the shared tree (similar to BIDIR). 285 For massive amounts of hosts sending (and receiving) multicast, the 286 shared tree (particularly with PIM-BIDIR) provides the best potential 287 scaling since no matter how many multicast sources exist within a 288 VLAN, the tree number stays the same. IGMP snooping, IGMP proxy, and 289 PIM-BIDIR have the potential to scale to the huge scaling numbers 290 required in a data center. 292 5. Challenges of using multicast in the Data Center 294 Data Center environments may create unique challenges for IP 295 Multicast. Data Center networks required a high amount of VM traffic 296 and mobility within and between DC networks. DC networks have large 297 numbers of servers. DC networks are often used with cloud 298 orchestration software. DC networks often use IP Multicast in their 299 unique environments. This section looks at the challenges of using 300 multicast within the challenging data center environment. 302 When IGMP/MLD Snooping is not implemented, ethernet switches will 303 flood multicast frames out of all switch-ports, which turns the 304 traffic into something more like a broadcast. 306 VRRP uses multicast heartbeat to communicate between routers. The 307 communication between the host and the default gateway is unicast. 309 The multicast heartbeat can be very chatty when there are thousands 310 of VRRP pairs with sub-second heartbeat calls back and forth. 312 Link-local multicast should scale well within one IP subnet 313 particularly with a large layer3 domain extending down to the access 314 or aggregation switches. But if multicast traverses beyond one IP 315 subnet, which is necessary for an overlay like VXLAN, you could 316 potentially have scaling concerns. If using a VXLAN overlay, it is 317 necessary to map the L2 multicast in the overlay to L3 multicast in 318 the underlay or do head end replication in the overlay and receive 319 duplicate frames on the first link from the router to the core 320 switch. The solution could be to run potentially thousands of PIM 321 messages to generate/maintain the required multicast state in the IP 322 underlay. The behavior of the upper layer, with respect to 323 broadcast/multicast, affects the choice of head end (*,G) or (S,G) 324 replication in the underlay, which affects the opex and capex of the 325 entire solution. A VXLAN, with thousands of logical groups, maps to 326 head end replication in the hypervisor or to IGMP from the hypervisor 327 and then PIM between the TOR and CORE 'switches' and the gateway 328 router. 330 Requiring IP multicast (especially PIM BIDIR) from the network can 331 prove challenging for data center operators especially at the kind of 332 scale that the VXLAN/NVGRE proposals require. This is also true when 333 the L2 topological domain is large and extended all the way to the L3 334 core. In data centers with highly virtualized servers, even small L2 335 domains may spread across many server racks (i.e. multiple switches 336 and router ports). 338 It's not uncommon for there to be 10-20 VMs per server in a 339 virtualized environment. One vendor reported a customer requesting a 340 scale to 400VM's per server. For multicast to be a viable solution 341 in this environment, the network needs to be able to scale to these 342 numbers when these VMs are sending/receiving multicast. 344 A lot of switching/routing hardware has problems with IP Multicast, 345 particularly with regards to hardware support of PIM-BIDIR. 347 Sending L2 multicast over a campus or data center backbone, in any 348 sort of significant way, is a new challenge enabled for the first 349 time by overlays. There are interesting challenges when pushing 350 large amounts of multicast traffic through a network, and have thus 351 far been dealt with using purpose-built networks. While the overlay 352 proposals have been careful not to impose new protocol requirements, 353 they have not addressed the issues of performance and scalability, 354 nor the large-scale availability of these protocols. 356 There is an unnecessary multicast stream flooding problem in the link 357 layer switches between the multicast source and the PIM First Hop 358 Router (FHR). The IGMP-Snooping Switch will forward multicast 359 streams to router ports, and the PIM FHR must receive all multicast 360 streams even if there is no request from receiver. This often leads 361 to waste of switch cache and link bandwidth when the multicast 362 streams are not actually required. [I-D.pim-umf-problem-statement] 363 details the problem and defines design goals for a generic mechanism 364 to restrain the unnecessary multicast stream flooding. 366 6. Layer 3 / Layer 2 Topological Variations 368 As discussed in [I-D.armd-problem-statement], there are a variety of 369 topological data center variations including L3 to Access Switches, 370 L3 to Aggregation Switches, and L3 in the Core only. Further 371 analysis is needed in order to understand how these variations affect 372 IP Multicast scalability 374 7. Address Resolution 376 7.1. Solicited-node Multicast Addresses for IPv6 address resolution 378 Solicited-node Multicast Addresses are used with IPv6 Neighbor 379 Discovery to provide the same function as the Address Resolution 380 Protocol (ARP) in IPv4. ARP uses broadcasts, to send an ARP 381 Requests, which are received by all end hosts on the local link. 382 Only the host being queried responds. However, the other hosts still 383 have to process and discard the request. With IPv6, a host is 384 required to join a Solicited-Node multicast group for each of its 385 configured unicast or anycast addresses. Because a Solicited-node 386 Multicast Address is a function of the last 24-bits of an IPv6 387 unicast or anycast address, the number of hosts that are subscribed 388 to each Solicited-node Multicast Address would typically be one 389 (there could be more because the mapping function is not a 1:1 390 mapping). Compared to ARP in IPv4, a host should not need to be 391 interrupted as often to service Neighbor Solicitation requests. 393 7.2. Direct Mapping for Multicast address resolution 395 With IPv4 unicast address resolution, the translation of an IP 396 address to a MAC address is done dynamically by ARP. With multicast 397 address resolution, the mapping from a multicast IP address to a 398 multicast MAC address is derived from direct mapping. In IPv4, the 399 mapping is done by assigning the low-order 23 bits of the multicast 400 IP address to fill the low-order 23 bits of the multicast MAC 401 address. When a host joins an IP multicast group, it instructs the 402 data link layer to receive frames that match the MAC address that 403 corresponds to the IP address of the multicast group. The data link 404 layer filters the frames and passes frames with matching destination 405 addresses to the IP module. Since the mapping from multicast IP 406 address to a MAC address ignores 5 bits of the IP address, groups of 407 32 multicast IP addresses are mapped to the same MAC address. As a 408 result a multicast MAC address cannot be uniquely mapped to a 409 multicast IPv4 address. Planning is required within an organization 410 to select IPv4 groups that are far enough away from each other as to 411 not end up with the same L2 address used. Any multicast address in 412 the [224-239].0.0.x and [224-239].128.0.x ranges should not be 413 considered. When sending IPv6 multicast packets on an Ethernet link, 414 the corresponding destination MAC address is a direct mapping of the 415 last 32 bits of the 128 bit IPv6 multicast address into the 48 bit 416 MAC address. It is possible for more than one IPv6 Multicast address 417 to map to the same 48 bit MAC address. 419 8. Acknowledgements 421 The authors would like to thank the many individuals who contributed 422 opinions on the ARMD wg mailing list about this topic: Linda Dunbar, 423 Anoop Ghanwani, Peter Ashwoodsmith, David Allan, Aldrin Isaac, Igor 424 Gashinsky, Michael Smith, Patrick Frejborg, Joel Jaeggli and Thomas 425 Narten. 427 9. IANA Considerations 429 This memo includes no request to IANA. 431 10. Security Considerations 433 No new security considerations result from this document 435 11. Informative References 437 [I-D.armd-problem-statement] 438 Narten, T., Karir, M., and I. Foo, 439 "draft-ietf-armd-problem-statement", February 2012. 441 [I-D.pim-umf-problem-statement] 442 Zhou, D., Deng, H., Shi, Y., Liu, H., and I. Bhattacharya, 443 "draft-dizhou-pim-umf-problem-statement", October 2010. 445 [RFC4601] Fenner, B., Handley, M., Holbrook, H., and I. Kouvelas, 446 "Protocol Independent Multicast - Sparse Mode (PIM-SM): 448 Protocol Specification (Revised)", RFC 4601, August 2006. 450 [RFC4607] Holbrook, H. and B. Cain, "Source-Specific Multicast for 451 IP", RFC 4607, August 2006. 453 [RFC5015] Handley, M., Kouvelas, I., Speakman, T., and L. Vicisano, 454 "Bidirectional Protocol Independent Multicast (BIDIR- 455 PIM)", RFC 5015, October 2007. 457 Author's Address 459 Mike McBride 460 Huawei Technologies 461 2330 Central Expressway 462 Santa Clara, CA 95050 463 USA 465 Email: michael.mcbride@huawei.com