idnits 2.17.1 draft-mahalingam-dutt-dcops-vxlan-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: ---------------------------------------------------------------------------- == The page length should not exceed 58 lines per page, but there was 20 longer pages, the longest (page 14) being 68 lines == It seems as if not all pages are separated by form feeds - found 0 form feeds but 20 pages Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** There are 345 instances of too long lines in the document, the longest one being 11 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 25 has weird spacing: '...t-Draft is s...' == Line 35 has weird spacing: '... at any ...' == Line 38 has weird spacing: '... The list ...' == Line 111 has weird spacing: '.... This is n...' == Line 124 has weird spacing: '... the indiv...' == (35 more instances...) -- The document date (February 24, 2012) is 4444 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Missing Reference: 'ECMP' is mentioned on line 119, but not defined == Unused Reference: 'RFC4601' is defined on line 731, but no explicit reference was found in the text == Unused Reference: 'RFC5015' is defined on line 735, but no explicit reference was found in the text == Unused Reference: 'RFC4541' is defined on line 739, but no explicit reference was found in the text -- Obsolete informational reference (is this intentional?): RFC 4601 (Obsoleted by RFC 7761) Summary: 1 error (**), 0 flaws (~~), 13 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group M.Mahalingam 3 Internet Draft VMware 4 Intended Status: Experimental D.Dutt 5 Expires: August 2012 K.Duda 6 Arista 7 P.Agarwal 8 Broadcom 9 L. Kreeger 10 Cisco 11 T. Sridhar 12 VMware 13 M.Bursell 14 Citrix 15 C.Wright 16 Red Hat 17 February 24, 2012 19 VXLAN: A Framework for Overlaying Virtualized Layer 2 Networks over 20 Layer 3 Networks 21 draft-mahalingam-dutt-dcops-vxlan-01.txt 23 Status of this Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF), its areas, and its working groups. Note that 30 other groups may also distribute working documents as Internet- 31 Drafts. 33 Internet-Drafts are draft documents valid for a maximum of six 34 months and may be updated, replaced, or obsoleted by other documents 35 at any time. It is inappropriate to use Internet-Drafts as 36 reference material or to cite them other than as "work in progress." 38 The list of current Internet-Drafts can be accessed at 39 http://www.ietf.org/ietf/1id-abstracts.txt 41 The list of Internet-Draft Shadow Directories can be accessed at 42 http://www.ietf.org/shadow.html 44 This Internet-Draft will expire on August 27, 2012. 46 Copyright Notice 48 Copyright (c) 2012 IETF Trust and the persons identified as the 49 document authors. All rights reserved. 51 This document is subject to BCP 78 and the IETF Trust's Legal 52 Provisions Relating to IETF Documents 53 (http://trustee.ietf.org/license-info) in effect on the date of 54 publication of this document. Please review these documents 55 carefully, as they describe your rights and restrictions with 56 respect to this document. 58 Abstract 60 This document describes Virtual eXtensible Local Area Network 61 (VXLAN), which is used to address the need for overlay networks 62 within virtualized data centers accommodating multiple tenants. The 63 scheme and the related protocols can be used in cloud service 64 provider and enterprise data center networks. 66 Table of Contents 68 1. Introduction...................................................3 69 1.1. Acronyms & Definitions....................................3 70 2. Conventions used in this document..............................4 71 3. VXLAN Problem Statement........................................5 72 3.1. Limitations imposed by Spanning Tree & VLAN Ranges........5 73 3.2. Multitenant Environments..................................5 74 3.3. Inadequate Table Sizes at ToR Switch......................6 75 4. Virtual eXtensible Local Area Network (VXLAN)..................6 76 4.1. Unicast VM to VM communication............................7 77 4.2. Broadcast Communication and Mapping to Multicast..........8 78 4.3. Physical Infrastructure Requirements......................9 79 5. VXLAN Frame Format.............................................9 80 6. VXLAN Deployment Scenarios....................................12 81 6.1. Inner VLAN Tag Handling..................................16 82 7. Security Considerations.......................................16 83 8. IANA Considerations...........................................17 84 9. Conclusion....................................................18 85 10. References...................................................18 86 10.1. Normative References....................................18 87 10.2. Informative References..................................18 88 11. Acknowledgments..............................................18 90 1. Introduction 92 Server virtualization has placed increased demands on the physical 93 network infrastructure. At a minimum, there is a need for more MAC 94 address table entries throughout the switched Ethernet network due 95 to potential attachment of hundreds of thousands of Virtual Machines 96 (VMs), each with its own MAC address. 98 Second, the VMs may be grouped according to their Virtual LAN 99 (VLAN). In a data center one might need thousands of VLANs to 100 partition the traffic according to the specific group that the VM 101 may belong to. The current VLAN limit of 4094 is inadequate in such 102 situations. A related requirement for virtualized environments is 103 having the Layer 2 network scale across the entire data center or 104 even between data centers for efficient allocation of compute, 105 network and storage resources. Using traditional approaches like 106 Spanning Tree Protocol (STP) for a loop free topology can result in 107 a large number of disabled links in such environments. 109 Another type of demand that is being placed on data centers is the 110 need to host multiple tenants, each with their own isolated network 111 domain. This is not economical to realize with dedicated 112 infrastructure, so network administrators opt to implement this over 113 a shared network. A concomitant problem is that each tenant may 114 independently assign MAC addresses and VLAN IDs leading to potential 115 duplication of these on the physical network. 117 The last scenario is the case where the network operator prefers to 118 use IP for interconnection of the physical infrastructure (e.g. to 119 achieve multipath scalability through Equal Cost Multipath [ECMP]) 120 while still preserving the Layer 2 model for inter-VM communication. 122 The scenarios described above lead to a requirement for an overlay 123 network. This overlay would be used to carry the MAC traffic from 124 the individual VMs in an encapsulated format over a logical 125 "tunnel". 127 This document details a framework termed Virtual eXtensible Local 128 Area Network (VXLAN) which provides such an encapsulation scheme to 129 address the various requirements specified above. 131 1.1. Acronyms & Definitions 133 ACL - Access Control List 135 ECMP - Equal Cost Multipath 136 IGMP - Internet Group Management Protocol 138 PIM - Protocol Independent Multicast 140 SPB - Shortest Path Bridging 142 STP - Spanning Tree Protocol 144 ToR - Top of Rack 146 TRILL - Transparent Interconnection of Lots of Links 148 VXLAN - Virtual eXtensible Local Area Network 150 VXLAN Segment - VXLAN Layer 2 overlay network over which VMs 152 communicate 154 VXLAN Overlay Network - another term for VXLAN Segment 156 VXLAN Gateway - an entity which forwards traffic between VXLAN 158 and non-VXLAN environments 160 VTEP - VXLAN Tunnel End Point - an entity which originates 161 and/or terminates VXLAN tunnels 163 VLAN - Virtual Local Area Network 165 VM - Virtual Machine 167 VNI - VXLAN Network Identifier (or VXLAN Segment ID) 169 2. Conventions used in this document 171 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 172 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 173 document are to be interpreted as described in RFC-2119 [RFC2119]. 175 In this document, these words will appear with that interpretation 176 only when in ALL CAPS. Lower case uses of these words are not to be 177 interpreted as carrying RFC-2119 significance. 179 3. VXLAN Problem Statement 181 This section details the problems that VXLAN is intended to address. 182 The focus is on the networking infrastructure within the data center 183 and the issues related to them. 185 3.1. Limitations imposed by Spanning Tree & VLAN Ranges 187 Current Layer 2 networks use the Spanning Tree Protocol (STP) to 188 avoid loops in the network due to duplicate paths. STP will turn off 189 links to avoid the replication and looping of frames. Some data 190 center operators see this as a problem with Layer 2 networks in 191 general since with STP they are effectively paying for more ports 192 and links than they can really use. In addition, resiliency due to 193 multipathing is not available with the STP model. Newer initiatives 194 like TRILL/Shortest Path Bridging (SPB) have been proposed to help 195 with multipathing and thus surmount some of the problems with STP. 196 STP limitations may also be avoided by configuring servers within a 197 rack to be on the same Layer 3 network with switching happening at 198 Layer 3 both within the rack and between racks. However, this is 199 incompatible with a Layer 2 model for inter-VM communication. 201 Another characteristic of Layer 2 data center networks is their use 202 of Virtual LANs (VLANs) to provide broadcast isolation. A 12 bit 203 VLAN ID is used in the Ethernet data frames to divide the larger 204 Layer 2 network into multiple broadcast domains. This has served 205 well for several data centers which require fewer than 4094 VLANs. 206 With the growing adoption of virtualization, this upper limit is 207 seeing pressure. Moreover, due to STP, several data centers limit 208 the number of VLANs that could be used. In addition, requirements 209 for multitenant environments accelerate the need for larger VLAN 210 limits, as discussed in Section 3.3. 212 3.2. Multitenant Environments 214 Cloud computing involves on demand elastic provisioning of resources 215 for multitenant environments. The most common example of cloud 216 computing is the public cloud, where a cloud service provider offers 217 these elastic services to multiple customers/tenants over the same 218 physical infrastructure. 220 Isolation of network traffic by tenant could be done via Layer 2 or 221 Layer 3 networks. For Layer 2 networks, VLANs are often used to 222 segregate traffic - so a tenant could be identified by its own VLAN, 223 for example. Due to the large number of tenants that a cloud 224 provider might service, the 4094 VLAN limit is often inadequate. In 225 addition, there is often a need for multiple VLANs per tenant, which 226 exacerbates the issue. 228 Another use case is cross pod expansion. A pod typically consists of 229 one or more racks of servers with associated network and storage 230 connectivity. Tenants may start off on a pod and, due to expansion, 231 require servers/VMs on other pods, especially the case when tenants 232 on the other pods are not fully utilizing all their resources. This 233 use case requires a "stretched" Layer 2 environment connecting the 234 individual servers/VMs. 236 Layer 3 networks are not a complete solution for multi tenancy 237 either. Two tenants might use the same set of Layer 3 addresses 238 within their networks which requires the cloud provider to provide 239 isolation in some other form. Further, requiring all tenants to use 240 IP excludes customers relying on direct Layer 2 or non-IP Layer 3 241 protocols for inter VM communication. 243 3.3. Inadequate Table Sizes at ToR Switch 245 Today's virtualized environments place additional demands on the MAC 246 address tables of Top of Rack (ToR) switches which connect to the 247 servers. Instead of just one MAC address per server link, the ToR 248 now has to learn the MAC addresses of the individual VMs (which 249 could range in the 100s per server). This is a requirement since 250 traffic from/to the VMs to the rest of the physical network will 251 traverse the link to the switch. A typical ToR switch could connect 252 to 24 or 48 servers depending upon the number of its server facing 253 ports. A data center might consist of several racks, so each ToR 254 switch would need to maintain an address table for the communicating 255 VMs across the various physical servers. This places a much larger 256 demand on the table capacity compared to non-virtualized 257 environments. 259 If the table overflows, the switch may stop learning new addresses 260 until idle entries age out, leading to significant flooding of 261 subsequent unknown destination frames. 263 4. Virtual eXtensible Local Area Network (VXLAN) 265 VXLAN (Virtual eXtensible Local Area Network) addresses the above 266 requirements of the Layer 2 and Layer 3 data center network 267 infrastructure in the presence of VMs in a multitenant environment. 268 It runs over the existing networking infrastructure and provides a 269 means to "stretch" a Layer 2 network. In short, VXLAN is a Layer 2 270 overlay scheme over a Layer 3 network. Each overlay is termed a 271 VXLAN segment. Only VMs within the same VXLAN segment can 272 communicate with each other. Each VXLAN segment is scoped through a 273 24 bit segment ID, hereafter termed the VXLAN Network Identifier 274 (VNI). This allows up to 16M VXLAN segments to coexist within the 275 same administrative domain. 277 The VNI scopes the inner MAC frame originated by the individual VM. 278 Thus, you could have overlapping MAC addresses across segments but 279 never have traffic "cross over" since the traffic is isolated using 280 the VNI qualifier. This qualifier is in an outer header envelope 281 over the inner MAC frame originated by the VM. In the following 282 sections, the term "VXLAN segment" is used interchangeably with the 283 term "VXLAN overlay network". 285 Due to this encapsulation, VXLAN could also be termed a tunneling 286 scheme to overlay Layer 2 networks on top of Layer 3 networks. The 287 tunnels are stateless, so each frame is encapsulated according to a 288 set of rules. The end point of the tunnel (VTEP) discussed in the 289 following sections is located within the hypervisor on the server 290 which houses the VM. Thus, the VNI and VXLAN related tunnel/outer 291 header encapsulation are known only to the VTEP - the VM never sees 292 it (see Figure 1). Note that it is possible that VTEPs could also be 293 on a physical switch or physical server and could be implemented in 294 software or hardware. One use case where the VTEP is a physical 295 switch is discussed in Section 6 VXLAN on VXLAN deployment 296 scenarios. 298 The following sections discuss typical traffic flow scenarios in a 299 VXLAN environment using one type of control scheme - data plane 300 learning. Here, the association of VM's MAC to VTEP's IP is 301 discovered via source learning. Multicast is used for carrying 302 unknown destination, broadcast and multicast frames. 304 In addition to a learning based control plane, there are other 305 schemes possible for the distribution of the VTEP IP to VM MAC 306 mapping information. Options could include a central directory based 307 lookup by the individual VTEPs, distribution of this mapping 308 information to the VTEPs by the central directory, and so on. These 309 are sometimes characterized as push and pull models respectively. 310 This draft will focus on the data plane learning scheme as the 311 control plane for VXLAN. 313 4.1. Unicast VM to VM communication 315 Consider a VM within a VXLAN overlay network. This VM is unaware of 316 VXLAN. To communicate with a VM on a different host, it sends a MAC 317 frame destined to the target as before. The VTEP on the physical 318 host looks up the VNI to which this VM is associated. It then 319 determines if the destination MAC is on the same segment. If so, an 320 outer header comprising an outer MAC, outer IP address and VXLAN 321 header (see Figure 1 in Section 5 for frame format) are inserted in 322 front of the original MAC frame. The final packet is transmitted out 323 to the destination. This is the IP address of the remote VTEP 324 connecting the destination VM (represented by the inner MAC 325 destination address). 327 Upon reception, the remote VTEP verifies that the VNI is a valid one 328 and is used by the destination VM. If so, the packet is stripped of 329 its outer header and passed on to the destination VM. The 330 destination VM never knows about the VNI or that the frame was 331 transported with a VXLAN encapsulation. 333 In addition to forwarding the packet to the destination VM, the 334 remote VTEP learns the Inner Source MAC to outer Source IP address 335 mapping. It stores this mapping in a table so that when the 336 destination VM sends a response packet, there is no need for an 337 "unknown destination" flooding of the response packet. 339 Determining the MAC address of the destination VM prior to the 340 transmission by the source VM is performed as with non-VXLAN 341 environments except as described below. Broadcast frames are used 342 but are encapsulated within a multicast packet, as detailed in the 343 next section. 345 4.2. Broadcast Communication and Mapping to Multicast 347 Consider the VM on the source host attempting to communicate with 348 the destination VM using IP. Assuming that they are both on the 349 same subnet, the VM sends out an ARP broadcast frame. In the non- 350 VXLAN environment, this frame would be sent out using MAC broadcast 351 which all switches carrying that VLAN. 353 With VXLAN, a header including the VXLAN VNI is inserted at the 354 beginning of the packet along with the IP header and UDP header. 355 However, this broadcast packet is sent out to the IP multicast group 356 on which that VXLAN overlay network is realized. 358 To effect this, we need to have a mapping between the VXLAN VNI and 359 the IP multicast group that it will use. This mapping is done at the 360 management layer and provided to the individual VTEPs through a 361 management channel. Using this mapping, the VTEP can provide IGMP 362 membership reports to the upstream switch/router to join/leave the 363 VXLAN related IP multicast groups as needed. This will enable 364 pruning of the leaf nodes for specific multicast traffic addresses 365 based on whether a member is available on this host using the 366 specific multicast address. In addition, use of multicast routing 367 protocols like Protocol Independent Multicast - Sparse Mode (PIM-SM) 368 will provide efficient multicast trees within the Layer 3 network. 370 The VTEP will use (*,G) joins. This is needed as the set of VXLAN 371 tunnel sources is unknown and may change often, as the VMs come 372 up/go down across different hosts. A side note here is that since 373 each VTEP can act as both the source and destination for multicast 374 packets, a protocol like PIM-bidir would be more efficient. 376 The destination VM sends a standard ARP response using IP unicast. 377 This frame will be encapsulated back to the VTEP connecting the 378 originating VM using IP unicast VXLAN encapsulation. This is 379 possible since the mapping of the ARP response's destination MAC to 380 the VXLAN tunnel end point IP was learned earlier through the ARP 381 request. 383 Another point to note is that multicast frames and "unknown MAC 384 destination" frames are also sent using the multicast tree, similar 385 to the broadcast frames. 387 4.3. Physical Infrastructure Requirements 389 When IP multicast is used within the network infrastructure, a 390 multicast routing protocol like PIM-SM can be used by the individual 391 Layer 3 IP routers/switches within the network. This is used to 392 build efficient multicast forwarding trees so that multicast frames 393 are only sent to those hosts which have requested to receive them. 395 Similarly, there is no requirement that the actual network 396 connecting the source VM and destination VM should be a Layer 3 397 network - VXLAN can also work over Layer 2 networks. In either case, 398 efficient multicast replication within the Layer 2 network can be 399 achieved using IGMP snooping. 401 5. VXLAN Frame Format 403 The VXLAN frame format is shown below. Parsing this from the bottom, 404 there is an inner MAC frame with its own Ethernet header with 405 source, destination MAC addresses along with the Ethernet type plus 406 an optional VLAN. One use case of the inner VLAN tag is with VM 407 based VLAN tagging in a virtualized environment. See Section 6 for 408 further details of inner VLAN tag handling. 410 The inner MAC frame is encapsulated with the following four headers 411 (starting from the innermost header): 413 O VXLAN Header: This is an 8 byte field which has: 415 o Flags (8 bits) where the I flag MUST be set to 1 for a valid 416 VXLAN Network ID (VNI). The remaining 7 bits (designated "R") are 417 reserved fields and MUST be set to zero. 419 o VXLAN Segment ID/VXLAN Network Identifier (VNI) - this is a 24 420 bit value used to designate the individual VXLAN overlay network 421 on which the communicating VMs are situated. VMs in different 422 VXLAN overlay networks cannot communicate with each other. 424 o Reserved fields (24 bits and 8 bits) - MUST be set to zero. 426 O Outer UDP Header: This is the outer UDP header with a source 427 port provided by the VTEP and the destination port being a well 428 known UDP port to be obtained by IANA assignment. It is recommended 429 that the source port be a hash of the inner Ethernet frame's 430 headers. This is to enable a level of entropy for ECMP/load 431 balancing of the VM to VM traffic across the VXLAN overlay. 433 The UDP checksum field SHOULD be transmitted as zero. When a packet 434 is received with a UDP checksum of zero, it MUST be accepted for 435 decapsulation. Optionally, if the encapsulating endpoint includes 436 a non-zero UDP checksum, it MUST be correctly calculated across the 437 entire packet including the IP header, UDP header, VXLAN header and 438 encapsulated MAC frame. When a decapsulating endpoint receives a 439 packet with a non-zero checksum it MAY choose to verify the 440 checksum value. If it chooses to perform such verification, and the 441 verification fails, the packet MUST be dropped. If the 442 decapsulating destination chooses not to perform the verification, 443 or performs it successfully, the packet MUST be accepted for 444 decapsulation. 446 O Outer IP Header: This is the outer IP header with the source IP 447 address indicating the IP address of the VTEP over which the 448 communicating VM (as represented by the inner source MAC address) is 449 running. The destination IP address is the IP address of the VTEP 450 connecting the communicating VM as represented by the inner 451 destination MAC address. 453 O Outer Ethernet Header (example): Figure 1 is an example of an 454 inner Ethernet frame encapsulated within an outer Ethernet + IP + 455 UDP + VXLAN header. The outer destination MAC address in this frame 456 may be the address of the target VTEP or of an intermediate Layer 3 457 router. The outer VLAN tag is optional. If present, it may be used 458 for delineating VXLAN traffic on the LAN. 460 0 1 2 3 461 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 463 Outer Ethernet Header: | 464 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 465 | Outer Destination MAC Address | 466 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 467 | Outer Destination MAC Address | Outer Source MAC Address | 468 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 469 | Outer Source MAC Address | 470 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 471 Optional Ethertype = C-Tag 802.1Q | Outer.VLAN Tag Information | 472 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 473 | Ethertype 0x0800 | 474 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 475 Outer IP Header: 476 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 477 |Version| IHL |Type of Service| Total Length | 478 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 479 | Identification |Flags| Fragment Offset | 480 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 481 | Time to Live | Protocol | Header Checksum | 482 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 483 | Outer Source Address | 484 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 485 | Outer Destination Address | 486 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 487 Outer UDP Header: 488 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 489 | Source Port = xxxx | Dest Port = VXLAN Port | 490 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 491 | UDP Length | UDP Checksum | 492 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 493 0 1 2 3 494 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 496 VXLAN Header: 497 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 498 |R|R|R|R|I|R|R|R| Reserved | 499 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 500 | VXLAN Network Identifier (VNI) | Reserved | 501 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 0 503 Inner Ethernet Header: | 504 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 505 | Inner Destination MAC Address | 506 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 507 | Inner Destination MAC Address | Inner Source MAC Address | 508 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 509 | Inner Source MAC Address | 510 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 511 Optional Ethertype = C-Tag [802.1Q] | Inner.VLAN Tag Information | 512 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 513 Payload: 514 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 515 | Ethertype of Original Payload | | 516 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | 517 | Original Ethernet Payload | 518 | | 519 | (Note that the original Ethernet Frame's FCS is not included) | 520 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 521 Frame Check Sequence: 522 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 523 | New FCS (Frame Check Sequence) for Outer Ethernet Frame | 524 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 526 Figure 1 VXLAN Frame Format 528 The frame format above shows tunneling of Ethernet frames using IPv4 529 for transport. Use of VXLAN with IPv6 transport will be addressed 530 in a future version of this draft. 532 6. VXLAN Deployment Scenarios 534 VXLAN is typically deployed in data centers on virtualized hosts, 535 which may be spread across multiple racks. The individual racks may 536 be parts of a different Layer 3 network or they could be in a single 537 Layer 2 network. The VXLAN segments/overlay networks are overlaid on 538 top of these Layer 2 or Layer 3 networks. 540 Consider Figure 2 below depicting two virtualized servers attached 541 to a Layer 3 infrastructure. The servers could be on the same rack, 542 or on different racks or potentially across data centers within the 543 same administrative domain. There are 4 VXLAN overlay networks 544 identified by the VNIs 22, 34, 74 and 98. Consider the case of VM1-1 545 in Server 1 and VM2-4 on Server 2 which are on the same VXLAN 546 overlay network identified by VNI 22. The VMs do not know about the 547 overlay networks and transport method since the encapsulation and 548 decapsulation happen transparently at the VTEPs on Servers 1 and 2. 549 The other overlay networks and the corresponding VMs are: VM1-2 on 550 Server 1 and VM2-1 on Server 2 both on VNI 34, VM1-3 on Server 1 and 551 VM2-2 on Server 2 on VNI 74, and finally VM1-4 on Server 1 and VM2-3 552 on Server 2 on VNI 98. 554 +------------+-------------+ 555 | Server 1 | 556 | +----+----+ +----+----+ | 557 | |VM1-1 | |VM1-2 | | 558 | |VNI 22 | |VNI 34 | | 559 | | | | | | 560 | +---------+ +---------+ | 561 | | 562 | +----+----+ +----+----+ | 563 | |VM1-3 | |VM1-4 | | 564 | |VNI 74 | |VNI 98 | | 565 | | | | | | 566 | +---------+ +---------+ | 567 | Hypervisor VTEP (IP1) | 568 +--------------------------+ 569 | 570 | 571 | 572 | 573 | 574 | 575 | +-------------+ 576 | | Layer 3 | 577 +---| Network | 578 | | 579 +-------------+ 580 | 581 | 582 + --------+ 583 | 584 +------------+-------------+ 585 | Server 2 | 586 | +----+----+ +----+----+ | 587 | |VM2-1 | |VM2-2 | | 588 | |VNI 34 | |VNI 74 | | 589 | | | | | | 590 | +---------+ +---------+ | 591 | | 592 | +----+----+ +----+----+ | 593 | |VM2-3 | |VM2-4 | | 594 | |VNI 98 | |VNI 22 | | 595 | | | | | | 596 | +---------+ +---------+ | 597 | Hypervisor VTEP (IP2) | 598 +--------------------------+ 600 Figure 2 VXLAN Deployment - VTEPs across a Layer 3 Network 602 One deployment scenario is where the tunnel termination point is a 603 physical server which understands VXLAN. Another scenario is where 604 nodes on a VXLAN overlay network need to communicate with nodes on 605 legacy networks which could be VLAN based. These nodes may be 606 physical nodes or virtual machines. To enable this communication, a 607 network can include VXLAN gateways (see Figure 3 below with a switch 608 acting as a VXLAN gateway) which forward traffic between VXLAN and 609 non-VXLAN environments. 611 Consider Figure 3 for the following discussion. For incoming frames 612 on the VXLAN connected interface, the gateway strips out the VXLAN 613 header and forwards to a physical port based on the destination MAC 614 address of the inner Ethernet frame. Decapsulated frames with the 615 inner VLAN ID SHOULD be discarded unless configured explicitly to be 616 passed on to the non-VXLAN interface. In the reverse direction, 617 incoming frames for the non-VXLAN interfaces are mapped to a 618 specific VXLAN overlay network based on the VLAN ID in the frame. 619 Unless configured explicitly to be passed on in the encapsulated 620 VXLAN frame, this VLAN ID is removed before the frame is 621 encapsulated for VXLAN. 623 These gateways which provide VXLAN tunnel termination functions 624 could be ToR/access switches or switches higher up in the data 625 center network topology - e.g. core or even WAN edge devices. The 626 last case (WAN edge) could involve a Provider Edge (PE) router which 627 terminates VXLAN tunnels in a hybrid cloud environment. Note that in 628 all these instances, the gateway functionality could be implemented 629 in software or hardware. 631 +---+-----+---+ +---+-----+---+ 632 | Server 1 | | Non VXLAN | 633 (VXLAN enabled)<-----+ +---->| server | 634 +-------------+ | | +-------------+ 635 | | 636 +---+-----+---+ | | +---+-----+---+ 637 |Server 2 | | | | Non VXLAN | 638 (VXLAN enabled)<-----+ +---+-----+---+ +---->| server | 639 +-------------+ | |Switch acting| | +-------------+ 640 |---| as VXLAN |-----| 641 +---+-----+---+ | | Gateway | 642 | Server 3 | | +-------------+ 643 (VXLAN enabled)<-----+ 644 +-------------+ | 645 | 646 +---+-----+---+ | 647 | Server 4 | | 648 (VXLAN enabled)<-----+ 649 +-------------+ 650 Figure 3 VXLAN Deployment - VXLAN Gateway 652 6.1. Inner VLAN Tag Handling 654 Inner VLAN Tag Handling in VTEP and VXLAN Gateway should conform to 655 the following: 657 Decapsulated VXLAN frames with the inner VLAN tag SHOULD be 658 discarded unless configured otherwise. On the encapsulation side, a 659 VTEP SHOULD NOT include an inner VLAN tag on tunnel packets unless 660 configured otherwise. When a VLAN-tagged packet is a candidate for 661 VXLAN tunneling, the encapsulating VTEP SHOULD strip the VLAN tag 662 unless configured otherwise. 664 7. Security Considerations 666 Traditionally, layer 2 networks can only be attacked from 'within' 667 by rogue endpoints - either by having inappropriate access to a LAN 668 and snooping on traffic or by injecting spoofed packets to 'take 669 over' another MAC address or by flooding and causing denial of 670 service. A MAC-over-IP mechanism for delivering Layer 2 traffic 671 significantly extends this attack surface. This can happen by rogues 672 injecting themselves into the network by subscribing to one or 673 more multicast groups that carry broadcast traffic for VXLAN 674 segments and also by sourcing MAC-over-UDP frames into the transport 675 network to inject spurious traffic, possibly to hijack MAC 676 addresses. 678 This proposal does not, at this time, incorporate specific measures 679 against such attacks, relying instead on other traditional 680 mechanisms layered on top of IP. This section, instead, sketches 681 out some possible approaches to security in the VXLAN environment. 683 Traditional Layer 2 attacks by rogue end points can be mitigated by 684 limiting the management and administrative scope of who deploys and 685 manages VMs/gateways in a VXLAN environment. In addition, such 686 administrative measures may be augmented by schemes like 802.1X for 687 admission control of individual end points. Also, the use of the 688 UDP based encapsulation of VXLAN enables exploiting the 5 tuple 689 based ACLs (Access Control Lists) functionality in physical 690 switches. 692 Tunneled traffic over the IP network can be secured with traditional 693 security mechanisms like IPsec that authenticate and optionally 694 encrypt VXLAN traffic. This will, of course, need to be coupled with 695 an authentication infrastructure for authorized endpoints to obtain 696 and distribute credentials. 698 VXLAN overlay networks are designated and operated over the existing 699 LAN infrastructure. To ensure that VXLAN end points and their VTEPs 700 are authorized on the LAN, it is recommended that a VLAN be 701 designated for VXLAN traffic and the servers/VTEPs send VXLAN 702 traffic over this VLAN to provide a measure of security. 704 In addition, VXLAN requires proper mapping of VNIs and VM membership 705 in these overlay networks. It is expected that this mapping be done 706 and communicated to the management entity on the VTEP and the 707 gateways using existing secure methods. 709 8. IANA Considerations 711 An IANA port will be requested for the VXLAN destination UDP port. 713 9. Conclusion 715 This document has introduced VXLAN, an overlay framework for 716 transporting MAC frames generated by VMs in isolated Layer 2 717 networks over an IP network. Through this scheme, it is possible to 718 stretch Layer 2 networks across Layer 3 networks. This finds use in 719 virtualized data center environments where Layer 2 networks may need 720 to span across the entire data center, or even between data centers. 722 10. References 724 10.1. Normative References 726 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 727 Requirement Levels", BCP 14, RFC 2119, March 1997. 729 10.2. Informative References 731 [RFC4601] Fenner, B., Handley, M., Holbrook, H., and Kouvelas, I., 732 "Protocol Independent Multicast - Sparse Mode (PIM-SM): Protocol 733 Specification", RFC 4601, August 2006. 735 [RFC5015] Handley, M., Kouvelas, I., Speakman, T., and Vicisano, L., 736 "Bidirectional Protocol Independent Multicast (BIDIR-PIM)", RFC 737 5015, October 2007. 739 [RFC4541] Christensen, M., Kimball, K., and Solensky, F., 740 "Considerations for Internet Group Management Protocol (IGMP) 741 and Multicast Listener Discovery (MLD) Snooping Switches", RFC 4541, 742 May 2006. 744 11. Acknowledgments 746 The authors wish to thank Ajit Sanzgiri for contributions to the 747 Security Considerations section and editorial inputs, Joseph Cheng, 748 Margaret Petrus and Milin Desai for their editorial reviews, inputs 749 and comments. 751 Authors' Addresses 753 Mallik Mahalingam 754 VMware Inc. 755 3401 Hillview 756 Palo Alto, CA 94304 758 Email: mallik@vmware.com 760 Dinesh G. Dutt 762 Email: ddutt.ietf@hobbesdutt.com 764 Kenneth Duda 765 Arista Networks 766 5470 Great America Parkway 767 Santa Clara, CA 95054 769 Email: kduda@aristanetworks.com 771 Puneet Agarwal 772 Broadcom Corporation 773 3151 Zanker Road 774 San Jose, CA 95134 776 Email: pagarwal@broadcom.com 778 Lawrence Kreeger 779 Cisco Systems, Inc. 780 170 W. Tasman Avenue 781 Palo Alto, CA 94304 783 Email: kreeger@cisco.com 785 T. Sridhar 786 VMware Inc. 787 3401 Hillview 788 Palo Alto, CA 94304 790 Email: tsridhar@vmware.com 791 Mike Bursell 792 Citrix Systems Research & Development Ltd. 793 Building 101 794 Cambridge Science Park 795 Milton Road 796 Cambridge CB4 0FY 797 United Kingdom 799 Email: mike.bursell@citrix.com 801 Chris Wright 802 Red Hat Inc. 803 1801 Varsity Drive 804 Raleigh, NC 27606 806 Email: chrisw@redhat.com