idnits 2.17.1 draft-ietf-nvo3-encap-03.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 378 has weird spacing: '...of TLVs from ...' == Line 506 has weird spacing: '...lso, if the...' -- The document date (July 1, 2019) is 1754 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'RFC2119' is mentioned on line 150, but not defined == Missing Reference: 'I-D.ietf-nvo3-vxlan-gpe' is mentioned on line 721, but not defined == Missing Reference: 'I-D.herbert-gue-extensions' is mentioned on line 711, but not defined == Missing Reference: 'I-D.ietf-nvo3-gue' is mentioned on line 735, but not defined == Unused Reference: 'KEYWORDS' is defined on line 622, but no explicit reference was found in the text == Unused Reference: 'Geneve' is defined on line 627, but no explicit reference was found in the text == Unused Reference: 'GUE' is defined on line 629, but no explicit reference was found in the text == Unused Reference: 'NSH' is defined on line 630, but no explicit reference was found in the text == Unused Reference: 'VXLAN-GPE' is defined on line 631, but no explicit reference was found in the text == Outdated reference: A later version (-05) exists of draft-smith-vxlan-group-policy-03 Summary: 2 errors (**), 0 flaws (~~), 13 warnings (==), 1 comment (--).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 Expires: January 2, 2020 July 1, 2019 7 NVO3 Encapsulation Considerations 8 draft-ietf-nvo3-encap-03 10 Abstract 12 As communicated by WG Chairs, the IETF NVO3 chairs and Routing Area 13 director have chartered a design team to take forward the 14 encapsulation discussion and see if there is potential to design a 15 common encapsulation that addresses the various technical concerns. 17 There are implications of different encapsulations in real 18 environments consisting of both software and hardware implementations 19 and spanning multiple data centers. For example, OAM functions such 20 as path MTU discovery become challenging with multiple encapsulations 21 along the data path. 23 The design team recommend Geneve with few modifications as the common 24 encapsulation, more details are described in section 7. 26 Status of this Memo 28 This Internet-Draft is submitted to IETF in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF), its areas, and its working groups. Note that 33 other groups may also distribute working documents as 34 Internet-Drafts. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 The list of current Internet-Drafts can be accessed at 42 http://www.ietf.org/1id-abstracts.html 44 The list of Internet-Draft Shadow Directories can be accessed at 45 http://www.ietf.org/shadow.html 47 Copyright and License Notice 49 Copyright (c) 2019 IETF Trust and the persons identified as the 50 document authors. All rights reserved. 52 This document is subject to BCP 78 and the IETF Trust's Legal 53 Provisions Relating to IETF Documents 54 (http://trustee.ietf.org/license-info) in effect on the date of 55 publication of this document. Please review these documents 56 carefully, as they describe your rights and restrictions with respect 57 to this document. Code Components extracted from this document must 58 include Simplified BSD License text as described in Section 4.e of 59 the Trust Legal Provisions and are provided without warranty as 60 described in the Simplified BSD License. 62 Table of Contents 64 1. Problem Statement . . . . . . . . . . . . . . . . . . . . . . . 4 65 2. Design Team Goals . . . . . . . . . . . . . . . . . . . . . . . 4 66 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 4 67 4. Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . 4 68 5. Issues with current Encapsulations . . . . . . . . . . . . . . 5 69 5.1 Geneve . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 70 5.2 GUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 71 5.3 VXLAN-GPE . . . . . . . . . . . . . . . . . . . . . . . . . 5 72 6. Common Encapsulation Considerations . . . . . . . . . . . . . . 6 73 6.1 Current Encapsulations . . . . . . . . . . . . . . . . . . . 6 74 6.2 Useful Extensions Use cases . . . . . . . . . . . . . . . . 6 75 6.2.1. Telemetry extensions. . . . . . . . . . . . . . . . . . 6 76 6.2.2. Security/Integrity extensions . . . . . . . . . . . . . 7 77 6.2.3. Group Base Policy . . . . . . . . . . . . . . . . . . . 7 78 6.3 Hardware Considerations . . . . . . . . . . . . . . . . . . 7 79 6.4 Extension Size . . . . . . . . . . . . . . . . . . . . . . . 8 80 6.5 Extension Ordering . . . . . . . . . . . . . . . . . . . . . 9 81 6.6 TLV vs Bit Fields . . . . . . . . . . . . . . . . . . . . . 9 82 6.7 Control Plane Considerations . . . . . . . . . . . . . . . . 10 83 6.8 Split NVE . . . . . . . . . . . . . . . . . . . . . . . . . 11 84 6.9 Larger VNI Considerations . . . . . . . . . . . . . . . . . 11 85 7. Design team recommendations . . . . . . . . . . . . . . . . . . 11 86 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14 87 9. Security Considerations . . . . . . . . . . . . . . . . . . . . 14 88 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 14 89 10.1 Normative References . . . . . . . . . . . . . . . . . . . 14 90 10.2 Informative References . . . . . . . . . . . . . . . . . . 15 91 11. Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . 15 92 11.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 15 93 11.2. Extensibility . . . . . . . . . . . . . . . . . . . . . . 15 94 11.2.1. Native Extensibility Support . . . . . . . . . . . . 15 95 11.2.2. Extension Parsing . . . . . . . . . . . . . . . . . . 15 96 11.2.3. Critical Extensions . . . . . . . . . . . . . . . . . 16 97 11.2.4. Maximal Header Length . . . . . . . . . . . . . . . . 16 98 11.3. Encapsulation Header . . . . . . . . . . . . . . . . . . 16 99 11.3.1. Virtual Network Identifier (VNI) . . . . . . . . . . 16 100 11.3.2. Next Protocol . . . . . . . . . . . . . . . . . . . . 16 101 11.3.3. Other Header Fields . . . . . . . . . . . . . . . . . 17 102 11.4. Comparison Summary . . . . . . . . . . . . . . . . . . . 17 103 Authors' Addresses (In alphabetical order) . . . . . . . . . . . . 18 105 1. Problem Statement 107 As communicated by WG Chairs, the NVO3 WG charter states that it may 108 produce requirements for network virtualization data planes based on 109 encapsulation of virtual network traffic over an IP-based underlay 110 data plane. Such requirements should consider OAM and security. Based 111 on these requirements the WG will select, extend, and/or develop one 112 or more data plane encapsulation format(s). 114 This has led to drafts describing three encapsulations being adopted 115 by the working group: 117 - draft-ietf-nvo3-geneve-03 119 - draft-ietf-nvo3-gue-04 121 - draft-ietf-nvo3-vxlan-gpe-02 123 Discussion on the list and in face-to-face meetings has identified a 124 number of technical problems with each of these encapsulations. 125 Furthermore, there was clear consensus at the IETF meeting in Berlin 126 that it is undesirable for the working group to progress more than 127 one data plane encapsulation. Although consensus could not be reached 128 on the list, the overall consensus was for a single encapsulation 129 (RFC2418, Section 3.3). Nonetheless there has been resistance to 130 converging on a single encapsulation format. 132 2. Design Team Goals 134 As communicated by WG Chairs, the design team should take one of the 135 proposed encapsulations and enhance it to address the technical 136 concerns. The simple evolution of deployed networks as well as 137 applicability to all locations in the NVO3 architecture are goals. 138 The DT should specifically avoid a design that is burdensome on 139 hardware implementations, but should allow future extensibility. The 140 chosen design should also operate well with ICMP and in ECMP 141 environments. If further extensibility is required, then it should be 142 done in such a manner that it does not require the consent of an 143 entity outside of the IETF. 145 3. Terminology 147 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 148 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 149 document are to be interpreted as described in RFC 2119 [RFC2119]. 151 4. Abbreviations 152 NVO3 Network Virtualization Overlays over Layer 3 154 OAM Operations, Administration, and Maintenance 156 TLV Type, Length, and Value 158 VNI Virtual Network Identifier 160 NVE Network Virtualization Edge 162 NVA Network Virtualization Authority 164 NIC Network interface card 166 Transit device Underlay network devices between NVE(s). 168 5. Issues with current Encapsulations 170 As summarized by WG Chairs. 172 5.1 Geneve 174 - Can't be implemented cost-effectively in all use cases because 175 variable length header and order of the TLVs makes is costly (in 176 terms of number of gates) to implement in hardware 178 - Header doesn't fit into largest commonly available parse buffer 179 (256 bytes in NIC). Cannot justify doubling buffer size unless it is 180 mandatory for hardware to process additional option fields. 182 5.2 GUE 184 - There were a significant number of objections related to the 185 complexity of implementation in hardware, similar to those noted for 186 Geneve above. 188 5.3 VXLAN-GPE 190 - GPE is not day-1 backwards compatible with VXLAN. Although the 191 frame format is similar, it uses a different UDP port, so would 192 require changes to existing implementations even if the rest of the 193 GPE frame is the same. 195 - GPE is insufficiently extensible. Numerous extensions and options 196 have been designed for GUE and Geneve. Note that these have not yet 197 been validated by the WG. 199 - Security e.g. of the VNI has not been addressed by GPE. Although a 200 shim header could be used for security and other extensions, this has 201 not been defined yet and its implications on offloading in NICs are 202 not understood. 204 6. Common Encapsulation Considerations 206 6.1 Current Encapsulations 208 Appendix A includes a detailed comparison between the three proposed 209 encapsulations. The comparison indicates several common properties, 210 but also three major differences among the encapsulations: 212 - Extensibility: Geneve and GUE were defined with built-in 213 extensibility, while VXLAN-GPE is not inherently extensible. Note 214 that any of the three encapsulations can be extended using the 215 Network Service Header (NSH). 217 - Extension method: Geneve is extensible using Type/Length/Value 218 (TLV) fields, while GUE uses a small set of possible extensions, and 219 a set of flags that indicate which extension is present. 221 - Length field: Geneve and GUE include a Length field, indicating the 222 length of the encapsulation header, while VXLAN-GPE does not include 223 such a field. 225 6.2 Useful Extensions Use cases 227 Non vendor specific TLV MUST follow the standardization process. The 228 following use cases for extensions shows that there is a strong 229 requirement to support variable length extensions with possible 230 different subtypes. 232 6.2.1. Telemetry extensions. 234 In several scenarios it is beneficial to make information about the 235 path a packet took through the network or through a network device as 236 well as associated telemetry information available to the operator. 238 This includes not only tasks like debugging, troubleshooting, as well 239 as network planning and network optimization but also policy or 240 service level agreement compliance checks. 242 Packet scheduling algorithms, especially for balancing traffic across 243 equal cost paths or links, often leverage information contained 244 within the packet, such as protocol number, IP-address or MAC- 245 address. Probe packets would thus either need to be sent from the 246 exact same endpoints with the exact same parameters, or probe packets 247 would need to be artificially constructed as "fake" packets and 248 inserted along the path. Both approaches are often not feasible from 249 an operational perspective, be it that access to the end-system is 250 not feasible, or that the diversity of parameters and associated 251 probe packets to be created is simply too large. An in-band telemetry 252 mechanism in extensions is an alternative in those cases. 254 6.2.2. Security/Integrity extensions 256 Since the currently proposed NVO3 encapsulations do not protect their 257 headers a single bit corruption in the VNI field could deliver a 258 packet to the wrong tenant. Extensions are needed to use any 259 sophisticated security. 261 The possibility of VNI spoofing with an NVO3 protocol is exacerbated 262 by the use of UDP. Systems typically have no restrictions on 263 applications being able to send to any UDP port so an unprivileged 264 application can trivially spoof for instance, VXLAN packets, 265 including using arbitrary VNIs. 267 One can envision HMAC-like support in some NVO3 extension to 268 authenticate the header and the outer IP addresses, thereby 269 preventing attackers from injecting packets with spoofed VNIs. 271 An other aspect of security is payload security. Essentially this is 272 to make packets that look like IP|UDP|NVO3 Encap|DTLS/IPSEC-ESP 273 Extension|payload. This is nice since we still have the UDP header 274 for ECMP, the NVO3 header is in plain text so it can by read by 275 network elements, and different security or other payload transforms 276 can be supported on a single UDP port (we don't need a separate UDP 277 for DTLS/IPSEC). 279 6.2.3. Group Base Policy 281 Another use case would be to carry the Group Based Policy (GBP) 282 source group information within a NVO3 header extension in a similar 283 manner as has been implemented for VXLAN [VXLAN-GBP]. This allows 284 various forms of policy such as access control and QoS to be applied 285 between abstract groups rather than coupled to specific endpoint 286 addresses. 288 6.3 Hardware Considerations 290 Hardware restrictions should be taken into consideration along with 291 future hardware enhancements that may provide more flexible metadata 292 processing. However, the set of options that need to and will be 293 implemented in hardware will be a subset of what is implemented in 294 software, since software NVEs are likely to grow features, and hence 295 option support, at a more rapid rate. 297 We note that it is hard to predict which options will be implemented 298 in which piece of hardware and when. That depends on whether the 299 hardware will be in the form of a NIC providing increasing offload 300 capabilities to software NVEs, or a switch chip being used as an NVE 301 gateway towards non-NVO3 parts of the network, or even an transit 302 devices that participates in the NVO3 dataplane e.g. for OAM 303 purposes. 305 A result of this is that it doesn't look useful to prescribe some 306 order of the option so that the ones that are likely to be 307 implemented in hardware come first; we can't decide such an order 308 when we define the options, however a control plane can enforce such 309 order for some hardware implementations. 311 We do know that hardware needs to initially be able to efficiently 312 skip over the NVO3 header to find the inner payload. That is needed 313 for both NICs doing e.g. TCP offload and transit devices and NVEs 314 applying policy/ACLs to the inner payload. 316 6.4 Extension Size 318 Extension header length has a significant impact to hardware and 319 software implementations. A total header length that is too small 320 will unnecessarily constrained software flexibility. A total header 321 length that is too large will place a nontrivial cost on hardware 322 implementations. Thus, the design team recommends that there be a 323 minimum and maximum total extension header length selected. The 324 maximum total header length is determined by the bits allocated for 325 the total extension header length field. The risk with this approach 326 is that it may be difficult to extend the total header size in the 327 future. The minimum total header length is determined by a 328 requirement in the specifications that all implementations must meet. 329 The risk with this approach is that all implementations will only 330 implement the minimum total header length which would then become the 331 de facto maximum total header length. The recommended minimum total 332 header length is 64 bytes. 334 Single Extension size should always be 4 bytes aligned. 336 The maximum length of a single option should be large enough to meet 337 the different extension use case requirements e.g. in-band telemetry 338 and future use. 340 6.5 Extension Ordering 342 In order to support hardware nodes at the tunnel endpoint or at the 343 transit that can process one or few extensions TLVs in TCAM. A 344 control plane in such a deployment can signal a capability to ensure 345 a specific TLV will always appear in a specific order for example the 346 first one in the packet. 348 The order of the TLVs should be HW friendly for both the sender and 349 the receiver and possibly the transit node too. 351 Transit nodes doesn't participate in control plane communication 352 between the end points and are not required to process the options 353 however, if they do, they need to process only a small subset of 354 options that will be consumed by tunnel endpoints. 356 6.6 TLV vs Bit Fields 358 If there is a well-known initial set of options that are likely to be 359 implemented in software and in hardware, it can be efficient to use 360 the bit-field approach as in GUE. However, as described in section 361 6.3, if options are added over time and different subsets of options 362 are likely to be implemented in different pieces of hardware, then it 363 would be hard for the IETF to specify which options should get the 364 early bit fields. TLVs are a lot more flexible, which avoids the need 365 to determine the relative importance different options. However, 366 general TLV of arbitrary order, size, and repetition of the same 367 order is difficult to implement in hardware. A middle ground is to 368 use TLV with restrictions on the size and alignment, observing that 369 individual TLVs can have a fixed length, and support in the control 370 plane such that an NVE will only receive options that to needs and 371 implements. The control plane approach can potentially be used to 372 control the order of the TLVs sent to a particular NVE. Note that 373 transit devices are not likely to participate in the control plane 374 hence to the extent that they need to participate in option 375 processing they need more effort, But transit devices would have 376 issues with future GUE bits being defined for future options as well. 378 A benefit of TLVs from a HW perspective is that they are self 379 describing i.e., all the information is in the TLV. In a Bit fields 380 approach the hardware needs to look up the bit to determine the 381 length of the data associated with the bit through some separate 382 table, which would add hardware complexity. 384 There are use cases where multiple modules of software are running on 385 NVE. This can be modules such as a diagnostic module by one vendor 386 that does packet sampling and another module from a different vendor 387 that does a firewall. Using a TLV format, it is easier to have 388 different software modules process different TLVs, which could be 389 standard extensions or vendor specific extensions defined by the 390 different vendors, without conflicting with each other. This can help 391 with hardware modularity as well. There are some implementations with 392 options that allows different software like mac learning and security 393 handle different options. 395 6.7 Control Plane Considerations 397 Given that we want to allow large flexibility and extensibility for 398 e.g. software NVEs, yet be able to support key extensions in less 399 flexible e.g. hardware NVEs, it is useful to consider the control 400 plane. By control plane in this context we mean both protocols such 401 as EVPN and others, and also deployment specific configuration. 403 If each NVE can express in the control plane that they only care 404 about particular extensions (could be a single extension, or a few), 405 and the source NVEs only include requested extensions in the NVO3 406 packets, then the target NVE can both use a simpler parser (e.g., a 407 TCAM might be usable to look for a single NVO3 extension) and the 408 depth of the inner payload in the NVO3 packet will be minimized. 409 Furthermore, if the target NVE cares about a few extensions and can 410 express in the control plane the desired order of those extensions in 411 the NVO3 packets, then it can provide useful functionality with 412 minimal hardware requirements. 414 Note that transit devices that are not aware of the NVO3 extensions 415 somewhat benefit from such an approach, since the inner payload is 416 less deep in the packet if no extraneous extensions are included in 417 the packet. However, in general a transit device is not likely to 418 participate in the NVO3 control plane. (However, configuration 419 mechanisms can take into account limitations of the transit devices 420 used in particular deployments.) 422 Note that in this approach different NVEs could desire different 423 (sets of) extensions, which means that the source NVE needs to be 424 able to place different sets of extensions in different NVO3 packets, 425 and perhaps in different order. It also assumes that underlay 426 multicast or replication servers are not used together with NVO3 427 extensions. 429 There is a need to consider mandatory extensions versus optional 430 extensions. Mandatory extensions require the receiver to drop the 431 packet if the extension is unknown. A control plane mechanism can 432 prevent the need for dropping unknown extensions, since they would 433 not be included to targets that do not support them. 435 The control planes defined today need to add the ability to describe 436 the different encapsulations. Thus perhaps EVPN, and any other 437 control plane protocol that the IETF defines, should have a way to 438 enumerate the supported NVO3 extensions and their order. 440 The WG should consider developing a separate draft on guidance for 441 option processing and control plane participation. This should 442 provide examples/guidance on range of usage models and deployments 443 scenarios for specific options and ordering that are relevant for 444 that specific deployment. This includes end points and middle boxes 445 using the options. So, having the control plane negotiate the 446 constraints is most appropriate and flexible way to address these 447 requirements. 449 6.8 Split NVE 451 If the working group sees a need for having the hosts send and 452 receive options in a split NVE case, this is possible using any of 453 the existing extensible encapsulations (Geneve, GUE, GPE+NSH) by 454 defining a way to carry those over other transports. NSH can already 455 be used over different transports. 457 If we need to do this with other encapsulations it can be done by 458 defining an Ether type for other encapsulations so that it can be 459 carried over Ethernet and 802.1Q. 461 If we need to carry other encapsulations over MPLS, it would require 462 an EVPN control plane to signal that other encapsulation header + 463 options will be present in front of the L2 packet. The VNI can be 464 ignored in the header, and the MPLS label will be the one used to 465 identify the EVPN L2 instance. 467 6.9 Larger VNI Considerations 469 We discussed whether we should make VNI 32-bits or larger. The 470 benefit of 24-bit VNI would be to avoid unnecessary changes with 471 existing proposals and implementations that are almost all, if not 472 all, are using 24-bit VNI. If we need a larger VNI, an extension can 473 be used to support that. 475 7. Design team recommendations 477 We concluded that Geneve is most suitable as a starting point for 478 proposed standard for network virtualization, for the following 479 reasons: 481 1. We studied whether VNI should be in base header or in extensions 482 and whether it should be 24-bit or 32-bit. The design team agreed 483 that VNI is critical information for network virtualization and MUST 484 be present in all packets. Design team also agreed that 24-bit VNI 485 matches the existing widely used encapsulation format i.e. VxLAN and 486 NVGRE and hence more suitable to use going forward. 488 2. Geneve has the total options length that allow skipping over the 489 options for NIC offload operations, and will allow transit devices to 490 view flow information in the inner payload. 492 3. We considered the option of using NSH with VxLAN-GPE but given 493 that NSH is targeted at service chaining and contains service 494 chaining information, it is less suitable for the network 495 virtualization use case. The other downside for VxLAN-GPE was lack of 496 header length in VxLAN-GPE and hence makes skipping over the headers 497 to process inner payload more difficult. Total Option Length is 498 present in Geneve. It is not possible to skip any options in the 499 middle with VxLAN-GPE. In principle a split between a base header and 500 a header with options is interesting (whether that options header is 501 NSH or some new header without ties to a service path). We explored 502 whether it would make sense to either use NSH for this, or define a 503 new NVO3 options header. However, we observed that this makes it 504 slightly harder to find the inner payload since the length field is 505 not in the NVO3 header itself. Thus one more field would have to be 506 extracted to compute the start of the inner payload. Also, if the 507 experience with IPv6 extension headers is a guidance, there would be 508 a risk that key pieces of hardware might not implement the options 509 header, resulting in future calls to deprecate its use. Making the 510 options part of the base NVO3 header has less of those issues. Even 511 though the implementation of any particular option can not be 512 predicted ahead of time, the option mechanism and ability to skip the 513 options is likely to be broadly implemented. 515 4. We compared the TLV vs Bit-fields style extension and it was 516 deemed that parsing both TLV and bit-fields is expensive and while 517 bit-fields may be simpler to parse, it is also more restrictive and 518 requires guessing which extensions will be widely implemented so they 519 can get early bit assignments, given that half the bits are already 520 assigned in GUE, a widely deployed extension may appear in a flag 521 extension, and this will require extra processing, to dig the flag 522 from the flag extension and then look for the extension itself. As 523 well Bit-fields are not flexible enough to address the requirements 524 from OAM, Telemetry and security extensions, for variable length 525 option and different subtypes of the same option. While TLV are more 526 flexible, a control plane can restrict the number of option TLVs as 527 well the order and size of the TLVs to make it simpler for a 528 dataplane implementation to handle. 530 5. We briefly discussed multi-vendor NVE case, and the need to allow 531 vendors to put their own extensions in the NVE header. This is 532 possible with TLVs. 534 6. We also agreed that the C bit in Geneve is helpful to allow 535 receiver NVE to easily decide whether to process options or not. For 536 example a UUID based packet trace and how an optional extension such 537 as that can be ignored by receiver NVE and thus make it easy for NVE 538 to skip over the options. Thus the C-bit remains as defined in 539 Geneve. 541 7. There are already some extensions that are being discussed (see 542 section 6.2) of varying sizes, by using Geneve option it is possible 543 to get in band parameters like: switch id, ingress port, egress port, 544 internal delay, and queue in telemetry defined extension TLV from 545 switches. It is also possible to add Security extension TLVs like 546 HMAC and DTLS/IPSEC to authenticate the Geneve packet header and 547 secure the Geneve packet payload by software or hardware tunnel 548 endpoints. As well, a Group Based Policy extension TLV can be 549 carried. 551 8. There are implemented Geneve options today in production. There 552 are as well new HW supporting Geneve TLV parsing. In addition In-band 553 Telemetry (INT) specification being developed by P4.org illustrates 554 the option of INT meta data carried over Geneve. OVN/OVS have also 555 defined some option TLV(s) for Geneve. 557 9. The DT has addressed the usage models while considering the 558 requirements and implementations in general that includes software 559 and hardware. 561 There seems to be interest to standardize some well known secure 562 option TLVs to secure the header and payload to guarantee 563 encapsulation header integrity and tenant data privacy. The design 564 team recommends that the working group consider standardizing such 565 option(s). 567 We recommend the following enhancements to Geneve to make it more 568 suitable to hardware and yet provide the flexibility for software: 570 We would propose a text such as, while TLV are more flexible, a 571 control plane can restrict the number of option TLVs as well the 572 order and size of the TLVs to make it simpler for a data plane 573 implementation in software or hardware to handle. For example, there 574 may be some critical information such as secure hash that must be 575 processed in certain order at lowest latency. 577 A control plane can negotiate a subset of option TLVs and certain TLV 578 ordering, as well can limit the total number of option TLVs present 579 in the packet, for example, to allow hardware capable of processing 580 fewer options. Hence, the control planes need to have the ability to 581 describe the supported TLVs subset and their order. 583 The Geneve draft could specify that the subset and order of option 584 TLVs should be configurable for each remote NVE in the absence of a 585 protocol control plane. 587 We recommend Geneve to follow fragmentation recommendations in 588 overlay services like PWE3, and L2/L3 VPN recommendation to guarantee 589 larger MTU for the tunnel overhead 590 https://tools.ietf.org/html/rfc3985#section-5.3 592 We request Geneve to provide a recommendation for critical bit 593 processing - text could look like how critical bits can be used with 594 control plane specifying the critical options. 596 Given that there is a telemetry option use case for a length of 256 597 bytes, we recommend Geneve to increase the Single TLV option length 598 to 256. 600 We request Geneve to address Requirements for OAM considerations for 601 alternate marking and for performance measurements that need 2 bits 602 in the header. And clarify the need of the current OAM bit in the 603 Geneve Header. 605 We recommend the WG to work on security options for Geneve. 607 8. Acknowledgements 609 The authors would like to thank Tom Herbert for providing the 610 motivation for the Security/Integrity extension, and for his valuable 611 comments, and would like to thank T. Sridhar for his valuable 612 comments and feedback. 614 9. Security Considerations 616 This document does not introduce any additional security constraints. 618 10. References 620 10.1 Normative References 622 [KEYWORDS] Bradner, S., "Key words for use in RFCs to Indicate 623 Requirement Levels", BCP 14, RFC 2119, March 1997. 625 10.2 Informative References 627 [Geneve] Generic Network Virtualization Encapsulation [I-D.ietf-nvo3- 628 geneve] 629 [GUE] Generic UDP Encapsulation [I-D.ietf-nvo3-gue] 630 [NSH] Network Service Header [I-D.ietf-sfc-nsh] 631 [VXLAN-GPE] Virtual eXtensible Local Area Network - Generic Protocol 632 Extension [I-D.ietf-nvo3-vxlan-gpe] 634 [VXLAN-GBP] VXLAN Group Policy Option - [I-D.draft-smith-vxlan-group- 635 policy-03] 637 11. Appendix A 639 11.1. Overview 641 This section presents a comparison of the three NVO3 encapsulation 642 proposals, Geneve, GUE, and VXLAN-GPE. The three encapsulations use 643 an outer UDP/IP transport. Geneve and VXLAN-GPE use an 8-octet 644 header, while GUE uses a 4-octet header. In addition to the base 645 header, optional extensions may be included in the encapsulation, as 646 discussed in Section 3.2 below. 648 11.2. Extensibility 650 11.2.1. Native Extensibility Support 652 The Geneve and GUE encapsulations both enable optional headers to be 653 incorporated at the end of the base encapsulation header. 655 VXLAN-GPE does not provide native support for header extensions. 656 However, as discussed in [I-D.ietf-nvo3-vxlan-gpe], extensibility can 657 be attained to some extent if the Network Service Header (NSH) [I- 658 D.ietf-sfc-nsh] is used immediately following the VXLAN-GPE header. 659 NSH supports either a fixed-size extension (MD Type 1), or a 660 variable-size TLV-based extension (MD Type 2). It should be noted 661 that NSH-over-VXLAN-GPE implies an additional overhead of the 8- 662 octets NSH header, in addition to the VXLAN-GPE header. 664 11.2.2. Extension Parsing 666 The Geneve Variable Length Options are defined as 667 Type/Length/Value(TLV) extensions. Similarly, VXLAN-GPE, when using 668 NSH, can include NSH TLV-based extensions. In contrast, GUE defines 669 a small set of possible extension fields (proposed in [I-D.herbert- 670 gue-extensions]), and a set of flags in the GUE header that indicate 671 for each extension type whether it is present or not. 673 TLV-based extensions, as defined in Geneve, provide the flexibility 674 for a large number of possible extension types. Similar behavior can 675 be supported in NSH-over-VXLAN-GPE when using MD Type 2. The flag- 676 based approach taken in GUE strives to simplify implementations by 677 defining a small number of possible extensions, used in a fixed 678 order. 680 The Geneve and GUE headers both include a length field, defining the 681 total length of the encapsulation, including the optional extensions. 683 The length field simplifies the parsing of transit devices that skip 684 the encapsulation header without parsing its extensions. 686 11.2.3. Critical Extensions 688 The Geneve encapsulation header includes the 'C' field, which 689 indicates whether the current Geneve header includes critical 690 options, which must be parsed by the tunnel endpoint. If the endpoint 691 is not able to process the critical option, the packet is discarded. 693 11.2.4. Maximal Header Length 695 The maximal header length in Geneve, including options, is 260 696 octets. GUE defines the maximal header to be 128 octets. VXLAN-GPE 697 uses a fixed-length header of 8 octets, unless NSH-over-VXLAN-GPE is 698 used, yielding an encapsulation header of up to 264 octets. 700 11.3. Encapsulation Header 702 11.3.1. Virtual Network Identifier (VNI) 704 The Geneve and VXLAN-GPE headers both include a 24-bit VNI field. 705 GUE, on the other hand, enables the use of a 32-bit field called 706 VNID; this field is not included in the GUE header, but was defined 707 as an optional extension in [I-D.herbert-gue-extensions]. 709 The VXLAN-GPE header includes the 'I' bit, indicating that the VNI 710 field is valid in the current header. A similar indicator is defined 711 as a flag in the GUE header [I-D.herbert-gue-extensions]. 713 11.3.2. Next Protocol 715 The three encapsulation headers include a field that specifies the 716 type of the next protocol header, which resides after the NVO3 717 encapsulation header. The Geneve header includes a 16-bit field that 718 uses the IEEE Ethertype convention. GUE uses an 8-bit field, which 719 uses the IANA Internet protocol numbering. The VXLAN-GPE header 720 incorporates an 8-bit Next Protocol field, using a VXLAN-GPE-specific 721 registry, defined in [I-D.ietf-nvo3-vxlan-gpe]. 723 The VXLAN-GPE header also includes the 'P' bit, which explicitly 724 indicates whether the Next Protocol field is present in the current 725 header. 727 11.3.3. Other Header Fields 729 The OAM bit, which is defined in Geneve and in VXLAN-GPE, indicates 730 whether the current packet is an OAM packet. The GUE header includes 731 a similar field, but uses different terminology; the GUE 'C-bit' 732 specifies whether the current packet is a control packet. Note that 733 the GUE control bit can potentially be used in a large set of 734 protocols that are not OAM protocols. However, the control packet 735 examples discussed in [I-D.ietf-nvo3-gue] are OAM-related. 737 Each of the three NVO3 encapsulation headers includes a 2-bit Version 738 field, which is currently defined to be zero. 740 The Geneve and VXLAN-GPE headers include reserved fields; 14 bits in 741 the Geneve header, and 27 bits in the VXLAN-GPE header are reserved. 743 11.4. Comparison Summary 745 The following table summarizes the comparison between the three NVO3 746 encapsulations. 748 +----------------+----------------+----------------+----------------+ 749 | | Geneve | GUE | VXLAN-GPE | 750 +----------------+----------------+----------------+----------------+ 751 | Outer transport| UDP/IP | UDP/IP | UDP/IP | 752 +----------------+----------------+----------------+----------------+ 753 | Base header | 8 octets | 4 octets | 8 octets | 754 | length | | | (16 octets | 755 | | | | using NSH) | 756 +----------------+----------------+----------------+----------------+ 757 | Extensibility |Variable length |Extension fields| No native ext- | 758 | | options | | ensibility. | 759 | | | | Extensible | 760 | | | | using NSH. | 761 +----------------+----------------+----------------+----------------+ 762 | Extension | TLV-based | Flag-based | TLV-based | 763 | parsing method | | |(using NSH with | 764 | | | | MD Type 2) | 765 +----------------+----------------+----------------+----------------+ 766 | Extension | Variable | Fixed | Variable | 767 | order | | | (using NSH) | 768 +----------------+----------------+----------------+----------------+ 769 | Length field | + | + | - | 770 +----------------+----------------+----------------+----------------+ 771 | Max Header | 260 octets | 128 octets | 8 octets | 772 | Length | | |(264 using NSH) | 773 +----------------+----------------+----------------+----------------+ 774 | Critical exte- | + | - | - | 775 | nsion bit | | | | 776 +----------------+----------------+----------------+----------------+ 777 | VNI field size | 24 bits | 32 bits | 24 bits | 778 | | | (extension) | | 779 +----------------+----------------+----------------+----------------+ 780 | Next protocol | 16 bits | 8 bits | 8 bits | 781 | field | Ethertype | Internet prot- | New registry | 782 | | registry | ocol registry | | 783 +----------------+----------------+----------------+----------------+ 784 | Next protocol | - | - | + | 785 | indicator | | | | 786 +----------------+----------------+----------------+----------------+ 787 | OAM / control | OAM bit | Control bit | OAM bit | 788 | field | | | | 789 +----------------+----------------+----------------+----------------+ 790 | Version field | 2 bits | 2 bits | 2 bits | 791 +----------------+----------------+----------------+----------------+ 792 | Reserved bits | 14 bits | - | 27 bits | 793 +----------------+----------------+----------------+----------------+ 795 Figure 1: NVO3 Encapsulation Comparison 797 Authors' Addresses (In alphabetical order) 799 Sami Boutros 800 VMware 801 Email: boutross@vmware.com 803 Ilango Ganga 804 Intel 805 Email: ilango.s.ganga@intel.com 807 Pankaj Garg 808 Microsoft 809 Email: pankajg@microsoft.com 811 Rajeev Manur 812 Broadcom 813 Email: rajeev.manur@broadcom.com 815 Tal Mizrahi 816 Marvell 817 Email: talmi@marvell.com 819 David Mozes 820 Email: mosesster@gmail.com 822 Erik Nordmark 823 Email: nordmark@sonic.net 825 Michael Smith 826 Cisco 827 Email: michsmit@cisco.com 829 Sam Aldrin 830 Google 831 Email: aldrin.ietf@gmail.com 833 Ignas Bagdonas 834 Equinix 835 Email: ibagdona.ietf@gmail.com