idnits 2.17.1 draft-ietf-rtgwg-dt-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.) ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 292: '...italized keyword MUST is used as defin...' RFC 2119 keyword, line 369: '...lancing procedure MUST choose the same...' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 2015) is 3209 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'I-D.xu-bier-encapsulation' is defined on line 1821, but no explicit reference was found in the text == Outdated reference: A later version (-19) exists of draft-ietf-tsvwg-rfc5405bis-02 ** Obsolete normative reference: RFC 2460 (Obsoleted by RFC 8200) ** Obsolete normative reference: RFC 5405 (Obsoleted by RFC 8085) ** Obsolete normative reference: RFC 6830 (Obsoleted by RFC 9300, RFC 9301) == Outdated reference: A later version (-02) exists of draft-herbert-remotecsumoffload-01 == Outdated reference: A later version (-08) exists of draft-ietf-nvo3-arch-03 == Outdated reference: A later version (-11) exists of draft-ietf-sfc-architecture-09 == Outdated reference: A later version (-15) exists of draft-ietf-tsvwg-circuit-breaker-01 == Outdated reference: A later version (-19) exists of draft-ietf-tsvwg-gre-in-udp-encap-07 == Outdated reference: A later version (-12) exists of draft-saldana-tsvwg-simplemux-02 == Outdated reference: A later version (-06) exists of draft-xu-bier-encapsulation-02 Summary: 5 errors (**), 0 flaws (~~), 10 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTGWG E. Nordmark (ed) 3 Internet-Draft Arista Networks 4 Intended status: Informational A. Tian 5 Expires: December 3, 2015 Ericsson Inc. 6 J. Gross 7 VMware 8 J. Hudson 9 Brocade Communications Systems, 10 Inc. 11 L. Kreeger 12 Cisco Systems, Inc. 13 P. Garg 14 Microsoft 15 P. Thaler 16 Broadcom Corporation 17 T. Herbert 18 Facebook 19 June 2015 21 Encapsulation Considerations 22 draft-ietf-rtgwg-dt-encap-00 24 Abstract 26 The IETF Routing Area director has chartered a design team to look at 27 common issues for the different data plane encapsulations being 28 discussed in the NVO3 and SFC working groups and also in the BIER 29 BoF, and also to look at the relationship between such encapsulations 30 in the case that they might be used at the same time. The purpose of 31 this design team is to discover, discuss and document considerations 32 across the different encapsulations in the different WGs/BoFs so that 33 we can reduce the number of wheels that need to be reinvented in the 34 future. 36 Status of this Memo 38 This Internet-Draft is submitted in full conformance with the 39 provisions of BCP 78 and BCP 79. 41 Internet-Drafts are working documents of the Internet Engineering 42 Task Force (IETF). Note that other groups may also distribute 43 working documents as Internet-Drafts. The list of current Internet- 44 Drafts is at http://datatracker.ietf.org/drafts/current/. 46 Internet-Drafts are draft documents valid for a maximum of six months 47 and may be updated, replaced, or obsoleted by other documents at any 48 time. It is inappropriate to use Internet-Drafts as reference 49 material or to cite them other than as "work in progress." 51 This Internet-Draft will expire on December 3, 2015. 53 Copyright Notice 55 Copyright (c) 2015 IETF Trust and the persons identified as the 56 document authors. All rights reserved. 58 This document is subject to BCP 78 and the IETF Trust's Legal 59 Provisions Relating to IETF Documents 60 (http://trustee.ietf.org/license-info) in effect on the date of 61 publication of this document. Please review these documents 62 carefully, as they describe your rights and restrictions with respect 63 to this document. Code Components extracted from this document must 64 include Simplified BSD License text as described in Section 4.e of 65 the Trust Legal Provisions and are provided without warranty as 66 described in the Simplified BSD License. 68 Table of Contents 70 1. Design Team Charter . . . . . . . . . . . . . . . . . . . . . 4 71 2. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 72 3. Common Issues . . . . . . . . . . . . . . . . . . . . . . . . 6 73 4. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 74 5. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 7 75 6. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 8 76 7. Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 77 8. Next-protocol indication . . . . . . . . . . . . . . . . . . . 9 78 9. MTU and Fragmentation . . . . . . . . . . . . . . . . . . . . 11 79 10. OAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 80 11. Security Considerations . . . . . . . . . . . . . . . . . . . 14 81 11.1. Encapsulation-specific considerations . . . . . . . . . . 14 82 11.2. Virtual network isolation . . . . . . . . . . . . . . . . 16 83 11.3. Packet level security . . . . . . . . . . . . . . . . . . 17 84 11.4. In summary: . . . . . . . . . . . . . . . . . . . . . . . 17 85 12. QoS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 86 13. Congestion Considerations . . . . . . . . . . . . . . . . . . 19 87 14. Header Protection . . . . . . . . . . . . . . . . . . . . . . 20 88 15. Extensibility Considerations . . . . . . . . . . . . . . . . . 22 89 16. Layering Considerations . . . . . . . . . . . . . . . . . . . 25 90 17. Service model . . . . . . . . . . . . . . . . . . . . . . . . 26 91 18. Hardware Friendly . . . . . . . . . . . . . . . . . . . . . . 27 92 18.1. Considerations for NIC offload . . . . . . . . . . . . . 28 93 19. Middlebox Considerations . . . . . . . . . . . . . . . . . . . 32 94 20. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . 33 95 21. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 34 96 22. Open Issues . . . . . . . . . . . . . . . . . . . . . . . . . 34 97 23. Change Log . . . . . . . . . . . . . . . . . . . . . . . . . . 35 98 24. References . . . . . . . . . . . . . . . . . . . . . . . . . . 36 99 24.1. Normative References . . . . . . . . . . . . . . . . . . 36 100 24.2. Informative References . . . . . . . . . . . . . . . . . 37 101 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 40 103 1. Design Team Charter 105 There have been multiple efforts over the years that have resulted in 106 new or modified data plane behaviors involving encapsulations. That 107 includes IETF efforts like MPLS, LISP, and TRILL but also industry 108 efforts like VXLAN and NVGRE. These collectively can be seen as a 109 source of insight into the properties that data planes need to meet. 110 The IETF is currently working on potentially new encapsulations in 111 NVO3 and SFC and considering working on BIER. In addition there is 112 work on tunneling in the INT area. 114 This is a short term design team chartered to collect and construct 115 useful advice to parties working on new or modified data plane 116 behaviors that include additional encapsulations. The goal is for 117 the group to document useful advice gathered from interacting with 118 ongoing efforts. An Internet Draft will be produced for IETF92 to 119 capture that advice, which will be discussed in RTGWG. 121 Data plane encapsulations face a set of common issues such as: 122 o How to provide entropy for ECMP 123 o Issues around packet size and fragmentation/reassembly 124 o OAM - what support is needed in an encapsulation format? 125 o Security and privacy. 126 o QoS 127 o Congestion Considerations 128 o IPv6 header protection (zero UDP checksum over IPv6 issue) 129 o Extensibility - e.g., for evolving OAM, security, and/or 130 congestion control 131 o Layering of multiple encapsulations e.g., SFC over NVO3 over BIER 132 The design team will provide advice on those issues. The intention 133 is that even where we have different encapsulations for different 134 purposes carrying different information, each such encapsulation 135 doesn't have to reinvent the wheel for the above common issues. 137 The design team will look across the routing area in particular at 138 SFC, NVO3 and BIER. It will not be involved in comparing or 139 analyzing any particular encapsulation formats proposed in those WGs 140 and BoFs but instead focus on common advice. 142 2. Overview 144 The references provide background information on NVO3, SFC, and BIER. 145 In particular, NVO3 is introduced in [RFC7364], [RFC7365], and 146 [I-D.ietf-nvo3-arch]. SFC is introduced in 147 [I-D.ietf-sfc-architecture] and [I-D.ietf-sfc-problem-statement]. 148 Finally, the information on BIER is in 149 [I-D.shepherd-bier-problem-statement], 151 [I-D.wijnands-bier-architecture], and 152 [I-D.wijnands-mpls-bier-encapsulation]. We assume the reader has 153 some basic familiarity with those proposed encapsulations. The 154 Related Work section points at some prior work that relates to the 155 encapsulation considerations in this document. 157 Encapsulation protocols typically have some unique information that 158 they need to carry. In some cases that information might be modified 159 along the path and in other cases it is constant. The in-flight 160 modifications has impacts on what it means to provide security for 161 the encapsulation headers. 162 o NVO3 carries a VNI Identifier edge to edge which is not modified. 163 There has been OAM discussions in the WG and it isn't clear 164 whether some of the OAM information might be modified in flight. 165 o SFC carries Service Function Path identification and service meta- 166 data. The meta-data might be modified as the packets follow the 167 service path. SFC talks of some loop avoidance mechanism which is 168 likely to result in modifications for for each hop in the service 169 chain even if the meta-data is unmodified. 170 o BIER carries a bitmap of egress ports to which a packet should be 171 delivered, and as the packet is forwarded down different paths 172 different bits are cleared in that bitmap. 174 Even if information isn't modified in flight there might be devices 175 that wish to inspect that information. For instance, one can 176 envision future NVO3 security devices which filter based on the 177 virtual network identifier. 179 The need for extensibility is different across the protocols 180 o NVO3 might need some extensions for OAM and security. 181 o SFC consists of Service Function Path identification plus carrying 182 service meta-data along a path, and different services might need 183 different types and amount of meta-data. 184 o BIER might need variable number of bits in their bitmaps, or other 185 future schemes to scale up to larger network. 186 The extensibility needs and constraints might be different when 187 considering hardware vs. software implementations of the 188 encapsulation headers. NIC hardware might have different constraints 189 than switch hardware. 191 As the IETF designs these encapsulations the different WGs solve the 192 issues for their own encapsulation. But there are likely to be 193 future cases when the different encapsulations are combined in the 194 same header. For instance, NVO3 might be a "transport" used to carry 195 SFC between the different hops in the service chain. 197 Most of the issues discussed in this document are not new. The IETF 198 and industry as specified and deployed many different encapsulation 199 or tunneling protocols over time, ranging from simple IP-in-IP and 200 GRE encapsulation, IPsec, pseudo-wires, session-based approached like 201 L2TP, and the use of MPLS control and data planes. IEEE 802 has also 202 defined layered encapsulation for Provider Backbone Bridges (PBB) and 203 IEEE 802.1Qbp (ECMP). This document tries to leverage what we 204 collectively have learned from that experience and summarize what 205 would be relevant for new encapsulations like NVO3, SFC, and BIER. 207 3. Common Issues 209 [This section is mostly a repeat of the charter but with a few 210 modifications and additions.] 212 Any new encapsulation protocol would need to address a large set of 213 issues that are not central to the new information that this protocol 214 intends to carry. The common issues explored in this document are: 215 o How to provide entropy for Equal Cost MultiPath (ECMP) routing 216 o Issues around packet size and fragmentation/reassembly 217 o Next header indication - each encapsulation might be able to carry 218 different payloads 219 o OAM - what support is needed in an encapsulation format? 220 o Security and privacy 221 o QoS 222 o Congestion Considerations 223 o Header protection 224 o Extensibility - e.g., for evolving OAM, security, and/or 225 congestion control 226 o Layering of multiple encapsulations e.g., SFC over NVO3 over BIER 227 o Importance of being friendly to hardware and software 228 implementations 230 The degree to which these common issues apply to a particular 231 encapsulation can differ based on the intended purpose of the 232 encapsulation. But it is useful to understand all of them before 233 determining which ones apply. 235 4. Scope 237 It is important to keep in mind what we are trying to cover and not 238 cover in this document and effort. This is 239 o A look across the three new encapsulations, while taking lots of 240 previous work into account 241 o Focus on the class of encapsulations that would run over IP/UDP. 242 That was done to avoid being distracted by the data-plane and 243 control-plane interaction, which is more significant for protocols 244 that are designed to run over "transports" that maintain session 245 or path state. 246 o We later expanded the scope somewhat to consider how the 247 encapsulations would play with MPLS "transport", which is 248 important because SFC and BIER seem to target being independent of 249 the underlying "transport" 251 However, this document and effort is NOT intended to: 252 o Design some new encapsulation header to rule them all 253 o Design yet another new NVO3 encapsulation header 254 o Try to select the best encapsulation header 255 o Evaluate any existing and proposed encapsulations 257 While the origin and focus of this document is the routing area and 258 in particular NVO3, SFC, and BIER, the considerations apply to other 259 encapsulations that are being defined in the IETF and elsewhere. 260 There seems to be an increase in the number of encapsulations being 261 defined to run over UDP, where there might already exist an 262 encapsulation over IP or Ethernet. Feedback on how these 263 considerations apply in those contexts is welcome. 265 5. Assumptions 267 The design center for the new encapsulations is a well-managed 268 network. That network can be a datacenter network (plus datacenter 269 interconnect) or a service provider network. Based on the existing 270 and proposed encapsulations in those environment it is reasonable to 271 make these assumptions: 272 o The MTU is carefully managed and configured. Hence an 273 encapsulation protocol can make the packets bigger without 274 resulting in a requirement for fragmentation and reassembly 275 between ingress and egress. (However, it might be useful to 276 detecting MTU misconfigurations.) 277 o In general an encapsulation needs some approach for congestion 278 management. But the assumptions are different than for arbitrary 279 Internet paths in that the underlay might be well-provisioned and 280 better policed at the edge, and due to multi-tenancy, the 281 congestion control in the endpoints might be even less trusted 282 than on the Internet at large. 284 The goal is to implement these encapsulations in hardware and 285 software hence we can't assume that the needs of either 286 implementation approach can trump the needs of the other. In 287 particular, around extensibility the needs and constraints might be 288 quite different. 290 6. Terminology 292 The capitalized keyword MUST is used as defined in 293 http://en.wikipedia.org/wiki/Julmust 295 TBD: Refer to existing documents for at least NVO3 and SFC 296 terminology. We use at least the VNI ID in this document. 298 7. Entropy 300 In many cases the encapsulation format needs to enable ECMP in 301 unmodified routers. Those routers might use different fields in TCP/ 302 UDP packets to do ECMP without a risk of reordering a flow. Note 303 that the same entropy might also be used at layer 2 e.g. for Link 304 Aggregation (LAG). 306 The common way to do ECMP-enabled encapsulation over IP today is to 307 add a UDP header and to use UDP with the UDP source port carrying 308 entropy from the inner/original packet headers as in LISP [RFC6830]. 309 The total entropy consists of 14 bits in the UDP source port (using 310 the ephemeral port range) plus the outer IP addresses which seems to 311 be sufficient for entropy; using outer IPv6 headers would give the 312 option for more entropy should it be needed in the future. 314 In some environments it might be fine to use all 16 bits of the port 315 range. However, middleboxes might make assumptions about the system 316 ports or user ports. But they should not make any assumptions about 317 the ports in the Dynamic and/or Private Port range, which have the 318 two MSBs set to 11b. 320 The UDP source port might change over the lifetime of an encapsulated 321 flow, for instance for DoS mitigation or re-balancing load across 322 ECMP. Such changes need to consider reordering if there are packets 323 in flight for the flow. 325 There is some interaction between entropy and OAM and extensibility 326 mechanism. It is desirable to be able to send OAM packets to follow 327 the same path as network packets. Hence OAM packets should use the 328 same entropy mechanism as data packets. While routers might use 329 information in addition the entropy field and outer IP header, they 330 can not use arbitrary parts of the encapsulation header since that 331 might result in OAM frames taking a different path. Likewise if 332 routers look past the encapsulation header they need to be aware of 333 the extensibility mechanism(s) in the encapsulation format to be able 334 to find the inner headers in the presence of extensions; OAM frames 335 might use some extensions e.g. for timestamps. 337 Architecturally the entropy and the next header field are really part 338 of enclosing delivery header. UDP with entropy goes hand-in-hand 339 with the outer IP header. Thus the UDP entropy is present for the 340 underlay IP routers the same way that an MPLS entropy label is 341 present for LSRs. The entropy above is all about providing entropy 342 for the outer delivery of the encapsulated packets. 344 It has been suggested that when IPv6 is used it would not be 345 necessary to add a UDP header for entropy, since the IPv6 flow label 346 can be used for entropy. (This assumes that there is an IP protocol 347 number for the encapsulation in addition to a UDP destination port 348 number since UDP would be used with IPv4 underlay. And any use of 349 UDP checksums would need to be replaced by an encaps-specific 350 checksum or secure hash.) While such an approach would save 8 bytes 351 of headers when the underlay is IPv6, it does assume that the 352 underlay routers use the flow label for ECMP, and it also would make 353 the IPv6 approach different than the IPv4 approach. Currently the 354 leaning is towards recommending using the UDP encapsulation for both 355 IPv4 and IPv6 underlay. The IPv6 flow label can be used for 356 additional entropy if need be. There is more detailed discussion for 357 using the IPv6 flow label for tunnels in [RFC6438]. 359 Note that in the proposed BIER encapsulation 360 [I-D.wijnands-mpls-bier-encapsulation], there is an an 8-bit field 361 which specifies an entropy value that can be used for load balancing 362 purposes. This entropy is for the BIER forwarding decisions, which 363 is independent of any outer delivery ECMP between BIER routers. Thus 364 it is not part of the delivery ECMP discussed in this section. 365 [Note: For any given bit in BIER (that identifies an exit from the 366 BIER domain) there might be multiple immediate next hops. The 367 BIER entropy field is used to select that next hop as part of BIER 368 processing. The BIER forwarding process may do equal cost load 369 balancing, but the load balancing procedure MUST choose the same 370 path for any two packets that have the same entropy value.] 372 In summary: 373 o The entropy is associated with the transport, that is an outer IP 374 header or MPLS. 375 o In the case of IP transport use 14 or 16 bits of UDP source port, 376 plus outer IPv6 flowid for entropy. 378 8. Next-protocol indication 380 Next-protocol indications appear in three different contexts for 381 encapsulations. 383 Firstly, the transport delivery mechanism for the encapsulations we 384 discuss in this document need some way to indicate which 385 encapsulation header (or other payload) comes next in the packet. 386 Some encapsulations might be identified by a UDP port; others might 387 be identified by an Ethernet type or IP protocol number. Which 388 approach is used is a function of the preceding header the same way 389 as IPv4 is identified by both an Ethernet type and an IP protocol 390 number (for IP-in-IP). In some cases the header type is implicit in 391 some session (L2TP) or path (MPLS) setup. But this is largely beyond 392 the control of the encapsulation protocol. For instance, if there is 393 a requirement to carry the encapsulation after an Ethernet header, 394 then an Ethernet type is needed. If required to be carried after an 395 IP/UDP header, then a UDP port number is needed. For UDP port 396 numbers there are considerations for port number conservation 397 described in [I-D.ietf-tsvwg-port-use]. 399 It is worth mentioning that in the MPLS case of no implicit protocol 400 type many forwarding devices peek at the first nibble of the payload 401 to determine whether to apply IPv4 or IPv6 L3/L4 hashes for load 402 balancing [RFC7325]. That behavior places some constraints on other 403 payloads carried over MPLS and some protocol define an initial 404 control word in the payload with a value of zero in its first nibble 405 [RFC4385] to avoid confusion with IPv4 and IPv6 payload headers. 407 Secondly, the encapsulation needs to indicate the type of its 408 payload, which is in scope for the design of the encapsulation. We 409 have existing protocols which use Ethernet types (such as GRE). Here 410 each encapsulation header can potentially makes its own choices 411 between: 412 o Use the Ethernet type space - makes it easy to carry existing L2 413 and L3 protocols including IPv4, IPv6, and Ethernet. 414 Disadvantages are that it is a 16 bit number and we probably need 415 far less than 100 values, and the number space is controlled by 416 the IEEE 802 RAC with its own allocation policies. 417 o Use the IP protocol number space - makes it easy to carry e.g., 418 ESP in addition to IP and Ethernet but brings in all existing 419 protocol numbers many of which would never be used directly on top 420 of the encapsulation protocol. IANA managed eight bit values, 421 presumably more difficult to get an assigned number than to get a 422 transport port assignment. 423 o Define their own next-protocol number space, which can use fewer 424 bits than an Ethernet type and give more flexibility, but at the 425 cost of administering that numbering space (presumably by the 426 IANA). 428 Thirdly, if the IETF ends up defining multiple encapsulations at 429 about the same time, and there is some chance that multiple such 430 encapsulations can be combined in the same packet, there is a 431 question whether it makes sense to use a common approach and 432 numbering space for the encapsulation across the different protocols. 433 A common approach might not be beneficial as long as there is only 434 one way to indicate e.g., SFC inside NVO3. 436 Many Internet protocols use fixed values (typically managed by the 437 IANA function) for their next-protocol field. That facilitates 438 interpretation of packets by middleboxes and e.g., for debugging 439 purposes, but might make the protocol evolution inflexible. Our 440 collective experience with MPLS shows an alternative where the label 441 can be viewed as an index to a table containing processing 442 instructions and the table content can be managed in different ways. 443 Encapsulations might want to consider the tradeoffs between such more 444 flexible versus more fixed approaches. 446 In summary: 447 o Would it be useful for the IETF come up with a common scheme for 448 encapsulation protocols? If not each encapsulation can define its 449 own scheme. 451 9. MTU and Fragmentation 453 A common approach today is to assume that the underlay have 454 sufficient MTU to carry the encapsulated packets without any 455 fragmentation and reassembly at the tunnel endpoints. That is 456 sufficient when the operator of the ingress and egress have full 457 control of the paths between those endpoints. And it makes for 458 simpler (hardware) implementations if fragmentation and reassembly 459 can be avoided. 461 However, even under that assumption it would be beneficial to be able 462 to detect when there is some misconfiguration causing packets to be 463 dropped due to MTU issues. One way to do this is to have the 464 encapsulator set the don't-fragment (DF) flag in the outer IPv4 465 header and receive and log any received ICMP "packet too big" (PTB) 466 errors. Note that no flag needs to be set in an outer IPv6 header 467 [RFC2460]. 469 Encapsulations could also define an optional tunnel fragmentation and 470 reassembly mechanism which would be useful in the case when the 471 operator doesn't have full control of the path, or when the protocol 472 gets deployed outside of its original intended context. Such a 473 mechanism would be required if the underlay might have a path MTU 474 which makes it impossible to carry at least 1518 bytes (if offering 475 Ethernet service), or at least 1280 (if offering IPv6 service). The 476 use of such a protocol mechanism could be triggered by receiving a 477 PTB. But such a mechanism might not be implemented by all 478 encapsulators and decapsulators. [Aerolink is one example of such a 479 protocol.] 481 Depending on the payload carried by the encapsulation there are some 482 additional possibilities: 483 o If payload is IPv4/6 then the underlay path MTU could be used to 484 report end-to-end path MTU. 485 o If the payload service is Ethernet/L2, then there is no such per 486 destination reporting mechanism. However, there is a LLDP TLV for 487 reporting max frame size; might be useful to report minimum to end 488 stations, but unmodified end stations would do nothing with that 489 TLV since they assume that the MTU is at least 1518. 491 In summary: 492 o In some deployments an encapsulation can assume well-managed MTU 493 hence no need for fragmentation and reassembly related to the 494 encapsulation. 495 o Even so, it makes sense for ingress to track any ICMP packet too 496 big addressed to ingress to be able to log any MTU 497 misconfigurations. 498 o Should an encapsulation protocol be deployed outside of the 499 original context it might very well need support for fragmentation 500 and reassembly. 502 10. OAM 504 The OAM area is seeing active development in the IETF with 505 discussions (at least) in NVO3 and SFC working groups, plus the new 506 LIME WG looking at architecture and YANG models. 508 The design team has take a narrow view of OAM to explore the 509 potential OAM implications on the encapsulation format. 511 In terms of what we have heard from the various working groups there 512 seem to be needs to: 513 o Be able to send out-of-band OAM messages - that potentially should 514 follow the same path through the network as some flow of data 515 packets. 516 * Such OAM messages should not accidentally be decapsulated and 517 forwarded to the end stations. 518 o Be able to add OAM information to data packets that are 519 encapsulated. Discussions have been around: 520 * Using a bit in the OAM to synchronize sampling of counters 521 between the encapsulator and decapsulator. 522 * Optional timestamps, sequence numbers, etc for more detailed 523 measurements between encapsulator and decapsulator. 525 o Usable for both proactive monitoring (akin to BFD) and reactive 526 checks (akin to traceroute to pin-point a failure) 528 To ensure that the OAM messages can follow the same path the OAM 529 messages need to get the same ECMP (and LAG hashing) results as a 530 given data flow. An encapsulator can choose between one of: 531 o Limit ECMP hashing to not look past the UDP header i.e. the 532 entropy needs to be in the source/destination IP and UDP ports 533 o Make OAM packets look the same as data packets i.e. the initial 534 part of the OAM payload has the inner Ethernet, IP, TCP/UDP 535 headers as a payload. (This approach was taken in TRILL out of 536 necessity since there is no UDP header.) Any OAM bit in the 537 encapsulation header must in any case be excluded from the 538 entropy. 540 There can be several ways to prevent OAM packets from accidentally 541 being forwarded to the end station using: 542 o A bit in the frame (as in TRILL) indicating OAM 543 o A next-protocol indication with a designated value for "none" or 544 "oam". 545 This assumes that the bit or next protocol, respectively, would not 546 affect entropy/ECMP in the underlay. However, the next-protocol 547 field might be used to provide differentiated treatment of packets 548 based on their payload; for instance a TCP vs. IPsec ESP payload 549 might be handled differently. Based on that observation it might be 550 undesirable to overload the next protocol with the OAM drop behavior, 551 resulting in a preference for having a bit to indicate that the 552 packet should be forwarded to the end station after decapsulation. 554 There has been suggestions that one (or more) marker bits in the 555 encaps header would be useful in order to delineate measurement 556 epochs on the encapsulator and decapsulator and use that to compare 557 counters to determine packet loss. 559 A result of the above is that OAM is likely to evolve and needs some 560 degree of extensibility from the encapsulation format; a bit or two 561 plus the ability to define additional larger extensions. 563 An open question is how to handle error messages or other reports 564 relating to OAM. One can think if such reporting as being associated 565 with the encapsulation the same way ICMP is associated with IP. 566 Would it make sense for the IETF to develop a common Encapsulation 567 Error Reporting Protocol as part of OAM, which can be used for 568 different encapsulations? And if so, what are the technical 569 challenges. For instance, how to avoid it being filtered as ICMP 570 often is? 572 A potential additional consideration for OAM is the possible future 573 existence of gateways that "stitch" together different dataplane 574 encapsulations and might want to carry OAM end-to-end across the 575 different encapsulations. 577 In summary: 578 o It makes sense to reserve a bit for "drop after decapsulation" for 579 OAM out-of-band. 580 o An encapsulation needs sufficient extensibility for OAM (such as 581 bits, timestamps, sequence numbers). That might be motivated by 582 in-band OAM but it would make sense to leverage the same 583 extensions for out-of band OAM. 584 o OAM places some constraints on use of entropy in forwarding 585 devices. 586 o Should IETF look into error reporting that is independent of the 587 specific encapsulation? 589 11. Security Considerations 591 Different encapsulation use cases will have different requirements 592 around security. For instance, when encapsulation is used to build 593 overlay networks for network virtualization, isolation between 594 virtual networks may be paramount. BIER support of multicast may 595 entail different security requirements than encapsulation for 596 unicast. 598 In real deployment, the security of the underlying network may be 599 considered for determining the level of security needed in the 600 encapsulation layer. However for the purposes of this discussion, we 601 assume that network security is out of scope and that the underlying 602 network does not itself provide adequate or as least uniform security 603 mechanisms for encapsulation. 605 There are at least three considerations for security: 606 o Anti-spoofing/virtual network isolation 607 o Interaction with packet level security such as IPsec or DTLS 608 o Privacy (e.g., VNI ID confidentially for NVO3) 610 This section uses a VNI ID in NVO3 as an example. A SFC or BIER 611 encapsulation is likely to have fields with similar security and 612 privacy requirements. 614 11.1. Encapsulation-specific considerations 616 Some of these considerations appear for a new encapsulation, and 617 others are more specific to network virtualization in datacenters. 619 o New attack vectors: 620 * DDOS on specific queued/paths by attempting to reproduce the 621 5-tuple hash for targeted connections. 622 * Entropy in outer 5-tuple may be too little or predictable. 623 * Leakage of identifying information in the encapsulation header 624 for an encrypted payload. 625 * Vulnerabilities of using global values in fields like VNI ID. 626 o Trusted versus untrusted tenants in network virtualization: 627 * The criticality of virtual network isolation depends on whether 628 tenants are trusted or untrusted. In the most extreme cases, 629 tenants might not only be untrusted but may be considered 630 hostile. 631 * For a trusted set of users (e.g. a private cloud) it may be 632 sufficient to have just a virtual network identifier to provide 633 isolation. Packets inadvertently crossing virtual networks 634 should be dropped similar to a TCP packet with a corrupted port 635 being received on the wrong connection. 636 * In the presence of untrusted users (e.g. a public cloud) the 637 virtual network identifier must be adequately protected against 638 corruption and verified for integrity. This case may warrant 639 keyed integrity. 640 o Different forms of isolation: 641 * Isolation could be blocking all traffic between tenants (or 642 except as allowed by some firewall) 643 * Could also be about performance isolation i.e. one tenant can 644 overload the network in a way that affects other tenants 645 * Physical isolation of traffic for different tenants in network 646 may be required, as well as required restrictions that tenants 647 may have on where their packets may be routed. 648 o New attack vectors from untrusted tenants: 649 * Third party VMs with untrusted tenants allows internally borne 650 attacks within data centers 651 * Hostile VMs inside the system may exist (e.g. public cloud) 652 * Internally launched DDOS 653 * Passive snooping for mis-delivered packets 654 * Mitigate damage and detection in event that a VM is able to 655 circumvent isolation mechanisms 656 o Tenant-provider relationship: 657 * Tenant might not trust provider, hypervisors, network 658 * Provider likely will need to provide SLA or a least a statement 659 on security 660 * Tenant may implement their own additional layers of security 661 * Regulation and certification considerations 662 o Trend towards tighter security: 663 * Tenants' data in network increases in volume and value, attacks 664 become more sophisticated 666 * Large DCs already encrypt everything on disk 667 * DCs likely to encrypt inter-DC traffic at this point, use TLS 668 to Internet. 669 * Encryption within DC is becoming more commonplace, becomes 670 ubiquitous when cost is low enough. 671 * Cost/performance considerations. Cost of support for strong 672 security has made strong network security in DCs prohibitive. 673 * Are there lessons from MacSec? 675 11.2. Virtual network isolation 677 The first requirement is isolation between virtual networks. Packets 678 sent in one virtual network should never be illegitimately received 679 by a node in another virtual network. Isolation should be protected 680 in the presence of malicious attacks or inadvertent packet 681 corruption. 683 The second requirement is sender authentication. Sender identity is 684 authenticated to prevent anti-spoofing. Even if an attacker has 685 access to the packets in the network, they cannot send packets into a 686 virtual network. This may have two possibilities: 687 o Pairwise sender authentication. Any two communicating hosts 688 negotiate a shared key. 689 o Group authentication. A group of hosts share a key (this may be 690 more appropriate for multicast of encapsulation). 692 Possible security solutions: 693 o Security cookie: This is similar to L2TP cookie mechanism 694 [RFC3931]. A shared plain text cookie is shared between 695 encapsulator and decapsulator. A receiver validates a packet by 696 evaluating if the cookie is correct for the virtual network and 697 address of a sender. Validation function is F(cookie, VNI ID, 698 source address). If cookie matches, accept packet, else drop. 699 Since cookie is plain text this method does not protect against an 700 eavesdropping. Cookies are set and may be rotated out of band. 701 o Secure hash: This is a stronger mechanism than simple cookies that 702 borrows from IPsec and PPP authentication methods. In this model 703 security field contains a secure hash of some fields in the packet 704 using a shared key. Hash function may be something like H(key, 705 VNI ID, address, salt). The salt ensures the hash is not the same 706 for every packet, and if it includes a sequence number may also 707 protect against replay attacks. 709 In any use of a shared key, periodic re-keying should be allowed. 710 This could include use of techniques like generation numbers, key 711 windows, etc. See [I-D.farrelll-mpls-opportunistic-encrypt] for an 712 example application. 714 We might see firewalls that are aware of the encapsulation and can 715 provide some defense in depth combined with the above example anti- 716 spoofing approaches. An example would be an NVO3-aware firewall 717 being able to check the VNI ID. 719 Separately and in addition to such filtering, there might be a desire 720 to completely block an encapsulation protocol at certain places in 721 the network, e.g., at the edge of a datacenter. Using a fixed 722 standard UDP destination port number for each encapsulation protocol 723 would facilitate such blocking. 725 11.3. Packet level security 727 An encapsulated packet may itself be encapsulated in IPsec (e.g. 728 ESP). This should be straightforward and in fact is what would 729 happen today in security gateways. In this case, there is no special 730 consideration for the fact that packet is encapsulated, however since 731 the encapsulation layer headers are included (part of encrypted data 732 for instance) we lose visibility in the network of the encapsulation. 734 The more interesting case is when security is applied to the 735 encapsulation payload. This will keep the encapsulation headers in 736 the outer header visible to the network (for instance in nvo3 we may 737 way to firewall based on VNI ID even if the payload is encrypted). 738 One possibility is to apply DTLS to the encapsulation payload. In 739 this model the protocol stack may be something like IP|UDP|Encap| 740 DTLS|encrypted_payload. The encapsulation and security should be 741 done together at an encapsulator and resolved at the decapsulator. 742 Since the encapsulation header is outside of the security coverage, 743 this may itself require security (like described above). 745 In both of the above the security associations (SAs) may be between 746 physical hosts, so for instance in nvo3 we can have packets of 747 different virtual networks using the same SA-- this should not be an 748 issue since it is the VNI ID that ensures isolation (which needs to 749 be secured also). 751 11.4. In summary: 753 o Encapsulations need extensibility mechanisms to be able to add 754 security features like cookies and secure hashes protecting the 755 encapsulation header. 756 o NVO3 probably has specific higher requirements relating to 757 isolation for network virtualization, which is in scope for the 758 NVO3 WG. 759 o Our collective IETF experience is that successful protocols get 760 deployed outside of the original intended context, hence the 761 initial assumptions about the threat model might become invalid. 763 That needs to be considered in the standardization of new 764 encapsulations. 766 12. QoS 768 In the Internet architecture we support QoS using the Differentiated 769 Services Code Points (DSCP) in the formerly named Type-of-Service 770 field in the IPv4 header, and in the Traffic-Class field in the IPv6 771 header. The ToS and TC fields also contain the two ECN bits, which 772 are discussed in Section 13. 774 We have existing specifications how to process those bits. See 775 [RFC2983] for diffserv handling, which specifies how the received 776 DSCP value is used to set the DSCP value in an outer IP header when 777 encapsulating. (There are also existing specifications how DSCP can 778 be mapped to layer2 priorities.) 780 Those specifications apply whether or not there is some intervening 781 headers (e.g., for NVO3 or SFC) between the inner and outer IP 782 headers. Thus the encapsulation considerations in this area are 783 mainly about applying the framework in [RFC2983]. 785 Note that the DSCP and ECN bits are not the only part of an inner 786 packet that might potentially affect the outer packet. For example, 787 [RFC2473] specifies handling of inner IPv6 hop-by-hop options that 788 effectively result in copying some options to the outer header. It 789 is simpler to not have future encapsulations depend on such copying 790 behavior. 792 There are some other considerations specific to doing OAM for 793 encapsulations. If OAM messages are used to measure latency, it 794 would make sense to treat them the same as data payloads. Thus they 795 need to have the same outer DSCP value as the data packets which they 796 wish to measure. 798 Due to OAM there are constraints on middleboxes in general. If 799 middleboxes inspect the packet past the outer IP+UDP and 800 encapsulation header and look for inner IP and TCP/UDP headers, that 801 might violate the assumption that OAM packets will be handled the 802 same as regular data packets. That issue is broader than just QoS - 803 applies to firewall filters etc. 805 In summary: 806 o Leverage the existing approach in [RFC2983] for DSCP handling. 808 13. Congestion Considerations 810 Additional encapsulation headers does not introduce anything new for 811 Explicit Congestion Notification. It is just like IP-in-IP and IPsec 812 tunnels which is specified in [RFC6040] in terms of how the ECN bits 813 in the inner and outer header are handled when encapsulating and 814 decapsulating packets. Thus new encapsulations can more or less 815 include that by reference. 817 There are additional considerations around carrying non-congestion 818 controlled traffic. These details have been worked out in 819 [I-D.ietf-mpls-in-udp]. As specified in [RFC5405]: "IP-based traffic 820 is generally assumed to be congestion-controlled, i.e., it is assumed 821 that the transport protocols generating IP-based traffic at the 822 sender already employ mechanisms that are sufficient to address 823 congestion on the path. Consequently, a tunnel carrying IP-based 824 traffic should already interact appropriately with other traffic 825 sharing the path, and specific congestion control mechanisms for the 826 tunnel are not necessary". Those considerations are being captured 827 in [I-D.ietf-tsvwg-rfc5405bis]. 829 For this reason, where an encapsulation method is used to carry IP 830 traffic that is known to be congestion controlled, the UDP tunnels 831 does not create an additional need for congestion control. Internet 832 IP traffic is generally assumed to be congestion-controlled. 833 Similarly, in general Layer 3 VPNs are carrying IP traffic that is 834 similarly assumed to be congestion controlled. 836 However, some of the encapsulations (at least NVO3) will be able to 837 carry arbitrary Layer 2 packets to provide an L2 service, in which 838 case one can not assume that the traffic is congestion controlled. 840 One could handle this by adding some congestion control support to 841 the encapsulation header (one instance of which would end up looking 842 like DCCP). However, if the underlay is well-provisioned and managed 843 as opposed to being arbitrary Internet path, it might be sufficient 844 to have a slower reaction to congestion induced by that traffic. 845 There is work underway on a notion of "circuit breakers" for this 846 purpose. See See [I-D.ietf-tsvwg-circuit-breaker]. Encapsulations 847 which carry arbitrary Layer 2 packets want to consider that ongoing 848 work. 850 If the underlay is provisioned in such a way that it can guarantee 851 sufficient capacity for non-congestion controlled Layer 2 traffic, 852 then such circuit breakers might not be needed. 854 Two other considerations appear in the context of these 855 encapsulations as applied to overlay networks: 857 o Protect against malicious end stations 858 o Ensure fairness and/or measure resource usage across multiple 859 tenants 860 Those issues are really orthogonal to the encapsulation, in that they 861 are present even when no new encapsulation header is in use. 862 However, the application of the new encapsulations are likely to be 863 in environments where those issues are becoming more important. 864 Hence it makes sense to consider them. 866 One could make the encapsulation header be extensible to that it can 867 carry sufficient information to be able to measure resource usage, 868 delays, and congestion. The suggestions in the OAM section about a 869 single bit for counter synchronization, and optional timestamps 870 and/or sequence numbers, could be part of such an approach. There 871 might also be additional congestion-control extensions to be carried 872 in the encapsulation. Overall this results in a consideration to 873 support sufficient extensibility in the encapsulation to handle 874 potential future developments in this space. 876 Coarse measurements are likely to suffice, at least for circuit- 877 breaker-like purposes, see [I-D.wei-tsvwg-tunnel-congestion-feedback] 878 and [I-D.briscoe-conex-data-centre] for examples on active work in 879 this area via use of ECN. [RFC6040] Appendix C is also relevant. 880 The outer ECN bits seem sufficient (at least when everything uses 881 ECN) to do this course measurements. Needs some more study for the 882 case when there are also drops; might need to exchange counters 883 between ingress and egress to handle drops. 885 Circuit breakers are not sufficient to make a network with different 886 congestion control when the goal is to provide a predictable service 887 to different tenants. The fallback would be to rate limit different 888 traffic. 890 In summary: 891 o Leverage the existing approach in [RFC6040] for ECN handling. 892 o If the encapsulation can carry non-IP, hence non-congestion 893 controlled traffic, then leverage the approach in 894 [I-D.ietf-mpls-in-udp]. 895 o "Watch this space" for circuit breakers. 897 14. Header Protection 899 Many UDP based encapsulations such as VXLAN [RFC7348] either 900 discourage or explicitly disallow the use of UDP checksums. The 901 reason is that the UDP checksum covers the entire payload of the 902 packet and switching ASICs are typically optimized to look at only a 903 small set of headers as the packet passes through the switch. In 904 these case, computing a checksum over the packet is very expensive. 905 (Software endpoints and the NICs used with them generally do not have 906 the same issue as they need to look at the entire packet anyways.) 908 The lack a header checksum creates the possibility that bit errors 909 can be introduced into any information carried by the new headers. 910 Specifically, in the case of IPv6, the assumption is that a transport 911 layer checksum - UDP in this case - will protect the IP addresses 912 through the inclusion of a pseudo-header in the calculation. This is 913 different from IPv4 on which many of these encapsulation protocols 914 are initially deployed which contains its own header checksum. In 915 addition to IP addresses, the encapsulation header often contains its 916 own information which is used for addressing packets or other high 917 value network functions. Without a checksum, this information is 918 potentially vulnerable - an issue regardless of whether the packet is 919 carried over IPv4 or IPv6. 921 Several protocols cite [RFC6935] and [RFC6936] as an exemption to the 922 IPv6 checksum requirements. However, these are intended to be 923 tailored to a fairly narrow set of circumstances - primarily relying 924 on sparseness of the address space to detect invalid values and well 925 managed networks - and are not a one size fits all solution. In 926 these cases, an analysis should be performed of the intended 927 environment, including the probability of errors being introduced and 928 the use of ECC memory in routing equipment. 930 Conceptually, the ideal solution to this problem is a checksum that 931 covers only the newly added headers of interest. There is little 932 value in the portion of the UDP checksum that covers the encapsulated 933 packet because that would generally be protected by other checksums 934 and this is the expensive portion to compute. In fact, this solution 935 already exists in the form of UDP-Lite and UDP based encapsulations 936 could be easily ported to run on top of it. Unfortunately, the main 937 value in using UDP as part of the encapsulation header is that it is 938 recognized by already deployed equipment for the purposes of ECMP, 939 RSS, and middlebox operations. As UDP-Lite uses a different protocol 940 number than UDP and it is not widely implemented in middleboxes, this 941 value is lost. A possible solution is to incorporate the same 942 partial-checksum concept as UDP-Lite or other header checksum 943 protection into the encapsulation header and continue using UDP as 944 the outer protocol. One potential challenge with this approach is 945 the use of NAT or other form of translation on the outer header will 946 result in an invalid checksum as the translator will not know to 947 update the encapsulation header. 949 The method chosen to protect headers is often related to the security 950 needs of the encapsulation mechanism. On one hand, the impact of a 951 poorly protected header is not limited to only data corruption but 952 can also introduce a security vulnerability in the form of 953 misdirected packets to an unauthorized recipient. Conversely, high 954 security protocols that already include a secure hash over the 955 valuable portion of the header (such as by encrypting the entire IP 956 packet using IPsec, or some secure hash of the encap header) do not 957 require additional checksum protection as the hash provides stronger 958 assurance than a simple checksum. 960 If the sender has included a checksum, then the receiver should 961 verify that checksum or, if incapable, drop the packet. The 962 assumption is that configuration and/or control-plane capability 963 exchanges can be used when different receiver have different checksum 964 validation capabilities. 966 In summary: 967 o Encapsulations need extensibility to be able to add checksum/CRC 968 for the encapsulation header itself. 969 o When the encapsulation has a checksum/CRC, include the IPv6 970 pseudo-header in it. 971 o The checksum/CRC can potentially be avoided when cryptographic 972 protection is applied to the encapsulation. 974 15. Extensibility Considerations 976 Protocol extensibility is the concept that a networking protocol may 977 be extended to include new use cases or functionality that were not 978 part of the original protocol specification. Extensibility may be 979 used to add security, control, management, or performance features to 980 a protocol. A solution may allow private extensions for 981 customization or experimentation. 983 Extending a protocol often implies that a protocol header must carry 984 new information. There are two usual methods to accomplish this: 985 1. Define or redefine the meaning of existing fields in a protocol 986 header. 987 2. Add new (optional) fields to the protocol header. 988 It is also possible to create a new protocol version, but this is 989 more associated with defining a protocol than extending it (IPv6 990 being a successor to IPv4 is an example of protocol versioning). 992 In some cases it might be more appropriate to define a new inner 993 protocol which can carry the new functionality instead of extending 994 the outer protocol. Examples where this works well is in the IP/ 995 transport split, where the earlier architecture had a single NCP 996 [RFC0033] protocol which carried both the hop-by-hop semantics which 997 are now in IP, and the end-to-end semantics which are now in TCP. 998 Such a split is effective when different nodes need to act upon the 999 different information. Applying this for general protocol 1000 extensibility through nesting is not well understood, and does result 1001 in longer header chains. Furthermore, our experience with IPv6 1002 extension headers [RFC2460] in middleboxes indicates that the header 1003 chaining approach does not help with middlebox traversal. 1005 Many protocol definitions include some number of reserved fields or 1006 bits which can be used for future extension. VXLAN is an example of 1007 a protocol that includes reserved bits which are subsequently being 1008 allocated for new purposes. Another technique employed is to re- 1009 purpose existing header fields with new meanings. A classic example 1010 of this is the definition of DSCP code point which redefines the ToS 1011 field originally specified in IPv4. When a field is redefined, some 1012 mechanism may be needed to ensure that all interested parties agree 1013 on the meaning of the field. The techniques of defining meaning for 1014 reserved bits or redefining existing fields have the advantage that a 1015 protocol header can be kept a fixed length. The disadvantage is that 1016 the extensibility is limited. For instance, the number reserved bits 1017 in a fixed protocol header is limited. For standard protocols the 1018 decision to commit to a definition for a field can be wrenching since 1019 it is difficult to retract later. Also, it is difficult to predict a 1020 priori how many reserved fields or bits to put into a protocol header 1021 to satisfy the extensions create over the lifetime of the protocol. 1023 Extending a protocol header with new fields can be done in several 1024 ways. 1025 o TLVs are a very popular method used in such protocols as IP and 1026 TCP. Depending on the type field size and structure, TLVs can 1027 offer a virtually unlimited range of extensions. A disadvantage 1028 of TLVs is that processing them can be verbose, quite complicated, 1029 several validations must often be done for each TLV, and there is 1030 no deterministic ordering for a list of TLVs. TCP serves as an 1031 example of a protocol where TLVs have been successfully used (i.e. 1032 required for protocol operation). IP is an example of a protocol 1033 that allows TLVs but are rarely used in practice (router fast 1034 paths usually that assume no IP options). Note that TCP TLVs are 1035 implemented in software as well as (NIC) hardware handling various 1036 forms of TCP offload. Additional discussions about hardware 1037 implications for extensibility is captured in Section 18. 1038 o Extension headers are closely related to TLVs. These also carry 1039 type/value information, but instead of being a list of TLVs within 1040 a single protocol header, each one is in its own protocol header. 1041 IPv6 extension headers and SFC NSH are examples of this technique. 1042 Similar to TLVs these offer a wide range of extensibility, but 1043 have similarly complex processing. Another difference with TLVs 1044 is that each extension header is idempotent. This is beneficial 1045 in cases where a protocol implements a push/pop model for header 1046 elements like service chaining, but makes it more difficult group 1047 correlated information within one protocol header. 1048 o A particular form of extension headers are the tags used by IEEE 1049 802 protocols. Those are similar to e.g., IPv6 extension headers 1050 but with the key difference that each tag is a fixed length header 1051 where the length is implicit in the tag value. Thus as long as a 1052 receiver can be programmed with a tag value to length map, it can 1053 skip those new tags. 1054 o Flag-fields are a non-TLV like method of extending a protocol 1055 header. The basic idea is that the header contains a set of 1056 flags, where each set flags corresponds to optional field that is 1057 present in the header. GRE is an example of a protocol that 1058 employs this mechanism. The fields are present in the header in 1059 the order of the flags, and the length of each field is fixed. 1060 Flag-fields are simpler to process compared to TLVs, having fewer 1061 validations and the order of the optional fields is deterministic. 1062 A disadvantage is that range of possible extensions with flag- 1063 fields is smaller than TLVs. 1065 The requirements for receiving unknown or unimplemented extensible 1066 elements in an encapsulation protocol (flags, TLVs, optional fields) 1067 need to be specified. There are two parties to consider, middle 1068 boxes and terminal endpoints of encapsulation (at the decapsulator). 1070 A protocol may allow or expect nodes in a path to modify fields in an 1071 encapsulation (example use of this is BIER). In this case, the 1072 middleboxes should follow the same requirements as nodes terminating 1073 the encapsulation. In the case that middle boxes do not modify the 1074 encapsulation, we can assume that they may still inspect any fields 1075 of the encapsulation. Missing or unknown fields should be accepted 1076 per protocol specification, however it is permissible for a site to 1077 implement a local policy otherwise (e.g. a firewall may drop packets 1078 with unknown options). 1080 For handling unknown options at terminal nodes, there are two 1081 possibilities: drop packet or accept while ignoring the unknown 1082 options. Many Internet protocols specify that reserved flags must be 1083 set to zero on transmission and ignored on reception. L2TP is 1084 example data protocol that has such flags. GRE is a notable 1085 exception to this rule, reserved flag bits 1-5 cannot be ignored 1086 [RFC2890]. For TCP and IPv4, implementations must ignore optional 1087 TLVs with unknown type; however in IPv6 if a packet contains an 1088 unknown extension header (unrecognized next header type) the packet 1089 must be dropped with an ICMP error message returned. The IPv6 1090 options themselves (encoded inside the destinations options or hop- 1091 by-hop options extension header) have more flexibility. There are 1092 bits in the option code are used to instruct the receiver whether to 1093 ignore, silently drop, or drop and send error if the option is 1094 unknown. Some protocols define a "mandatory bit" that can is set 1095 with TLVs to indicate that an option must not be ignored. 1096 Conceptually, optional data elements can only be ignored if they are 1097 idempotent and do not alter how the rest of the packet is parsed or 1098 processed. 1100 Depending on what type of protocol evolution one can predict, it 1101 might make sense to have a way for a sender to express that the 1102 packet should be dropped by a terminal node which does not understand 1103 the new information. In other cases it would make sense to have the 1104 receiver silently ignore the new info. The former can be expressed 1105 by having a version field in the encapsulation, or a notion of 1106 "mandatory bit" as discussed above. 1108 A security mechanism which use some form secure hash over the 1109 encapsulation header would need to be able to know which extensions 1110 can be changed in flight. 1112 In summary: 1113 o Encapsulations need the ability to be extended to handle e.g., the 1114 OAM or security aspects discussed in this document. 1115 o Practical experience seems to tell us that extensibility 1116 mechanisms which are not in use on day one might result in 1117 immediate ossification by lack of implementation support. In some 1118 cases that has occurred in routers and in other cases in 1119 middleboxes. Hence devising ways where the extensibility 1120 mechanisms are in use seems important. 1122 16. Layering Considerations 1124 One can envision that SFC might use NVO3 as a delivery/transport 1125 mechanism. With more imagination that in turn might be delivered 1126 using BIER. Thus it is useful to think about what things look like 1127 when we have BIER+NVO3+SFC+payload. Also, if NVO3 is widely deployed 1128 there might be cases of NVO3 nesting where a customer uses NVO3 to 1129 provide network virtualization e.g., across departments. That 1130 customer uses a service provider which happens to use NVO3 to provide 1131 transport for their customers.Thus NVO3 in NVO3 might happen. 1133 A key question we set out to answer is what the packets might look 1134 like in such a case, and in particular whether we would end up with 1135 multiple UDP headers for entropy. 1137 Based on the discussion in the Entropy section, the entropy is 1138 associated with the outer delivery IP header. Thus if there are 1139 multiple IP headers there would be a UDP header for each one of the 1140 IP headers. But SFC does not require its own IP header. So a case 1141 of NVO3+SFC would be IP+UDP+NVO3+SFC. A nested NVO3 encapsulation 1142 would have independent IP+UDP headers. 1144 The layering also has some implications for middleboxes. 1145 o A device on the path between the ingress and egress is allowed to 1146 transparently inspect all layers of the protocol stack and drop or 1147 forward, but not transparently modify anything but the layer in 1148 which they operate. What this means is that an IP router is 1149 allowed modify the outer IP ttl and ECN bits, but not the 1150 encapsulation header or inner headers and payload. And a BIER 1151 router is allowed to modify the BIER header. 1152 o Alternatively such a device can become visible at a higher layer. 1153 E.g., a middlebox could a middlebox could first decapsulate, 1154 perform some function then encapsulate; which means it will 1155 generate a new encapsulation header. 1157 The design team asked itself some additional questions: 1158 o Would it make sense to have a common encapsulation base header 1159 (for OAM, security?, etc) and then followed by the specific 1160 information for NVO3, SFC, BIER? Given that there are separate 1161 proposals and the set of information needing to be carried 1162 differs, and the extensibility needs might be different, it would 1163 be difficult and not that useful to have a common base header. 1164 o With a base header in place, one could view the different 1165 functions (NVO3, SFC, and BIER) as different extensions to that 1166 base header resulting in encodings which are more space optimal by 1167 not repeating the same base header. The base header would only be 1168 repeated when there is an additional IP (and hence UDP) header. 1169 That could mean a single length field (to skip to get to the 1170 payload after all the encapsulation headers). That might be 1171 technically feasible, but it would create a lot of dependencies 1172 between different WGs making it harder to make progress. Compare 1173 with the potential savings in packet size. 1175 17. Service model 1177 The IP service is lossy and subject to reordering. In order to avoid 1178 a performance impact on transports like TCP the handling of packets 1179 is designed to avoid reordering packets that are in the same 1180 transport flow (which is typically identified by the 5-tuple). But 1181 across such flows the receiver can see different ordering for a given 1182 sender. That is the case for a unicast vs. a multicast flow from the 1183 same sender. 1185 There is a general tussle between the desire for high capacity 1186 utilization across a multipath network and the impact on packet 1187 ordering within the same flow (which results in lower transport 1188 protocol performance). That isn't affected by the introduction of an 1189 encapsulation. However, the encapsulation comes with some entropy, 1190 and there might be cases where folks want to change that in response 1191 to overload or failures. For instance, one might want to change UDP 1192 source port to try different ECMP route. Such changes can result in 1193 packet reordering within a flow, hence would need to be done 1194 infrequently and with care e.g., by identifying packet trains. 1196 There might be some applications/services which are not able to 1197 handle reordering across flows. The IETF has defined pseudo-wires 1198 [RFC3985] which provides the ability to ensure ordering (implemented 1199 using sequence numbers and/or timestamps). 1201 Architectural such services would make sense, but as a separate layer 1202 on top of an encapsulation protocol. They could be deployed between 1203 ingress and egress of a tunnel which uses some encaps. Potentially 1204 the tunnel control points at the ingress and egress could become a 1205 platform for fixing suboptimal behavior elsewhere in the network. 1206 That would clearly be undesirable in the general case. However, 1207 handling encapsulation of non-IP traffic hence non-congestion- 1208 controlled traffic is likely to be required, which implies some 1209 fairness and/or QoS policing on the ingress and egress devices. 1211 But the tunnels could potentially do more like increase reliability 1212 (retransmissions, FEC) or load spreading using e.g. MP-TCP between 1213 ingress and egress. 1215 18. Hardware Friendly 1217 Hosts, switches and routers often leverage capabilities in the 1218 hardware to accelerate packet encapsulation, decapsulation and 1219 forwarding. 1221 Some design considerations in encapsulation that leverage these 1222 hardware capabilities may result in more efficiently packet 1223 processing and higher overall protocol throughput. 1225 While "hardware friendliness" can be viewed as unnecessary 1226 considerations for a design, part of the motivation for considering 1227 this is ease of deployment; being able to leverage existing NIC and 1228 switch chips for at least a useful subset of the functionality that 1229 the new encapsulation provides. The other part is the ease of 1230 implementing new NICs and switch/router chips that support the 1231 encapsulation at ever increasing line rates. 1233 [disclaimer] There are many different types of hardware in any given 1234 network, each maybe better at some tasks while worse at others. We 1235 would still recommend protocol designers to examine the specific 1236 hardware that are likely to be used in their networks and make 1237 decisions on a case by case basis. 1239 Some considerations are: 1240 o Keep the encap header small. Switches and routers usually only 1241 read the first small number of bytes into the fast memory for 1242 quick processing and easy manipulation. The bulk of the packets 1243 are usually stored in slow memory. A big encap header may not fit 1244 and additional read from the slow memory will hurt the overall 1245 performance and throughput. 1246 o Put important information at the beginning of the encapsulation 1247 header. The reasoning is similar as explained in the previous 1248 point. If important information are located at the beginning of 1249 the encapsulation header, the packet may be processed with smaller 1250 number of bytes to be read into the fast memory and improve 1251 performance. 1252 o Avoid full packet checksums in the encapsulation if possible. 1253 Encapsulations should instead consider adding their own checksum 1254 which covers the encapsulation header and any IPv6 pseudo-header. 1255 The motivation is that most of the switch/router hardware make 1256 switching/forwarding decisions by reading and examining only the 1257 first certain number of bytes in the packet. Most of the body of 1258 the packet do not need to be processed normally. If we are 1259 concerned of preventing packet to be misdelivered due to memory 1260 errors, consider only perform header checksums. Note that NIC 1261 chips can typically already do full packet checksums for TCP/UDP, 1262 while adding a header checksum might require adding some hardware 1263 support. 1264 o Place important information at fixed offset in the encapsulation 1265 header. Packet processing hardware may be capable of parallel 1266 processing. If important information can be found at fixed 1267 offset, different part of the encapsulation header may be 1268 processed by different hardware units in parallel (for example 1269 multiple table lookups may be launched in parallel). It is easier 1270 for hardware to handle optional information when the information, 1271 if present, can be found in ideally one place, but in general, in 1272 as few places as possible. That facilitates parallel processing. 1273 TLV encoding with unconstrained order typically does not have that 1274 property. 1275 o Limit the number of header combinations. In many cases the 1276 hardware can explore different combinations of headers in 1277 parallel, however there is some added cost for this. 1279 18.1. Considerations for NIC offload 1281 This section provides guidelines to provide support of common 1282 offloads for encapsulation in Network Interface Cards (NICs). 1283 Offload mechanisms are techniques that are implemented separately 1284 from the normal protocol implementation of a host networking stack 1285 and are intended to optimize or speed up protocol processing. 1286 Hardware offload is performed within a NIC device on behalf of a 1287 host. 1289 There are three basic offload techniques of interest: 1290 o Receive multi queue 1291 o Checksum offload 1292 o Segmentation offload 1294 18.1.1. Receive multi-queue 1296 Contemporary NICs support multiple receive descriptor queues (multi- 1297 queue). Multi-queue enables load balancing of network processing for 1298 a NIC across multiple CPUs. On packet reception, a NIC must select 1299 the appropriate queue for host processing. Receive Side Scaling 1300 (RSS) is a common method which uses the flow hash for a packet to 1301 index an indirection table where each entry stores a queue number. 1303 UDP encapsulation, where the source port is used for entropy, should 1304 be compatible with multi-queue NICs that support five-tuple hash 1305 calculation for UDP/IP packets as input to RSS. The source port 1306 ensures classification of the encapsulated flow even in the case that 1307 the outer source and destination addresses are the same for all flows 1308 (e.g. all flows are going over a single tunnel). 1310 18.1.2. Checksum offload 1312 Many NICs provide capabilities to calculate standard ones complement 1313 payload checksum for packets in transmit or receive. When using 1314 encapsulation over UDP there are at least two checksums that may be 1315 of interest: the encapsulated packet's transport checksum, and the 1316 UDP checksum in the outer header. 1318 18.1.2.1. Transmit checksum offload 1320 NICs may provide a protocol agnostic method to offload transmit 1321 checksum (NETIF_F_HW_CSUM in Linux parlance) that can be used with 1322 UDP encapsulation. In this method the host provides checksum related 1323 parameters in a transmit descriptor for a packet. These parameters 1324 include the starting offset of data to checksum, the length of data 1325 to checksum, and the offset in the packet where the computed checksum 1326 is to be written. The host initializes the checksum field to pseudo 1327 header checksum. In the case of encapsulated packet, the checksum 1328 for an encapsulated transport layer packet, a TCP packet for 1329 instance, can be offloaded by setting the appropriate checksum 1330 parameters. 1332 NICs typically can offload only one transmit checksum per packet, so 1333 simultaneously offloading both an inner transport packet's checksum 1334 and the outer UDP checksum is likely not possible. In this case 1335 setting UDP checksum to zero (per above discussion) and offloading 1336 the inner transport packet checksum might be acceptable. 1338 There is a proposal in [I-D.herbert-remotecsumoffload] to leverage 1339 NIC checksum offload when an encapsulator is co-resident with a host. 1341 18.1.2.2. Receive checksum offload 1343 Protocol encapsulation is compatible with NICs that perform a 1344 protocol agnostic receive checksum (CHECKSUM_COMPLETE in Linux 1345 parlance). In this technique, a NIC computes a ones complement 1346 checksum over all (or some predefined portion) of a packet. The 1347 computed value is provided to the host stack in the packet's receive 1348 descriptor. The host driver can use this checksum to "patch up" and 1349 validate any inner packet transport checksum, as well as the outer 1350 UDP checksum if it is non-zero. 1352 Many legacy NICs don't provide checksum-complete but instead provide 1353 an indication that a checksum has been verified (CHECKSUM_UNNECESSARY 1354 in Linux). Usually, such validation is only done for simple TCP/IP 1355 or UDP/IP packets. If a NIC indicates that a UDP checksum is valid, 1356 the checksum-complete value for the UDP packet is the "not" of the 1357 pseudo header checksum. In this way, checksum-unnecessary can be 1358 converted to checksum-complete. So if the NIC provides checksum- 1359 unnecessary for the outer UDP header in an encapsulation, checksum 1360 conversion can be done so that the checksum-complete value is derived 1361 and can be used by the stack to validate an checksums in the 1362 encapsulated packet. 1364 18.1.3. Segmentation offload 1366 Segmentation offload refers to techniques that attempt to reduce CPU 1367 utilization on hosts by having the transport layers of the stack 1368 operate on large packets. In transmit segmentation offload, a 1369 transport layer creates large packets greater than MTU size (Maximum 1370 Transmission Unit). It is only at much lower point in the stack, or 1371 possibly the NIC, that these large packets are broken up into MTU 1372 sized packet for transmission on the wire. Similarly, in receive 1373 segmentation offload, small packets are coalesced into large, greater 1374 than MTU size packets at a point low in the stack receive path or 1375 possibly in a device. The effect of segmentation offload is that the 1376 number of packets that need to be processed in various layers of the 1377 stack is reduced, and hence CPU utilization is reduced. 1379 18.1.3.1. Transmit Segmentation Offload 1381 Transmit Segmentation Offload (TSO) is a NIC feature where a host 1382 provides a large (larger than MTU size) TCP packet to the NIC, which 1383 in turn splits the packet into separate segments and transmits each 1384 one. This is useful to reduce CPU load on the host. 1386 The process of TSO can be generalized as: 1387 o Split the TCP payload into segments which allow packets with size 1388 less than or equal to MTU. 1389 o For each created segment: 1390 1. Replicate the TCP header and all preceding headers of the 1391 original packet. 1392 2. Set payload length fields in any headers to reflect the length 1393 of the segment. 1394 3. Set TCP sequence number to correctly reflect the offset of the 1395 TCP data in the stream. 1396 4. Recompute and set any checksums that either cover the payload 1397 of the packet or cover header which was changed by setting a 1398 payload length. 1400 Following this general process, TSO can be extended to support TCP 1401 encapsulation UDP. For each segment the Ethernet, outer IP, UDP 1402 header, encapsulation header, inner IP header if tunneling, and TCP 1403 headers are replicated. Any packet length header fields need to be 1404 set properly (including the length in the outer UDP header), and 1405 checksums need to be set correctly (including the outer UDP checksum 1406 if being used). 1408 To facilitate TSO with encapsulation it is recommended that optional 1409 fields should not contain values that must be updated on a per 1410 segment basis-- for example an encapsulation header should not 1411 include checksums, lengths, or sequence numbers that refer to the 1412 payload. If the encapsulation header does not contain such fields 1413 then the TSO engine only needs to copy the bits in the encapsulation 1414 header when creating each segment and does not need to parse the 1415 encapsulation header. 1417 18.1.3.2. Large Receive Offload 1419 Large Receive Offload (LRO) is a NIC feature where packets of a TCP 1420 connection are reassembled, or coalesced, in the NIC and delivered to 1421 the host as one large packet. This feature can reduce CPU 1422 utilization in the host. 1424 LRO requires significant protocol awareness to be implemented 1425 correctly and is difficult to generalize. Packets in the same flow 1426 need to be unambiguously identified. In the presence of tunnels or 1427 network virtualization, this may require more than a five-tuple match 1428 (for instance packets for flows in two different virtual networks may 1429 have identical five-tuples). Additionally, a NIC needs to perform 1430 validation over packets that are being coalesced, and needs to 1431 fabricate a single meaningful header from all the coalesced packets. 1433 The conservative approach to supporting LRO for encapsulation would 1434 be to assign packets to the same flow only if they have identical 1435 five-tuple and were encapsulated the same way. That is the outer IP 1436 addresses, the outer UDP ports, encapsulated protocol, encapsulation 1437 headers, and inner five tuple are all identical. 1439 18.1.3.3. In summary: 1441 In summary, for NIC offload: 1442 o The considerations for using full UDP checksums are different for 1443 NIC offload than for implementations in forwarding devices like 1444 routers and switches. 1445 o Be judicious about encapsulations that change fields on a per- 1446 packet basis, since such behavior might make it hard to use TSO. 1448 19. Middlebox Considerations 1450 This document has touched upon middleboxes in different section. The 1451 reason for this is as encapsulations get widely deployed one would 1452 expect different forms of middleboxes might become aware of the 1453 encapsulation protocol just as middleboxes have been made aware of 1454 other protocols where there are business and deployment 1455 opportunities. Such middleboxes are likely to do more than just drop 1456 packets based on the UDP port number used by an encapsulation 1457 protocol. 1459 We note that various forms of encapsulation gateways that stitch one 1460 encapsulation protocol together with another form of protocol could 1461 have similar effects. 1463 An example of a middlebox that could see some use would be an NVO3- 1464 aware firewall that would filter on the VNI IDs to provide some 1465 defense in depth inside or across NVO3 datacenters. 1467 A question for the IETF is whether we should document what to do or 1468 what not to do in such middleboxes. This document touches on areas 1469 of OAM and ECMP as it relates to middleboxes and it might make sense 1470 to document how encapsulation-aware middleboxes should avoid 1471 unintended consequences in those (and perhaps other) areas. 1473 In summary: 1475 o We are likely to see middleboxes that at least parse the headers 1476 for successful new encapsulations. 1477 o Should the IETF document considerations for what not to do in such 1478 middleboxes? 1480 20. Related Work 1482 The IETF and industry has defined encapsulations for a long time, 1483 with examples like GRE [RFC2890], VXLAN [RFC7348], and NVGRE 1484 [I-D.sridharan-virtualization-nvgre] being able to carry arbitrary 1485 Ethernet payloads, and various forms of IP-in-IP and IPsec 1486 encapsulations that can carry IP packets. As part of NVO3 there has 1487 been additional proposals like Geneve [I-D.gross-geneve] and GUE 1488 [I-D.herbert-gue] which look at more extensibility. NSH 1489 [I-D.quinn-sfc-nsh] is an example of an encapsulation that tries to 1490 provide extensibility mechanisms which target both hardware and 1491 software implementations. 1493 There is also a large body of work around MPLS encapsulations 1494 [RFC3032]. The MPLS-in-UDP work [I-D.ietf-mpls-in-udp] and GRE over 1495 UDP [I-D.ietf-tsvwg-gre-in-udp-encap] have worked on some of the 1496 common issues around checksum and congestion control. MPLS also 1497 introduced a entropy label [RFC6790]. There is also a proposal for 1498 MPLS encryption [I-D.farrelll-mpls-opportunistic-encrypt]. 1500 The idea to use a UDP encapsulation with a UDP source port for 1501 entropy for the underlay routers' ECMP dates back to LISP [RFC6830]. 1503 The pseudo-wire work [RFC3985] is interesting in the notion of 1504 layering additional services/characteristics such as ordered delivery 1505 or timely deliver on top of an encapsulation. That layering approach 1506 might be useful for the new encapsulations as well. For instance, 1507 the control word [RFC4385]. There is also material on congestion 1508 control for pseudo-wires in [I-D.ietf-pwe3-congcons]. 1510 Both MPLS and L2TP [RFC3931] rely on some control or signaling to 1511 establish state (for the path/labels in the case of MPLS, and for the 1512 session in the case of L2TP). The NVO3, SFC, and BIER encapsulations 1513 will also have some separation between the data plane and control 1514 plane, but the type of separation appears to be different. 1516 IEEE 802.1 has defined encapsulations for L2 over L2, in the form of 1517 Provider backbone Bridging (PBB) [IEEE802.1Q-2014] and Equal Cost 1518 Multipath (ECMP) [IEEE802.1Q-2014]. The latter includes something 1519 very similar to the way the UDP source port is used as entropy: "The 1520 flow hash, carried in an F-TAG, serves to distinguish frames 1521 belonging to different flows and can be used in the forwarding 1522 process to distribute frames over equal cost paths" 1524 TRILL, which is also a L2 over L2 encapsulation, took a different 1525 approach to entropy but preserved the ability for OAM frames 1526 [RFC7174] to use the same entropy hence ECMP path as data frames. In 1527 [I-D.ietf-trill-oam-fm] there 96 bytes of headers for entropy in the 1528 OAM frames, followed by the actual OAM content. This ensures that 1529 any headers, which fit in those 96 bytes except the OAM bit in the 1530 TRILL header, can be used for ECMP hashing. 1532 As encapsulations evolve there might be a desire to fit multiple 1533 inner packets into one outer packet. The work in 1534 [I-D.saldana-tsvwg-simplemux] might be interesting for that purpose. 1536 21. Acknowledgements 1538 The authors acknowledge the comments from Alia Atlas, Fred Baker, 1539 David Black, Bob Briscoe, Stewart Bryant, Mike Cox, Andy Malis, Radia 1540 Perlman, Carlos Pignataro, Jamal Hadi Salim, Michael Smith, and Lucy 1541 Yong. 1543 22. Open Issues 1545 o Middleboxes: 1546 * Due to OAM there are constraints on middleboxes in general. If 1547 middleboxes inspect the packet past the outer IP+UDP and 1548 encapsulation header and look for inner IP and TCP/UDP headers, 1549 that might violate the assumption that OAM packets will be 1550 handled the same as regular data packets. That issue is 1551 broader than just QoS - applies to firewall filters etc. 1552 * Firewalls looking at inner payload? How does that work for OAM 1553 frames? Even if it only drops ... TRILL approach might be an 1554 option? Would that encourage more middleboxes making the 1555 network more fragile? 1556 * Editorially perhaps we should pull the above two into a 1557 separate section about middlebox considerations? 1558 o Next-protocol indication - should it be common across different 1559 encapsulation headers? We will have different ways to indicate 1560 the presence of the first encapsulation header in a packet (could 1561 be a UDP destination port, an Ethernet type, etc depending on the 1562 outer delivery header). But for the next protocol past an 1563 encapsulation header one could envision creating or adoption a 1564 common scheme. Such a would also need to be able to identify 1565 following headers like Ethernet, IPv4/IPv6, ESP, etc. 1567 o Common OAM error reporting protocol? 1568 o There is discussion about timestamps, sequence numbers, etc in 1569 three different parts of the document. OAM, Congestion 1570 Considerations, and Service Model, where the latter argues that a 1571 pseudo-wire service should really be layered on top of the 1572 encapsulation using its own header. Those recommendations seem to 1573 be at odds with each other. Do we envision sequence numbers, 1574 timestamps, etc as potential extensions for OAM and CC? If so, 1575 those extensions could be used to provide a service which doesn't 1576 reorder packets. 1578 23. Change Log 1580 The changes from draft-rtg-dt-encap-01 based on feedback at the 1581 Dallas IETF meeting: 1582 o Setting the context that not all common issues might apply to all 1583 encapsulations, but that they should all be understood before 1584 being dismissed. 1585 o Clarified that IPv6 flow label is useful for entropy in 1586 combination with a UDP source port. 1587 o Editorially added a "summary" set of bullets to most sections. 1588 o Editorial clarifications in the next protocol section to more 1589 clearly state the three areas. 1590 o Folded the two next protocol sections into one. 1591 o Mention the MPLS first nibble issue in the next protocol section. 1592 o Mention that viewing the next protocol as an index to a table with 1593 processing instructions can provide additional flexibility in the 1594 protocol evolution. 1595 o For the OAM "don't forward to end stations" added that defining a 1596 bit seems better than using a special next-protocol value. 1597 o Added mention of DTLS in addition to IPsec for security. 1598 o Added some mention of IPv6 hob-by-hop options of other headers 1599 than potentially can be copied from inner to outer header. 1600 o Added text on architectural considerations when it might make 1601 sense to define an additional header/protocol as opposed to using 1602 the extensibility mechanism in the existing encapsulation 1603 protocol. 1604 o Clarified the "unconstrained TLVs" in the hardware friendly 1605 section. 1606 o Clarified the text around checksum verification and full vs. 1607 header checksums. 1608 o Added wording that the considerations might apply for encaps 1609 outside of the routing area. 1610 o Added references to draft-ietf-pwe3-congcons, 1611 draft-ietf-tsvwg-rfc5405bis, RFC2473, and RFC7325 1613 o Removed reference to RFC3948. 1614 o Updated the acknowledgements section. 1615 o Added this change log section. 1617 24. References 1619 24.1. Normative References 1621 [I-D.ietf-tsvwg-rfc5405bis] 1622 Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1623 Guidelines", draft-ietf-tsvwg-rfc5405bis-02 (work in 1624 progress), April 2015. 1626 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1627 (IPv6) Specification", RFC 2460, December 1998. 1629 [RFC2473] Conta, A. and S. Deering, "Generic Packet Tunneling in 1630 IPv6 Specification", RFC 2473, December 1998. 1632 [RFC2890] Dommety, G., "Key and Sequence Number Extensions to GRE", 1633 RFC 2890, September 2000. 1635 [RFC2983] Black, D., "Differentiated Services and Tunnels", 1636 RFC 2983, October 2000. 1638 [RFC3032] Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y., 1639 Farinacci, D., Li, T., and A. Conta, "MPLS Label Stack 1640 Encoding", RFC 3032, January 2001. 1642 [RFC3931] Lau, J., Townsley, M., and I. Goyret, "Layer Two Tunneling 1643 Protocol - Version 3 (L2TPv3)", RFC 3931, March 2005. 1645 [RFC3985] Bryant, S. and P. Pate, "Pseudo Wire Emulation Edge-to- 1646 Edge (PWE3) Architecture", RFC 3985, March 2005. 1648 [RFC4385] Bryant, S., Swallow, G., Martini, L., and D. McPherson, 1649 "Pseudowire Emulation Edge-to-Edge (PWE3) Control Word for 1650 Use over an MPLS PSN", RFC 4385, February 2006. 1652 [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines 1653 for Application Designers", BCP 145, RFC 5405, 1654 November 2008. 1656 [RFC6040] Briscoe, B., "Tunnelling of Explicit Congestion 1657 Notification", RFC 6040, November 2010. 1659 [RFC6438] Carpenter, B. and S. Amante, "Using the IPv6 Flow Label 1660 for Equal Cost Multipath Routing and Link Aggregation in 1661 Tunnels", RFC 6438, November 2011. 1663 [RFC6790] Kompella, K., Drake, J., Amante, S., Henderickx, W., and 1664 L. Yong, "The Use of Entropy Labels in MPLS Forwarding", 1665 RFC 6790, November 2012. 1667 [RFC6830] Farinacci, D., Fuller, V., Meyer, D., and D. Lewis, "The 1668 Locator/ID Separation Protocol (LISP)", RFC 6830, 1669 January 2013. 1671 [RFC6935] Eubanks, M., Chimento, P., and M. Westerlund, "IPv6 and 1672 UDP Checksums for Tunneled Packets", RFC 6935, April 2013. 1674 [RFC6936] Fairhurst, G. and M. Westerlund, "Applicability Statement 1675 for the Use of IPv6 UDP Datagrams with Zero Checksums", 1676 RFC 6936, April 2013. 1678 [RFC7174] Salam, S., Senevirathne, T., Aldrin, S., and D. Eastlake, 1679 "Transparent Interconnection of Lots of Links (TRILL) 1680 Operations, Administration, and Maintenance (OAM) 1681 Framework", RFC 7174, May 2014. 1683 [RFC7325] Villamizar, C., Kompella, K., Amante, S., Malis, A., and 1684 C. Pignataro, "MPLS Forwarding Compliance and Performance 1685 Requirements", RFC 7325, August 2014. 1687 [RFC7348] Mahalingam, M., Dutt, D., Duda, K., Agarwal, P., Kreeger, 1688 L., Sridhar, T., Bursell, M., and C. Wright, "Virtual 1689 eXtensible Local Area Network (VXLAN): A Framework for 1690 Overlaying Virtualized Layer 2 Networks over Layer 3 1691 Networks", RFC 7348, August 2014. 1693 [RFC7364] Narten, T., Gray, E., Black, D., Fang, L., Kreeger, L., 1694 and M. Napierala, "Problem Statement: Overlays for Network 1695 Virtualization", RFC 7364, October 2014. 1697 [RFC7365] Lasserre, M., Balus, F., Morin, T., Bitar, N., and Y. 1698 Rekhter, "Framework for Data Center (DC) Network 1699 Virtualization", RFC 7365, October 2014. 1701 24.2. Informative References 1703 [I-D.briscoe-conex-data-centre] 1704 Briscoe, B. and M. Sridharan, "Network Performance 1705 Isolation in Data Centres using Congestion Policing", 1706 draft-briscoe-conex-data-centre-02 (work in progress), 1707 February 2014. 1709 [I-D.farrelll-mpls-opportunistic-encrypt] 1710 Farrel, A. and S. Farrell, "Opportunistic Security in MPLS 1711 Networks", draft-farrelll-mpls-opportunistic-encrypt-05 1712 (work in progress), June 2015. 1714 [I-D.gross-geneve] 1715 Gross, J., Sridhar, T., Garg, P., Wright, C., Ganga, I., 1716 Agarwal, P., Duda, K., Dutt, D., and J. Hudson, "Geneve: 1717 Generic Network Virtualization Encapsulation", 1718 draft-gross-geneve-02 (work in progress), October 2014. 1720 [I-D.herbert-gue] 1721 Herbert, T., Yong, L., and O. Zia, "Generic UDP 1722 Encapsulation", draft-herbert-gue-03 (work in progress), 1723 March 2015. 1725 [I-D.herbert-remotecsumoffload] 1726 Herbert, T., "Remote checksum offload for encapsulation", 1727 draft-herbert-remotecsumoffload-01 (work in progress), 1728 November 2014. 1730 [I-D.ietf-mpls-in-udp] 1731 Xu, X., Sheth, N., Yong, L., Callon, R., and D. Black, 1732 "Encapsulating MPLS in UDP", draft-ietf-mpls-in-udp-11 1733 (work in progress), January 2015. 1735 [I-D.ietf-nvo3-arch] 1736 Black, D., Hudson, J., Kreeger, L., Lasserre, M., and T. 1737 Narten, "An Architecture for Overlay Networks (NVO3)", 1738 draft-ietf-nvo3-arch-03 (work in progress), March 2015. 1740 [I-D.ietf-pwe3-congcons] 1741 Stein, Y., Black, D., and B. Briscoe, "Pseudowire 1742 Congestion Considerations", draft-ietf-pwe3-congcons-02 1743 (work in progress), July 2014. 1745 [I-D.ietf-sfc-architecture] 1746 Halpern, J. and C. Pignataro, "Service Function Chaining 1747 (SFC) Architecture", draft-ietf-sfc-architecture-09 (work 1748 in progress), June 2015. 1750 [I-D.ietf-sfc-problem-statement] 1751 Quinn, P. and T. Nadeau, "Service Function Chaining 1752 Problem Statement", draft-ietf-sfc-problem-statement-13 1753 (work in progress), February 2015. 1755 [I-D.ietf-trill-oam-fm] 1756 Senevirathne, T., Finn, N., Salam, S., Kumar, D., 1757 Eastlake, D., Aldrin, S., and L. Yizhou, "TRILL Fault 1758 Management", draft-ietf-trill-oam-fm-11 (work in 1759 progress), October 2014. 1761 [I-D.ietf-tsvwg-circuit-breaker] 1762 Fairhurst, G., "Network Transport Circuit Breakers", 1763 draft-ietf-tsvwg-circuit-breaker-01 (work in progress), 1764 March 2015. 1766 [I-D.ietf-tsvwg-gre-in-udp-encap] 1767 Crabbe, E., Yong, L., Xu, X., and T. Herbert, "GRE-in-UDP 1768 Encapsulation", draft-ietf-tsvwg-gre-in-udp-encap-07 (work 1769 in progress), July 2015. 1771 [I-D.ietf-tsvwg-port-use] 1772 Touch, J., "Recommendations on Using Assigned Transport 1773 Port Numbers", draft-ietf-tsvwg-port-use-11 (work in 1774 progress), April 2015. 1776 [I-D.quinn-sfc-nsh] 1777 Quinn, P., Guichard, J., Surendra, S., Smith, M., 1778 Henderickx, W., Nadeau, T., Agarwal, P., Manur, R., 1779 Chauhan, A., Halpern, J., Majee, S., Elzur, U., Melman, 1780 D., Garg, P., McConnell, B., Wright, C., and K. Kevin, 1781 "Network Service Header", draft-quinn-sfc-nsh-07 (work in 1782 progress), February 2015. 1784 [I-D.saldana-tsvwg-simplemux] 1785 Saldana, J., "Simplemux. A generic multiplexing protocol", 1786 draft-saldana-tsvwg-simplemux-02 (work in progress), 1787 January 2015. 1789 [I-D.shepherd-bier-problem-statement] 1790 Shepherd, G., Dolganow, A., and a. 1791 arkadiy.gulko@thomsonreuters.com, "Bit Indexed Explicit 1792 Replication (BIER) Problem Statement", 1793 draft-shepherd-bier-problem-statement-02 (work in 1794 progress), February 2015. 1796 [I-D.sridharan-virtualization-nvgre] 1797 Garg, P. and Y. Wang, "NVGRE: Network Virtualization using 1798 Generic Routing Encapsulation", 1799 draft-sridharan-virtualization-nvgre-08 (work in 1800 progress), April 2015. 1802 [I-D.wei-tsvwg-tunnel-congestion-feedback] 1803 Wei, X., Zhu, L., Deng, L., and B. Briscoe, "Tunnel 1804 Congestion Feedback", 1805 draft-wei-tsvwg-tunnel-congestion-feedback-04 (work in 1806 progress), June 2015. 1808 [I-D.wijnands-bier-architecture] 1809 Wijnands, I., Rosen, E., Dolganow, A., Przygienda, T., and 1810 S. Aldrin, "Multicast using Bit Index Explicit 1811 Replication", draft-wijnands-bier-architecture-05 (work in 1812 progress), March 2015. 1814 [I-D.wijnands-mpls-bier-encapsulation] 1815 Wijnands, I., Rosen, E., Dolganow, A., Tantsura, J., and 1816 S. Aldrin, "Encapsulation for Bit Index Explicit 1817 Replication in MPLS Networks", 1818 draft-wijnands-mpls-bier-encapsulation-02 (work in 1819 progress), December 2014. 1821 [I-D.xu-bier-encapsulation] 1822 Xu, X., Somasundaram, S., Jacquenet, C., and R. Raszuk, 1823 "BIER Encapsulation", draft-xu-bier-encapsulation-02 (work 1824 in progress), February 2015. 1826 [IEEE802.1Q-2014] 1827 IEEE, "IEEE Standard for Local and metropolitan area 1828 networks--Bridges and Bridged Networks", IEEE Std 802.1Q- 1829 2014, 2014, 1830 . 1832 (Access Controlled link within page) 1834 [RFC0033] Crocker, S., "New Host-Host Protocol", RFC 33, 1835 February 1970. 1837 Authors' Addresses 1839 Erik Nordmark 1840 Arista Networks 1841 5453 Great America Parkway 1842 Santa Clara, CA 95054 1843 USA 1845 Email: nordmark@arista.com 1846 Albert Tian 1847 Ericsson Inc. 1848 300 Holger Way 1849 San Jose, California 95134 1850 USA 1852 Email: albert.tian@ericsson.com 1854 Jesse Gross 1855 VMware 1856 3401 Hillview Ave. 1857 Palo Alto, CA 94304 1858 USA 1860 Email: jgross@vmware.com 1862 Jon Hudson 1863 Brocade Communications Systems, Inc. 1864 130 Holger Way 1865 San Jose, CA 95134 1866 USA 1868 Email: jon.hudson@gmail.com 1870 Lawrence Kreeger 1871 Cisco Systems, Inc. 1872 170 W. Tasman Avenue 1873 San Jose, CA 95134 1874 USA 1876 Email: kreeger@cisco.com 1878 Pankaj Garg 1879 Microsoft 1880 1 Microsoft Way 1881 Redmond, WA 98052 1882 USA 1884 Email: pankajg@microsoft.com 1885 Patricia Thaler 1886 Broadcom Corporation 1887 3151 Zanker Road 1888 San Jose, CA 95134 1889 USA 1891 Email: pthaler@broadcom.com 1893 Tom Herbert 1894 Facebook 1895 1 Hacker Way 1896 Menlo Park, CA 94052 1897 USA 1899 Email: tom@herbertland.com