idnits 2.17.1 draft-ietf-mboned-dc-deploy-00.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 (February 18, 2013) is 4078 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: '224-239' is mentioned on line 410, 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 February 18, 2013 4 Expires: August 22, 2013 6 Multicast in the Data Center Overview 7 draft-ietf-mboned-dc-deploy-00 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 August 22, 2013. 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 critical and is enabled on their 97 data center switches and routers. For this latter group, there are 98 several uses of multicast in their data centers. An understanding of 99 the uses of that multicast is important in order to properly support 100 these applications in the ever evolving data centers. If, for 101 instance, the majority of the applications are discovering/signaling 102 each other, using multicast, there may be better ways to support them 103 then using multicast. If, however, the multicasting of data is 104 occurring in large volumes, there is a need for good data center 105 overlay multicast support. The applications either fall into the 106 category of those that leverage L2 multicast for discovery or of 107 those that require L3 support and likely span multiple subnets. 109 2.1. Client-Server Applications 111 IPTV servers use multicast to deliver content from the data center to 112 end users. IPTV is typically a one to many application where the 113 hosts are configured for IGMPv3, the switches are configured with 114 IGMP snooping, and the routers are running PIM-SSM mode. Often 115 redundant servers are sending multicast streams into the network and 116 the network is forwarding the data across diverse paths. 118 Windows Media servers send multicast streaming to clients. Windows 119 Media Services streams to an IP multicast address and all clients 120 subscribe to the IP address to receive the same stream. This allows 121 a single stream to be played simultaneously by multiple clients and 122 thus reducing bandwidth utilization. 124 Market data relies extensively on IP multicast to deliver stock 125 quotes from the data center to a financial services provider and then 126 to the stock analysts. The most critical requirement of a multicast 127 trading floor is that it be highly available. The network must be 128 designed with no single point of failure and in a way the network can 129 respond in a deterministic manner to any failure. Typically 130 redundant servers (in a primary/backup or live live mode) are sending 131 multicast streams into the network and the network is forwarding the 132 data across diverse paths (when duplicate data is sent by multiple 133 servers). 135 With publish and subscribe servers, a separate message is sent to 136 each subscriber of a publication. With multicast publish/subscribe, 137 only one message is sent, regardless of the number of subscribers. 138 In a publish/subscribe system, client applications, some of which are 139 publishers and some of which are subscribers, are connected to a 140 network of message brokers that receive publications on a number of 141 topics, and send the publications on to the subscribers for those 142 topics. The more subscribers there are in the publish/subscribe 143 system, the greater the improvement to network utilization there 144 might be with multicast. 146 2.2. Non Client-Server Multicast Applications 148 Routers, running Virtual Routing Redundancy Protocol (VRRP), 149 communicate with one another using a multicast address. VRRP packets 150 are sent, encapsulated in IP packets, to 224.0.0.18. A failure to 151 receive a multicast packet from the master router for a period longer 152 than three times the advertisement timer causes the backup routers to 153 assume that the master router is dead. The virtual router then 154 transitions into an unsteady state and an election process is 155 initiated to select the next master router from the backup routers. 156 This is fulfilled through the use of multicast packets. Backup 157 router(s) are only to send multicast packets during an election 158 process. 160 Overlays may use IP multicast to virtualize L2 multicasts. IP 161 multicast is used to reduce the scope of the L2-over-UDP flooding to 162 only those hosts that have expressed explicit interest in the 163 frames.VXLAN, for instance, is an encapsulation scheme to carry L2 164 frames over L3 networks. The VXLAN Tunnel End Point (VTEP) 165 encapsulates frames inside an L3 tunnel. VXLANs are identified by a 166 24 bit VXLAN Network Identifier (VNI). The VTEP maintains a table of 167 known destination MAC addresses, and stores the IP address of the 168 tunnel to the remote VTEP to use for each. Unicast frames, between 169 VMs, are sent directly to the unicast L3 address of the remote VTEP. 170 Multicast frames are sent to a multicast IP group associated with the 171 VNI. Underlying IP Multicast protocols (PIM-SM/SSM/BIDIR) are used 172 to forward multicast data across the overlay. 174 The Ganglia application relies upon multicast for distributed 175 discovery and monitoring of computing systems such as clusters and 176 grids. It has been used to link clusters across university campuses 177 and can scale to handle clusters with 2000 nodes 179 Windows Server, cluster node exchange, relies upon the use of 180 multicast heartbeats between servers. Only the other interfaces in 181 the same multicast group use the data. Unlike broadcast, multicast 182 traffic does not need to be flooded throughout the network, reducing 183 the chance that unnecessary CPU cycles are expended filtering traffic 184 on nodes outside the cluster. As the number of nodes increases, the 185 ability to replace several unicast messages with a single multicast 186 message improves node performance and decreases network bandwidth 187 consumption. Multicast messages replace unicast messages in two 188 components of clustering: 190 o Heartbeats: The clustering failure detection engine is based on a 191 scheme whereby nodes send heartbeat messages to other nodes. 192 Specifically, for each network interface, a node sends a heartbeat 193 message to all other nodes with interfaces on that network. 194 Heartbeat messages are sent every 1.2 seconds. In the common case 195 where each node has an interface on each cluster network, there 196 are N * (N - 1) unicast heartbeats sent per network every 1.2 197 seconds in an N-node cluster. With multicast heartbeats, the 198 message count drops to N multicast heartbeats per network every 199 1.2 seconds, because each node sends 1 message instead of N - 1. 200 This represents a reduction in processing cycles on the sending 201 node and a reduction in network bandwidth consumed. 203 o Regroup: The clustering membership engine executes a regroup 204 protocol during a membership view change. The regroup protocol 205 algorithm assumes the ability to broadcast messages to all cluster 206 nodes. To avoid unnecessary network flooding and to properly 207 authenticate messages, the broadcast primitive is implemented by a 208 sequence of unicast messages. Converting the unicast messages to 209 a single multicast message conserves processing power on the 210 sending node and reduces network bandwidth consumption. 212 Multicast addresses in the 224.0.0.x range are considered link local 213 multicast addresses. They are used for protocol discovery and are 214 flooded to every port. For example, OSPF uses 224.0.0.5 and 215 224.0.0.6 for neighbor and DR discovery. These addresses are 216 reserved and will not be constrained by IGMP snooping. These 217 addresses are not to be used by any application. 219 3. L2 Multicast Protocols in the Data Center 221 The switches, in between the servers and the routers, rely upon igmp 222 snooping to bound the multicast to the ports leading to interested 223 hosts and to L3 routers. A switch will, by default, flood multicast 224 traffic to all the ports in a broadcast domain (VLAN). IGMP snooping 225 is designed to prevent hosts on a local network from receiving 226 traffic for a multicast group they have not explicitly joined. It 227 provides switches with a mechanism to prune multicast traffic from 228 links that do not contain a multicast listener (an IGMP client). 229 IGMP snooping is a L2 optimization for L3 IGMP. 231 IGMP snooping, with proxy reporting or report suppression, actively 232 filters IGMP packets in order to reduce load on the multicast router. 233 Joins and leaves heading upstream to the router are filtered so that 234 only the minimal quantity of information is sent. The switch is 235 trying to ensure the router only has a single entry for the group, 236 regardless of how many active listeners there are. If there are two 237 active listeners in a group and the first one leaves, then the switch 238 determines that the router does not need this information since it 239 does not affect the status of the group from the router's point of 240 view. However the next time there is a routine query from the router 241 the switch will forward the reply from the remaining host, to prevent 242 the router from believing there are no active listeners. It follows 243 that in active IGMP snooping, the router will generally only know 244 about the most recently joined member of the group. 246 In order for IGMP, and thus IGMP snooping, to function, a multicast 247 router must exist on the network and generate IGMP queries. The 248 tables (holding the member ports for each multicast group) created 249 for snooping are associated with the querier. Without a querier the 250 tables are not created and snooping will not work. Furthermore IGMP 251 general queries must be unconditionally forwarded by all switches 252 involved in IGMP snooping. Some IGMP snooping implementations 253 include full querier capability. Others are able to proxy and 254 retransmit queries from the multicast router. 256 In source-only networks, however, which presumably describes most 257 data center networks, there are no IGMP hosts on switch ports to 258 generate IGMP packets. Switch ports are connected to multicast 259 source ports and multicast router ports. The switch typically learns 260 about multicast groups from the multicast data stream by using a type 261 of source only learning (when only receiving multicast data on the 262 port, no IGMP packets). The switch forwards traffic only to the 263 multicast router ports. When the switch receives traffic for new IP 264 multicast groups, it will typically flood the packets to all ports in 265 the same VLAN. This unnecessary flooding can impact switch 266 performance. 268 4. L3 Multicast Protocols in the Data Center 270 There are three flavors of PIM used for Multicast Routing in the Data 271 Center: PIM-SM [RFC4601], PIM-SSM [RFC4607], and PIM-BIDIR [RFC5015]. 272 SSM provides the most efficient forwarding between sources and 273 receivers and is most suitable for one to many types of multicast 274 applications. State is built for each S,G channel therefore the more 275 sources and groups there are, the more state there is in the network. 276 BIDIR is the most efficient shared tree solution as one tree is built 277 for all S,G's, therefore saving state. But it is not the most 278 efficient in forwarding path between sources and receivers. SSM and 279 BIDIR are optimizations of PIM-SM. PIM-SM is still the most widely 280 deployed multicast routing protocol. PIM-SM can also be the most 281 complex. PIM-SM relies upon a RP (Rendezvous Point) to set up the 282 multicast tree and then will either switch to the SPT (shortest path 283 tree), similar to SSM, or stay on the shared tree (similar to BIDIR). 284 For massive amounts of hosts sending (and receiving) multicast, the 285 shared tree (particularly with PIM-BIDIR) provides the best potential 286 scaling since no matter how many multicast sources exist within a 287 VLAN, the tree number stays the same. IGMP snooping, IGMP proxy, and 288 PIM-BIDIR have the potential to scale to the huge scaling numbers 289 required in a data center. 291 5. Challenges of using multicast in the Data Center 293 Data Center environments may create unique challenges for IP 294 Multicast. Data Center networks required a high amount of VM traffic 295 and mobility within and between DC networks. DC networks have large 296 numbers of servers. DC networks are often used with cloud 297 orchestration software. DC networks often use IP Multicast in their 298 unique environments. This section looks at the challenges of using 299 multicast within the challenging data center environment. 301 When IGMP/MLD Snooping is not implemented, ethernet switches will 302 flood multicast frames out of all switch-ports, which turns the 303 traffic into something more like a broadcast. 305 VRRP uses multicast heartbeat to communicate between routers. The 306 communication between the host and the default gateway is unicast. 307 The multicast heartbeat can be very chatty when there are thousands 308 of VRRP pairs with sub-second heartbeat calls back and forth. 310 Link-local multicast should scale well within one IP subnet 311 particularly with a large layer3 domain extending down to the access 312 or aggregation switches. But if multicast traverses beyond one IP 313 subnet, which is necessary for an overlay like VXLAN, you could 314 potentially have scaling concerns. If using a VXLAN overlay, it is 315 necessary to map the L2 multicast in the overlay to L3 multicast in 316 the underlay or do head end replication in the overlay and receive 317 duplicate frames on the first link from the router to the core 318 switch. The solution could be to run potentially thousands of PIM 319 messages to generate/maintain the required multicast state in the IP 320 underlay. The behavior of the upper layer, with respect to 321 broadcast/multicast, affects the choice of head end (*,G) or (S,G) 322 replication in the underlay, which affects the opex and capex of the 323 entire solution. A VXLAN, with thousands of logical groups, maps to 324 head end replication in the hypervisor or to IGMP from the hypervisor 325 and then PIM between the TOR and CORE 'switches' and the gateway 326 router. 328 Requiring IP multicast (especially PIM BIDIR) from the network can 329 prove challenging for data center operators especially at the kind of 330 scale that the VXLAN/NVGRE proposals require. This is also true when 331 the L2 topological domain is large and extended all the way to the L3 332 core. In data centers with highly virtualized servers, even small L2 333 domains may spread across many server racks (i.e. multiple switches 334 and router ports). 336 It's not uncommon for there to be 10-20 VMs per server in a 337 virtualized environment. One vendor reported a customer requesting a 338 scale to 400VM's per server. For multicast to be a viable solution 339 in this environment, the network needs to be able to scale to these 340 numbers when these VMs are sending/receiving multicast. 342 A lot of switching/routing hardware has problems with IP Multicast, 343 particularly with regards to hardware support of PIM-BIDIR. 345 Sending L2 multicast over a campus or data center backbone, in any 346 sort of significant way, is a new challenge enabled for the first 347 time by overlays. There are interesting challenges when pushing 348 large amounts of multicast traffic through a network, and have thus 349 far been dealt with using purpose-built networks. While the overlay 350 proposals have been careful not to impose new protocol requirements, 351 they have not addressed the issues of performance and scalability, 352 nor the large-scale availability of these protocols. 354 There is an unnecessary multicast stream flooding problem in the link 355 layer switches between the multicast source and the PIM First Hop 356 Router (FHR). The IGMP-Snooping Switch will forward multicast 357 streams to router ports, and the PIM FHR must receive all multicast 358 streams even if there is no request from receiver. This often leads 359 to waste of switch cache and link bandwidth when the multicast 360 streams are not actually required. [I-D.pim-umf-problem-statement] 361 details the problem and defines design goals for a generic mechanism 362 to restrain the unnecessary multicast stream flooding. 364 6. Layer 3 / Layer 2 Topological Variations 366 As discussed in [I-D.armd-problem-statement], there are a variety of 367 topological data center variations including L3 to Access Switches, 368 L3 to Aggregation Switches, and L3 in the Core only. Further 369 analysis is needed in order to understand how these variations affect 370 IP Multicast scalability 372 7. Address Resolution 374 7.1. Solicited-node Multicast Addresses for IPv6 address resolution 376 Solicited-node Multicast Addresses are used with IPv6 Neighbor 377 Discovery to provide the same function as the Address Resolution 378 Protocol (ARP) in IPv4. ARP uses broadcasts, to send an ARP 379 Requests, which are received by all end hosts on the local link. 380 Only the host being queried responds. However, the other hosts still 381 have to process and discard the request. With IPv6, a host is 382 required to join a Solicited-Node multicast group for each of its 383 configured unicast or anycast addresses. Because a Solicited-node 384 Multicast Address is a function of the last 24-bits of an IPv6 385 unicast or anycast address, the number of hosts that are subscribed 386 to each Solicited-node Multicast Address would typically be one 387 (there could be more because the mapping function is not a 1:1 388 mapping). Compared to ARP in IPv4, a host should not need to be 389 interrupted as often to service Neighbor Solicitation requests. 391 7.2. Direct Mapping for Multicast address resolution 393 With IPv4 unicast address resolution, the translation of an IP 394 address to a MAC address is done dynamically by ARP. With multicast 395 address resolution, the mapping from a multicast IP address to a 396 multicast MAC address is derived from direct mapping. In IPv4, the 397 mapping is done by assigning the low-order 23 bits of the multicast 398 IP address to fill the low-order 23 bits of the multicast MAC 399 address. When a host joins an IP multicast group, it instructs the 400 data link layer to receive frames that match the MAC address that 401 corresponds to the IP address of the multicast group. The data link 402 layer filters the frames and passes frames with matching destination 403 addresses to the IP module. Since the mapping from multicast IP 404 address to a MAC address ignores 5 bits of the IP address, groups of 405 32 multicast IP addresses are mapped to the same MAC address. As a 406 result a multicast MAC address cannot be uniquely mapped to a 407 multicast IPv4 address. Planning is required within an organization 408 to select IPv4 groups that are far enough away from each other as to 409 not end up with the same L2 address used. Any multicast address in 410 the [224-239].0.0.x and [224-239].128.0.x ranges should not be 411 considered. When sending IPv6 multicast packets on an Ethernet link, 412 the corresponding destination MAC address is a direct mapping of the 413 last 32 bits of the 128 bit IPv6 multicast address into the 48 bit 414 MAC address. It is possible for more than one IPv6 Multicast address 415 to map to the same 48 bit MAC address. 417 8. Acknowledgements 419 The authors would like to thank the many individuals who contributed 420 opinions on the ARMD wg mailing list about this topic: Linda Dunbar, 421 Anoop Ghanwani, Peter Ashwoodsmith, David Allan, Aldrin Isaac, Igor 422 Gashinsky, Michael Smith, Patrick Frejborg, Joel Jaeggli and Thomas 423 Narten. 425 9. IANA Considerations 427 This memo includes no request to IANA. 429 10. Security Considerations 431 No security considerations at this time. 433 11. Informative References 435 [I-D.armd-problem-statement] 436 Narten, T., Karir, M., and I. Foo, 437 "draft-ietf-armd-problem-statement", February 2012. 439 [I-D.pim-umf-problem-statement] 440 Zhou, D., Deng, H., Shi, Y., Liu, H., and I. Bhattacharya, 441 "draft-dizhou-pim-umf-problem-statement", October 2010. 443 [RFC4601] Fenner, B., Handley, M., Holbrook, H., and I. Kouvelas, 444 "Protocol Independent Multicast - Sparse Mode (PIM-SM): 445 Protocol Specification (Revised)", RFC 4601, August 2006. 447 [RFC4607] Holbrook, H. and B. Cain, "Source-Specific Multicast for 448 IP", RFC 4607, August 2006. 450 [RFC5015] Handley, M., Kouvelas, I., Speakman, T., and L. Vicisano, 451 "Bidirectional Protocol Independent Multicast (BIDIR- 452 PIM)", RFC 5015, October 2007. 454 Author's Address 456 Mike McBride 457 Huawei Technologies 458 2330 Central Expressway 459 Santa Clara, CA 95050 460 USA 462 Email: michael.mcbride@huawei.com