idnits 2.17.1 draft-ietf-nvo3-encap-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: ---------------------------------------------------------------------------- No issues found here. 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.) ** There are 46 instances of too long lines in the document, the longest one being 2 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 398 has weird spacing: '...of TLVs from ...' == Line 526 has weird spacing: '...lso, if the...' -- The document date (June 7, 2017) is 2513 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'RFC2119' is mentioned on line 170, but not defined == Missing Reference: 'I-D.ietf-nvo3-vxlan-gpe' is mentioned on line 741, but not defined == Missing Reference: 'I-D.herbert-gue-extensions' is mentioned on line 731, but not defined == Missing Reference: 'I-D.ietf-nvo3-gue' is mentioned on line 755, but not defined == Unused Reference: 'KEYWORDS' is defined on line 642, but no explicit reference was found in the text == Unused Reference: 'Geneve' is defined on line 647, but no explicit reference was found in the text == Unused Reference: 'GUE' is defined on line 649, but no explicit reference was found in the text == Unused Reference: 'NSH' is defined on line 650, but no explicit reference was found in the text == Unused Reference: 'VXLAN-GPE' is defined on line 651, but no explicit reference was found in the text == Outdated reference: A later version (-05) exists of draft-smith-vxlan-group-policy-03 Summary: 2 errors (**), 0 flaws (~~), 13 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 INTERNET-DRAFT Sami Boutros(Ed.) 3 Intended Status: Informational VMware 5 Ilango Ganga 6 Intel 8 Pankaj Garg 9 Microsoft 11 Rajeev Manur 12 Broadcom 14 Tal Mizrahi 15 Marvell 17 David Mozes 18 Mellanox 20 Erik Nordmark 22 Michael Smith 23 Cisco 25 Expires: December 9, 2017 June 7, 2017 27 NVO3 Encapsulation Considerations 28 draft-ietf-nvo3-encap-00 30 Abstract 32 As communicated by WG Chairs, the IETF NVO3 chairs and Routing Area 33 director have chartered a design team to take forward the 34 encapsulation discussion and see if there is potential to design a 35 common encapsulation that addresses the various technical concerns. 37 There are implications of different encapsulations in real 38 environments consisting of both software and hardware implementations 39 and spanning multiple data centers. For example, OAM functions such 40 as path MTU discovery become challenging with multiple encapsulations 41 along the data path. 43 The design team recommend Geneve with few modifications as the common 44 encapsulation, more details are described in section 7. 46 Status of this Memo 48 This Internet-Draft is submitted to IETF in full conformance with the 49 provisions of BCP 78 and BCP 79. 51 Internet-Drafts are working documents of the Internet Engineering 52 Task Force (IETF), its areas, and its working groups. Note that 53 other groups may also distribute working documents as 54 Internet-Drafts. 56 Internet-Drafts are draft documents valid for a maximum of six months 57 and may be updated, replaced, or obsoleted by other documents at any 58 time. It is inappropriate to use Internet-Drafts as reference 59 material or to cite them other than as "work in progress." 61 The list of current Internet-Drafts can be accessed at 62 http://www.ietf.org/1id-abstracts.html 64 The list of Internet-Draft Shadow Directories can be accessed at 65 http://www.ietf.org/shadow.html 67 Copyright and License Notice 69 Copyright (c) 2017 IETF Trust and the persons identified as the 70 document authors. All rights reserved. 72 This document is subject to BCP 78 and the IETF Trust's Legal 73 Provisions Relating to IETF Documents 74 (http://trustee.ietf.org/license-info) in effect on the date of 75 publication of this document. Please review these documents 76 carefully, as they describe your rights and restrictions with respect 77 to this document. Code Components extracted from this document must 78 include Simplified BSD License text as described in Section 4.e of 79 the Trust Legal Provisions and are provided without warranty as 80 described in the Simplified BSD License. 82 Table of Contents 84 1. Problem Statement . . . . . . . . . . . . . . . . . . . . . . . 4 85 2. Design Team Goals . . . . . . . . . . . . . . . . . . . . . . . 4 86 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 4 87 4. Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . 4 88 5. Issues with current Encapsulations . . . . . . . . . . . . . . 5 89 5.1 Geneve . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 90 5.2 GUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 91 5.3 VXLAN-GPE . . . . . . . . . . . . . . . . . . . . . . . . . 5 92 6. Common Encapsulation Considerations . . . . . . . . . . . . . . 6 93 6.1 Current Encapsulations . . . . . . . . . . . . . . . . . . . 6 94 6.2 Useful Extensions Use cases . . . . . . . . . . . . . . . . 6 95 6.2.1. Telemetry extensions. . . . . . . . . . . . . . . . . . 6 96 6.2.2. Security/Integrity extensions . . . . . . . . . . . . . 7 97 6.2.3. Group Base Policy . . . . . . . . . . . . . . . . . . . 7 98 6.3 Hardware Considerations . . . . . . . . . . . . . . . . . . 7 99 6.4 Extension Size . . . . . . . . . . . . . . . . . . . . . . . 8 100 6.5 Extension Ordering . . . . . . . . . . . . . . . . . . . . . 9 101 6.6 TLV vs Bit Fields . . . . . . . . . . . . . . . . . . . . . 9 102 6.7 Control Plane Considerations . . . . . . . . . . . . . . . . 10 103 6.8 Split NVE . . . . . . . . . . . . . . . . . . . . . . . . . 11 104 6.9 Larger VNI Considerations . . . . . . . . . . . . . . . . . 11 105 7. Design team recommendations . . . . . . . . . . . . . . . . . . 11 106 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14 107 9. Security Considerations . . . . . . . . . . . . . . . . . . . . 14 108 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 14 109 10.1 Normative References . . . . . . . . . . . . . . . . . . . 14 110 10.2 Informative References . . . . . . . . . . . . . . . . . . 15 111 11. Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . 15 112 11.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 15 113 11.2. Extensibility . . . . . . . . . . . . . . . . . . . . . . 15 114 11.2.1. Native Extensibility Support . . . . . . . . . . . . 15 115 11.2.2. Extension Parsing . . . . . . . . . . . . . . . . . . 15 116 11.2.3. Critical Extensions . . . . . . . . . . . . . . . . . 16 117 11.2.4. Maximal Header Length . . . . . . . . . . . . . . . . 16 118 11.3. Encapsulation Header . . . . . . . . . . . . . . . . . . 16 119 11.3.1. Virtual Network Identifier (VNI) . . . . . . . . . . 16 120 11.3.2. Next Protocol . . . . . . . . . . . . . . . . . . . . 16 121 11.3.3. Other Header Fields . . . . . . . . . . . . . . . . . 17 122 11.4. Comparison Summary . . . . . . . . . . . . . . . . . . . 17 123 Authors' Addresses (In alphabetical order) . . . . . . . . . . . . 18 125 1. Problem Statement 127 As communicated by WG Chairs, the NVO3 WG charter states that it may 128 produce requirements for network virtualization data planes based on 129 encapsulation of virtual network traffic over an IP-based underlay 130 data plane. Such requirements should consider OAM and security. Based 131 on these requirements the WG will select, extend, and/or develop one 132 or more data plane encapsulation format(s). 134 This has led to drafts describing three encapsulations being adopted 135 by the working group: 137 - draft-ietf-nvo3-geneve-03 139 - draft-ietf-nvo3-gue-04 141 - draft-ietf-nvo3-vxlan-gpe-02 143 Discussion on the list and in face-to-face meetings has identified a 144 number of technical problems with each of these encapsulations. 145 Furthermore, there was clear consensus at the IETF meeting in Berlin 146 that it is undesirable for the working group to progress more than 147 one data plane encapsulation. Although consensus could not be reached 148 on the list, the overall consensus was for a single encapsulation 149 (RFC2418, Section 3.3). Nonetheless there has been resistance to 150 converging on a single encapsulation format. 152 2. Design Team Goals 154 As communicated by WG Chairs, the design team should take one of the 155 proposed encapsulations and enhance it to address the technical 156 concerns. The simple evolution of deployed networks as well as 157 applicability to all locations in the NVO3 architecture are goals. 158 The DT should specifically avoid a design that is burdensome on 159 hardware implementations, but should allow future extensibility. The 160 chosen design should also operate well with ICMP and in ECMP 161 environments. If further extensibility is required, then it should be 162 done in such a manner that it does not require the consent of an 163 entity outside of the IETF. 165 3. Terminology 167 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 168 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 169 document are to be interpreted as described in RFC 2119 [RFC2119]. 171 4. Abbreviations 172 NVO3 Network Virtualization Overlays over Layer 3 174 OAM Operations, Administration, and Maintenance 176 TLV Type, Length, and Value 178 VNI Virtual Network Identifier 180 NVE Network Virtualization Edge 182 NVA Network Virtualization Authority 184 NIC Network interface card 186 Transit device Underlay network devices between NVE(s). 188 5. Issues with current Encapsulations 190 As summarized by WG Chairs. 192 5.1 Geneve 194 - Can't be implemented cost-effectively in all use cases because 195 variable length header and order of the TLVs makes is costly (in 196 terms of number of gates) to implement in hardware 198 - Header doesn't fit into largest commonly available parse buffer 199 (256 bytes in NIC). Cannot justify doubling buffer size unless it is 200 mandatory for hardware to process additional option fields. 202 5.2 GUE 204 - There were a significant number of objections related to the 205 complexity of implementation in hardware, similar to those noted for 206 Geneve above. 208 5.3 VXLAN-GPE 210 - GPE is not day-1 backwards compatible with VXLAN. Although the 211 frame format is similar, it uses a different UDP port, so would 212 require changes to existing implementations even if the rest of the 213 GPE frame is the same. 215 - GPE is insufficiently extensible. Numerous extensions and options 216 have been designed for GUE and Geneve. Note that these have not yet 217 been validated by the WG. 219 - Security e.g. of the VNI has not been addressed by GPE. Although a 220 shim header could be used for security and other extensions, this has 221 not been defined yet and its implications on offloading in NICs are 222 not understood. 224 6. Common Encapsulation Considerations 226 6.1 Current Encapsulations 228 Appendix A includes a detailed comparison between the three proposed 229 encapsulations. The comparison indicates several common properties, 230 but also three major differences among the encapsulations: 232 - Extensibility: Geneve and GUE were defined with built-in 233 extensibility, while VXLAN-GPE is not inherently extensible. Note 234 that any of the three encapsulations can be extended using the 235 Network Service Header (NSH). 237 - Extension method: Geneve is extensible using Type/Length/Value 238 (TLV) fields, while GUE uses a small set of possible extensions, and 239 a set of flags that indicate which extension is present. 241 - Length field: Geneve and GUE include a Length field, indicating the 242 length of the encapsulation header, while VXLAN-GPE does not include 243 such a field. 245 6.2 Useful Extensions Use cases 247 Non vendor specific TLV MUST follow the standardization process. The 248 following use cases for extensions shows that there is a strong 249 requirement to support variable length extensions with possible 250 different subtypes. 252 6.2.1. Telemetry extensions. 254 In several scenarios it is beneficial to make information about the 255 path a packet took through the network or through a network device as 256 well as associated telemetry information available to the operator. 258 This includes not only tasks like debugging, troubleshooting, as well 259 as network planning and network optimization but also policy or 260 service level agreement compliance checks. 262 Packet scheduling algorithms, especially for balancing traffic across 263 equal cost paths or links, often leverage information contained 264 within the packet, such as protocol number, IP-address or MAC- 265 address. Probe packets would thus either need to be sent from the 266 exact same endpoints with the exact same parameters, or probe packets 267 would need to be artificially constructed as "fake" packets and 268 inserted along the path. Both approaches are often not feasible from 269 an operational perspective, be it that access to the end-system is 270 not feasible, or that the diversity of parameters and associated 271 probe packets to be created is simply too large. An in-band telemetry 272 mechanism in extensions is an alternative in those cases. 274 6.2.2. Security/Integrity extensions 276 Since the currently proposed NVO3 encapsulations do not protect their 277 headers a single bit corruption in the VNI field could deliver a 278 packet to the wrong tenant. Extensions are needed to use any 279 sophisticated security. 281 The possibility of VNI spoofing with an NVO3 protocol is exacerbated 282 by the use of UDP. Systems typically have no restrictions on 283 applications being able to send to any UDP port so an unprivileged 284 application can trivially spoof for instance, VXLAN packets, 285 including using arbitrary VNIs. 287 One can envision HMAC-like support in some NVO3 extension to 288 authenticate the header and the outer IP addresses, thereby 289 preventing attackers from injecting packets with spoofed VNIs. 291 An other aspect of security is payload security. Essentially this is 292 to make packets that look like IP|UDP|NVO3 Encap|DTLS/IPSEC-ESP 293 Extension|payload. This is nice since we still have the UDP header 294 for ECMP, the NVO3 header is in plain text so it can by read by 295 network elements, and different security or other payload transforms 296 can be supported on a single UDP port (we don't need a separate UDP 297 for DTLS/IPSEC). 299 6.2.3. Group Base Policy 301 Another use case would be to carry the Group Based Policy (GBP) 302 source group information within a NVO3 header extension in a similar 303 manner as has been implemented for VXLAN [VXLAN-GBP]. This allows 304 various forms of policy such as access control and QoS to be applied 305 between abstract groups rather than coupled to specific endpoint 306 addresses. 308 6.3 Hardware Considerations 310 Hardware restrictions should be taken into consideration along with 311 future hardware enhancements that may provide more flexible metadata 312 processing. However, the set of options that need to and will be 313 implemented in hardware will be a subset of what is implemented in 314 software, since software NVEs are likely to grow features, and hence 315 option support, at a more rapid rate. 317 We note that it is hard to predict which options will be implemented 318 in which piece of hardware and when. That depends on whether the 319 hardware will be in the form of a NIC providing increasing offload 320 capabilities to software NVEs, or a switch chip being used as an NVE 321 gateway towards non-NVO3 parts of the network, or even an transit 322 devices that participates in the NVO3 dataplane e.g. for OAM 323 purposes. 325 A result of this is that it doesn't look useful to prescribe some 326 order of the option so that the ones that are likely to be 327 implemented in hardware come first; we can't decide such an order 328 when we define the options, however a control plane can enforce such 329 order for some hardware implementations. 331 We do know that hardware needs to initially be able to efficiently 332 skip over the NVO3 header to find the inner payload. That is needed 333 for both NICs doing e.g. TCP offload and transit devices and NVEs 334 applying policy/ACLs to the inner payload. 336 6.4 Extension Size 338 Extension header length has a significant impact to hardware and 339 software implementations. A total header length that is too small 340 will unnecessarily constrained software flexibility. A total header 341 length that is too large will place a nontrivial cost on hardware 342 implementations. Thus, the design team recommends that there be a 343 minimum and maximum total extension header length selected. The 344 maximum total header length is determined by the bits allocated for 345 the total extension header length field. The risk with this approach 346 is that it may be difficult to extend the total header size in the 347 future. The minimum total header length is determined by a 348 requirement in the specifications that all implementations must meet. 349 The risk with this approach is that all implementations will only 350 implement the minimum total header length which would then become the 351 de facto maximum total header length. The recommended minimum total 352 header length is 64 bytes. 354 Single Extension size should always be 4 bytes aligned. 356 The maximum length of a single option should be large enough to meet 357 the different extension use case requirements e.g. in-band telemetry 358 and future use. 360 6.5 Extension Ordering 362 In order to support hardware nodes at the tunnel endpoint or at the 363 transit that can process one or few extensions TLVs in TCAM. A 364 control plane in such a deployment can signal a capability to ensure 365 a specific TLV will always appear in a specific order for example the 366 first one in the packet. 368 The order of the TLVs should be HW friendly for both the sender and 369 the receiver and possibly the transit node too. 371 Transit nodes doesn't participate in control plane communication 372 between the end points and are not required to process the options 373 however, if they do, they need to process only a small subset of 374 options that will be consumed by tunnel endpoints. 376 6.6 TLV vs Bit Fields 378 If there is a well-known initial set of options that are likely to be 379 implemented in software and in hardware, it can be efficient to use 380 the bit-field approach as in GUE. However, as described in section 381 6.3, if options are added over time and different subsets of options 382 are likely to be implemented in different pieces of hardware, then it 383 would be hard for the IETF to specify which options should get the 384 early bit fields. TLVs are a lot more flexible, which avoids the need 385 to determine the relative importance different options. However, 386 general TLV of arbitrary order, size, and repetition of the same 387 order is difficult to implement in hardware. A middle ground is to 388 use TLV with restrictions on the size and alignment, observing that 389 individual TLVs can have a fixed length, and support in the control 390 plane such that an NVE will only receive options that to needs and 391 implements. The control plane approach can potentially be used to 392 control the order of the TLVs sent to a particular NVE. Note that 393 transit devices are not likely to participate in the control plane 394 hence to the extent that they need to participate in option 395 processing they need more effort, But transit devices would have 396 issues with future GUE bits being defined for future options as well. 398 A benefit of TLVs from a HW perspective is that they are self 399 describing i.e., all the information is in the TLV. In a Bit fields 400 approach the hardware needs to look up the bit to determine the 401 length of the data associated with the bit through some separate 402 table, which would add hardware complexity. 404 There are use cases where multiple modules of software are running on 405 NVE. This can be modules such as a diagnostic module by one vendor 406 that does packet sampling and another module from a different vendor 407 that does a firewall. Using a TLV format, it is easier to have 408 different software modules process different TLVs, which could be 409 standard extensions or vendor specific extensions defined by the 410 different vendors, without conflicting with each other. This can help 411 with hardware modularity as well. There are some implementations with 412 options that allows different software like mac learning and security 413 handle different options. 415 6.7 Control Plane Considerations 417 Given that we want to allow large flexibility and extensibility for 418 e.g. software NVEs, yet be able to support key extensions in less 419 flexible e.g. hardware NVEs, it is useful to consider the control 420 plane. By control plane in this context we mean both protocols such 421 as EVPN and others, and also deployment specific configuration. 423 If each NVE can express in the control plane that they only care 424 about particular extensions (could be a single extension, or a few), 425 and the source NVEs only include requested extensions in the NVO3 426 packets, then the target NVE can both use a simpler parser (e.g., a 427 TCAM might be usable to look for a single NVO3 extension) and the 428 depth of the inner payload in the NVO3 packet will be minimized. 429 Furthermore, if the target NVE cares about a few extensions and can 430 express in the control plane the desired order of those extensions in 431 the NVO3 packets, then it can provide useful functionality with 432 minimal hardware requirements. 434 Note that transit devices that are not aware of the NVO3 extensions 435 somewhat benefit from such an approach, since the inner payload is 436 less deep in the packet if no extraneous extensions are included in 437 the packet. However, in general a transit device is not likely to 438 participate in the NVO3 control plane. (However, configuration 439 mechanisms can take into account limitations of the transit devices 440 used in particular deployments.) 442 Note that in this approach different NVEs could desire different 443 (sets of) extensions, which means that the source NVE needs to be 444 able to place different sets of extensions in different NVO3 packets, 445 and perhaps in different order. It also assumes that underlay 446 multicast or replication servers are not used together with NVO3 447 extensions. 449 There is a need to consider mandatory extensions versus optional 450 extensions. Mandatory extensions require the receiver to drop the 451 packet if the extension is unknown. A control plane mechanism can 452 prevent the need for dropping unknown extensions, since they would 453 not be included to targets that do not support them. 455 The control planes defined today need to add the ability to describe 456 the different encapsulations. Thus perhaps EVPN, and any other 457 control plane protocol that the IETF defines, should have a way to 458 enumerate the supported NVO3 extensions and their order. 460 The WG should consider developing a separate draft on guidance for 461 option processing and control plane participation. This should 462 provide examples/guidance on range of usage models and deployments 463 scenarios for specific options and ordering that are relevant for 464 that specific deployment. This includes end points and middle boxes 465 using the options. So, having the control plane negotiate the 466 constraints is most appropriate and flexible way to address these 467 requirements. 469 6.8 Split NVE 471 If the working group sees a need for having the hosts send and 472 receive options in a split NVE case, this is possible using any of 473 the existing extensible encapsulations (Geneve, GUE, GPE+NSH) by 474 defining a way to carry those over other transports. NSH can already 475 be used over different transports. 477 If we need to do this with other encapsulations it can be done by 478 defining an Ether type for other encapsulations so that it can be 479 carried over Ethernet and 802.1Q. 481 If we need to carry other encapsulations over MPLS, it would require 482 an EVPN control plane to signal that other encapsulation header + 483 options will be present in front of the L2 packet. The VNI can be 484 ignored in the header, and the MPLS label will be the one used to 485 identify the EVPN L2 instance. 487 6.9 Larger VNI Considerations 489 We discussed whether we should make VNI 32-bits or larger. The 490 benefit of 24-bit VNI would be to avoid unnecessary changes with 491 existing proposals and implementations that are almost all, if not 492 all, are using 24-bit VNI. If we need a larger VNI, an extension can 493 be used to support that. 495 7. Design team recommendations 497 We concluded that Geneve is most suitable as a starting point for 498 proposed standard for network virtualization, for the following 499 reasons: 501 1. We studied whether VNI should be in base header or in extensions 502 and whether it should be 24-bit or 32-bit. The design team agreed 503 that VNI is critical information for network virtualization and MUST 504 be present in all packets. Design team also agreed that 24-bit VNI 505 matches the existing widely used encapsulation format i.e. VxLAN and 506 NVGRE and hence more suitable to use going forward. 508 2. Geneve has the total options length that allow skipping over the 509 options for NIC offload operations, and will allow transit devices to 510 view flow information in the inner payload. 512 3. We considered the option of using NSH with VxLAN-GPE but given 513 that NSH is targeted at service chaining and contains service 514 chaining information, it is less suitable for the network 515 virtualization use case. The other downside for VxLAN-GPE was lack of 516 header length in VxLAN-GPE and hence makes skipping over the headers 517 to process inner payload more difficult. Total Option Length is 518 present in Geneve. It is not possible to skip any options in the 519 middle with VxLAN-GPE. In principle a split between a base header and 520 a header with options is interesting (whether that options header is 521 NSH or some new header without ties to a service path). We explored 522 whether it would make sense to either use NSH for this, or define a 523 new NVO3 options header. However, we observed that this makes it 524 slightly harder to find the inner payload since the length field is 525 not in the NVO3 header itself. Thus one more field would have to be 526 extracted to compute the start of the inner payload. Also, if the 527 experience with IPv6 extension headers is a guidance, there would be 528 a risk that key pieces of hardware might not implement the options 529 header, resulting in future calls to deprecate its use. Making the 530 options part of the base NVO3 header has less of those issues. Even 531 though the implementation of any particular option can not be 532 predicted ahead of time, the option mechanism and ability to skip the 533 options is likely to be broadly implemented. 535 4. We compared the TLV vs Bit-fields style extension and it was 536 deemed that parsing both TLV and bit-fields is expensive and while 537 bit-fields may be simpler to parse, it is also more restrictive and 538 requires guessing which extensions will be widely implemented so they 539 can get early bit assignments, given that half the bits are already 540 assigned in GUE, a widely deployed extension may appear in a flag 541 extension, and this will require extra processing, to dig the flag 542 from the flag extension and then look for the extension itself. As 543 well Bit-fields are not flexible enough to address the requirements 544 from OAM, Telemetry and security extensions, for variable length 545 option and different subtypes of the same option. While TLV are more 546 flexible, a control plane can restrict the number of option TLVs as 547 well the order and size of the TLVs to make it simpler for a 548 dataplane implementation to handle. 550 5. We briefly discussed multi-vendor NVE case, and the need to allow 551 vendors to put their own extensions in the NVE header. This is 552 possible with TLVs. 554 6. We also agreed that the C bit in Geneve is helpful to allow 555 receiver NVE to easily decide whether to process options or not. For 556 example a UUID based packet trace and how an optional extension such 557 as that can be ignored by receiver NVE and thus make it easy for NVE 558 to skip over the options. Thus the C-bit remains as defined in 559 Geneve. 561 7. There are already some extensions that are being discussed (see 562 section 6.2) of varying sizes, by using Geneve option it is possible 563 to get in band parameters like: switch id, ingress port, egress port, 564 internal delay, and queue in telemetry defined extension TLV from 565 switches. It is also possible to add Security extension TLVs like 566 HMAC and DTLS/IPSEC to authenticate the Geneve packet header and 567 secure the Geneve packet payload by software or hardware tunnel 568 endpoints. As well, a Group Based Policy extension TLV can be 569 carried. 571 8. There are implemented Geneve options today in production. There 572 are as well new HW supporting Geneve TLV parsing. In addition In-band 573 Telemetry (INT) specification being developed by P4.org illustrates 574 the option of INT meta data carried over Geneve. OVN/OVS have also 575 defined some option TLV(s) for Geneve. 577 9. The DT has addressed the usage models while considering the 578 requirements and implementations in general that includes software 579 and hardware. 581 There seems to be interest to standardize some well known secure 582 option TLVs to secure the header and payload to guarantee 583 encapsulation header integrity and tenant data privacy. The design 584 team recommends that the working group consider standardizing such 585 option(s). 587 We recommend the following enhancements to Geneve to make it more 588 suitable to hardware and yet provide the flexibility for software: 590 We would propose a text such as, while TLV are more flexible, a 591 control plane can restrict the number of option TLVs as well the 592 order and size of the TLVs to make it simpler for a data plane 593 implementation in software or hardware to handle. For example, there 594 may be some critical information such as secure hash that must be 595 processed in certain order at lowest latency. 597 A control plane can negotiate a subset of option TLVs and certain TLV 598 ordering, as well can limit the total number of option TLVs present 599 in the packet, for example, to allow hardware capable of processing 600 fewer options. Hence, the control planes need to have the ability to 601 describe the supported TLVs subset and their order. 603 The Geneve draft could specify that the subset and order of option 604 TLVs should be configurable for each remote NVE in the absence of a 605 protocol control plane. 607 We recommend Geneve to follow fragmentation recommendations in 608 overlay services like PWE3, and L2/L3 VPN recommendation to guarantee 609 larger MTU for the tunnel overhead 610 https://tools.ietf.org/html/rfc3985#section-5.3 612 We request Geneve to provide a recommendation for critical bit 613 processing - text could look like how critical bits can be used with 614 control plane specifying the critical options. 616 Given that there is a telemetry option use case for a length of 256 617 bytes, we recommend Geneve to increase the Single TLV option length 618 to 256. 620 We request Geneve to address Requirements for OAM considerations for 621 alternate marking and for performance measurements that need 2 bits 622 in the header. And clarify the need of the current OAM bit in the 623 Geneve Header. 625 We recommend the WG to work on security options for Geneve. 627 8. Acknowledgements 629 The authors would like to thank Tom Herbert for providing the 630 motivation for the Security/Integrity extension, and for his valuable 631 comments, and would like to thank T. Sridhar for his valuable 632 comments and feedback. 634 9. Security Considerations 636 This document does not introduce any additional security constraints. 638 10. References 640 10.1 Normative References 642 [KEYWORDS] Bradner, S., "Key words for use in RFCs to Indicate 643 Requirement Levels", BCP 14, RFC 2119, March 1997. 645 10.2 Informative References 647 [Geneve] Generic Network Virtualization Encapsulation [I-D.ietf-nvo3- 648 geneve] 649 [GUE] Generic UDP Encapsulation [I-D.ietf-nvo3-gue] 650 [NSH] Network Service Header [I-D.ietf-sfc-nsh] 651 [VXLAN-GPE] Virtual eXtensible Local Area Network - Generic Protocol 652 Extension [I-D.ietf-nvo3-vxlan-gpe] 654 [VXLAN-GBP] VXLAN Group Policy Option - [I-D.draft-smith-vxlan-group- 655 policy-03] 657 11. Appendix A 659 11.1. Overview 661 This section presents a comparison of the three NVO3 encapsulation 662 proposals, Geneve, GUE, and VXLAN-GPE. The three encapsulations use 663 an outer UDP/IP transport. Geneve and VXLAN-GPE use an 8-octet 664 header, while GUE uses a 4-octet header. In addition to the base 665 header, optional extensions may be included in the encapsulation, as 666 discussed in Section 3.2 below. 668 11.2. Extensibility 670 11.2.1. Native Extensibility Support 672 The Geneve and GUE encapsulations both enable optional headers to be 673 incorporated at the end of the base encapsulation header. 675 VXLAN-GPE does not provide native support for header extensions. 676 However, as discussed in [I-D.ietf-nvo3-vxlan-gpe], extensibility can 677 be attained to some extent if the Network Service Header (NSH) [I- 678 D.ietf-sfc-nsh] is used immediately following the VXLAN-GPE header. 679 NSH supports either a fixed-size extension (MD Type 1), or a 680 variable-size TLV-based extension (MD Type 2). It should be noted 681 that NSH-over-VXLAN-GPE implies an additional overhead of the 8- 682 octets NSH header, in addition to the VXLAN-GPE header. 684 11.2.2. Extension Parsing 686 The Geneve Variable Length Options are defined as 687 Type/Length/Value(TLV) extensions. Similarly, VXLAN-GPE, when using 688 NSH, can include NSH TLV-based extensions. In contrast, GUE defines 689 a small set of possible extension fields (proposed in [I-D.herbert- 690 gue-extensions]), and a set of flags in the GUE header that indicate 691 for each extension type whether it is present or not. 693 TLV-based extensions, as defined in Geneve, provide the flexibility 694 for a large number of possible extension types. Similar behavior can 695 be supported in NSH-over-VXLAN-GPE when using MD Type 2. The flag- 696 based approach taken in GUE strives to simplify implementations by 697 defining a small number of possible extensions, used in a fixed 698 order. 700 The Geneve and GUE headers both include a length field, defining the 701 total length of the encapsulation, including the optional extensions. 703 The length field simplifies the parsing of transit devices that skip 704 the encapsulation header without parsing its extensions. 706 11.2.3. Critical Extensions 708 The Geneve encapsulation header includes the 'C' field, which 709 indicates whether the current Geneve header includes critical 710 options, which must be parsed by the tunnel endpoint. If the endpoint 711 is not able to process the critical option, the packet is discarded. 713 11.2.4. Maximal Header Length 715 The maximal header length in Geneve, including options, is 260 716 octets. GUE defines the maximal header to be 128 octets. VXLAN-GPE 717 uses a fixed-length header of 8 octets, unless NSH-over-VXLAN-GPE is 718 used, yielding an encapsulation header of up to 264 octets. 720 11.3. Encapsulation Header 722 11.3.1. Virtual Network Identifier (VNI) 724 The Geneve and VXLAN-GPE headers both include a 24-bit VNI field. 725 GUE, on the other hand, enables the use of a 32-bit field called 726 VNID; this field is not included in the GUE header, but was defined 727 as an optional extension in [I-D.herbert-gue-extensions]. 729 The VXLAN-GPE header includes the 'I' bit, indicating that the VNI 730 field is valid in the current header. A similar indicator is defined 731 as a flag in the GUE header [I-D.herbert-gue-extensions]. 733 11.3.2. Next Protocol 735 The three encapsulation headers include a field that specifies the 736 type of the next protocol header, which resides after the NVO3 737 encapsulation header. The Geneve header includes a 16-bit field that 738 uses the IEEE Ethertype convention. GUE uses an 8-bit field, which 739 uses the IANA Internet protocol numbering. The VXLAN-GPE header 740 incorporates an 8-bit Next Protocol field, using a VXLAN-GPE-specific 741 registry, defined in [I-D.ietf-nvo3-vxlan-gpe]. 743 The VXLAN-GPE header also includes the 'P' bit, which explicitly 744 indicates whether the Next Protocol field is present in the current 745 header. 747 11.3.3. Other Header Fields 749 The OAM bit, which is defined in Geneve and in VXLAN-GPE, indicates 750 whether the current packet is an OAM packet. The GUE header includes 751 a similar field, but uses different terminology; the GUE 'C-bit' 752 specifies whether the current packet is a control packet. Note that 753 the GUE control bit can potentially be used in a large set of 754 protocols that are not OAM protocols. However, the control packet 755 examples discussed in [I-D.ietf-nvo3-gue] are OAM-related. 757 Each of the three NVO3 encapsulation headers includes a 2-bit Version 758 field, which is currently defined to be zero. 760 The Geneve and VXLAN-GPE headers include reserved fields; 14 bits in 761 the Geneve header, and 27 bits in the VXLAN-GPE header are reserved. 763 11.4. Comparison Summary 765 The following table summarizes the comparison between the three NVO3 766 encapsulations. 768 +----------------+----------------+----------------+----------------+ 769 | | Geneve | GUE | VXLAN-GPE | 770 +----------------+----------------+----------------+----------------+ 771 | Outer transport| UDP/IP | UDP/IP | UDP/IP | 772 +----------------+----------------+----------------+----------------+ 773 | Base header | 8 octets | 4 octets | 8 octets | 774 | length | | | (16 octets | 775 | | | | using NSH) | 776 +----------------+----------------+----------------+----------------+ 777 | Extensibility |Variable length |Extension fields| No native ext- | 778 | | options | | ensibility. | 779 | | | | Extensible | 780 | | | | using NSH. | 781 +----------------+----------------+----------------+----------------+ 782 | Extension | TLV-based | Flag-based | TLV-based | 783 | parsing method | | |(using NSH with | 784 | | | | MD Type 2) | 785 +----------------+----------------+----------------+----------------+ 786 | Extension | Variable | Fixed | Variable | 787 | order | | | (using NSH) | 788 +----------------+----------------+----------------+----------------+ 789 | Length field | + | + | - | 790 +----------------+----------------+----------------+----------------+ 791 | Max Header | 260 octets | 128 octets | 8 octets | 792 | Length | | |(264 using NSH) | 793 +----------------+----------------+----------------+----------------+ 794 | Critical exte- | + | - | - | 795 | nsion bit | | | | 796 +----------------+----------------+----------------+----------------+ 797 | VNI field size | 24 bits | 32 bits | 24 bits | 798 | | | (extension) | | 799 +----------------+----------------+----------------+----------------+ 800 | Next protocol | 16 bits | 8 bits | 8 bits | 801 | field | Ethertype | Internet prot- | New registry | 802 | | registry | ocol registry | | 803 +----------------+----------------+----------------+----------------+ 804 | Next protocol | - | - | + | 805 | indicator | | | | 806 +----------------+----------------+----------------+----------------+ 807 | OAM / control | OAM bit | Control bit | OAM bit | 808 | field | | | | 809 +----------------+----------------+----------------+----------------+ 810 | Version field | 2 bits | 2 bits | 2 bits | 811 +----------------+----------------+----------------+----------------+ 812 | Reserved bits | 14 bits | - | 27 bits | 813 +----------------+----------------+----------------+----------------+ 815 Figure 1: NVO3 Encapsulation Comparison 817 Authors' Addresses (In alphabetical order) 819 Sami Boutros 820 VMware 821 Email: sboutros@vmware.com 823 Ilango Ganga 824 Intel 825 Email: ilango.s.ganga@intel.com 827 Pankaj Garg 828 Microsoft 829 Email: pankajg@microsoft.com 831 Rajeev Manur 832 Broadcom 833 Email: rajeev.manur@broadcom.com 835 Tal Mizrahi 836 Marvell 837 Email: talmi@marvell.com 839 David Mozes 840 Mellanox 841 Email: davidm@mellanox.com 843 Erik Nordmark 844 Arista Networks 845 Email: nordmark@sonic.net 847 Michael Smith 848 Cisco 849 Email: michsmit@cisco.com 851 Sam Aldrin 852 Google 853 Email: aldrin.ietf@gmail.com 855 Ignas Bagdonas 856 Equinix 857 Email: ibagdona.ietf@gmail.com