idnits 2.17.1 draft-ietf-rtgwg-enterprise-pa-multihoming-07.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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 11, 2018) is 2145 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'RFC1122' is defined on line 1987, but no explicit reference was found in the text == Unused Reference: 'RFC1123' is defined on line 1992, but no explicit reference was found in the text == Unused Reference: 'RFC3582' is defined on line 2017, but no explicit reference was found in the text == Unused Reference: 'RFC4116' is defined on line 2022, but no explicit reference was found in the text == Unused Reference: 'RFC4218' is defined on line 2035, but no explicit reference was found in the text == Unused Reference: 'RFC4219' is defined on line 2039, but no explicit reference was found in the text == Unused Reference: 'RFC7157' is defined on line 2052, but no explicit reference was found in the text == Unused Reference: 'RFC8200' is defined on line 2075, but no explicit reference was found in the text == Unused Reference: 'I-D.baker-ipv6-isis-dst-src-routing' is defined on line 2082, but no explicit reference was found in the text == Unused Reference: 'I-D.baker-rtgwg-src-dst-routing-use-cases' is defined on line 2087, but no explicit reference was found in the text == Unused Reference: 'I-D.boutier-babel-source-specific' is defined on line 2093, but no explicit reference was found in the text == Unused Reference: 'I-D.huitema-shim6-ingress-filtering' is defined on line 2098, but no explicit reference was found in the text == Unused Reference: 'I-D.xu-src-dst-bgp' is defined on line 2119, but no explicit reference was found in the text == Unused Reference: 'PATRICIA' is defined on line 2124, but no explicit reference was found in the text == Unused Reference: 'RFC7788' is defined on line 2177, but no explicit reference was found in the text == Unused Reference: 'RFC8041' is defined on line 2181, but no explicit reference was found in the text == Unused Reference: 'RFC8305' is defined on line 2186, but no explicit reference was found in the text ** Obsolete normative reference: RFC 3315 (Obsoleted by RFC 8415) ** Obsolete normative reference: RFC 4242 (Obsoleted by RFC 8415) == Outdated reference: A later version (-11) exists of draft-ietf-intarea-provisioning-domains-02 == Outdated reference: A later version (-07) exists of draft-ietf-rtgwg-dst-src-routing-06 -- Obsolete informational reference (is this intentional?): RFC 3736 (Obsoleted by RFC 8415) -- Obsolete informational reference (is this intentional?): RFC 4941 (Obsoleted by RFC 8981) Summary: 2 errors (**), 0 flaws (~~), 20 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Routing Working Group F. Baker 3 Internet-Draft 4 Intended status: Informational C. Bowers 5 Expires: December 13, 2018 Juniper Networks 6 J. Linkova 7 Google 8 June 11, 2018 10 Enterprise Multihoming using Provider-Assigned Addresses without Network 11 Prefix Translation: Requirements and Solution 12 draft-ietf-rtgwg-enterprise-pa-multihoming-07 14 Abstract 16 Connecting an enterprise site to multiple ISPs using provider- 17 assigned addresses is difficult without the use of some form of 18 Network Address Translation (NAT). Much has been written on this 19 topic over the last 10 to 15 years, but it still remains a problem 20 without a clearly defined or widely implemented solution. Any 21 multihoming solution without NAT requires hosts at the site to have 22 addresses from each ISP and to select the egress ISP by selecting a 23 source address for outgoing packets. It also requires routers at the 24 site to take into account those source addresses when forwarding 25 packets out towards the ISPs. 27 This document attempts to define a complete solution to this problem. 28 It covers the behavior of routers to forward traffic taking into 29 account source address, and it covers the behavior of host to select 30 appropriate source addresses. It also covers any possible role that 31 routers might play in providing information to hosts to help them 32 select appropriate source addresses. In the process of exploring 33 potential solutions, this documents also makes explicit requirements 34 for how the solution would be expected to behave from the perspective 35 of an enterprise site network administrator . 37 Status of This Memo 39 This Internet-Draft is submitted in full conformance with the 40 provisions of BCP 78 and BCP 79. 42 Internet-Drafts are working documents of the Internet Engineering 43 Task Force (IETF). Note that other groups may also distribute 44 working documents as Internet-Drafts. The list of current Internet- 45 Drafts is at https://datatracker.ietf.org/drafts/current/. 47 Internet-Drafts are draft documents valid for a maximum of six months 48 and may be updated, replaced, or obsoleted by other documents at any 49 time. It is inappropriate to use Internet-Drafts as reference 50 material or to cite them other than as "work in progress." 52 This Internet-Draft will expire on December 13, 2018. 54 Copyright Notice 56 Copyright (c) 2018 IETF Trust and the persons identified as the 57 document authors. All rights reserved. 59 This document is subject to BCP 78 and the IETF Trust's Legal 60 Provisions Relating to IETF Documents 61 (https://trustee.ietf.org/license-info) in effect on the date of 62 publication of this document. Please review these documents 63 carefully, as they describe your rights and restrictions with respect 64 to this document. Code Components extracted from this document must 65 include Simplified BSD License text as described in Section 4.e of 66 the Trust Legal Provisions and are provided without warranty as 67 described in the Simplified BSD License. 69 Table of Contents 71 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 72 2. Requirements Language . . . . . . . . . . . . . . . . . . . . 6 73 3. Enterprise Multihoming Requirements . . . . . . . . . . . . . 6 74 3.1. Simple ISP Connectivity with Connected SERs . . . . . . . 6 75 3.2. Simple ISP Connectivity Where SERs Are Not Directly 76 Connected . . . . . . . . . . . . . . . . . . . . . . . . 8 77 3.3. Enterprise Network Operator Expectations . . . . . . . . 9 78 3.4. More complex ISP connectivity . . . . . . . . . . . . . . 12 79 3.5. ISPs and Provider-Assigned Prefixes . . . . . . . . . . . 14 80 3.6. Simplified Topologies . . . . . . . . . . . . . . . . . . 15 81 4. Generating Source-Prefix-Scoped Forwarding Tables . . . . . 15 82 5. Mechanisms For Hosts To Choose Good Source Addresses In A 83 Multihomed Site . . . . . . . . . . . . . . . . . . . . . . . 22 84 5.1. Source Address Selection Algorithm on Hosts . . . . . . . 24 85 5.2. Selecting Source Address When Both Uplinks Are Working . 27 86 5.2.1. Distributing Address Selection Policy Table with 87 DHCPv6 . . . . . . . . . . . . . . . . . . . . . . . 27 88 5.2.2. Controlling Source Address Selection With Router 89 Advertisements . . . . . . . . . . . . . . . . . . . 27 90 5.2.3. Controlling Source Address Selection With ICMPv6 . . 29 91 5.2.4. Summary of Methods For Controlling Source Address 92 Selection To Implement Routing Policy . . . . . . . . 31 93 5.3. Selecting Source Address When One Uplink Has Failed . . . 32 94 5.3.1. Controlling Source Address Selection With DHCPv6 . . 33 95 5.3.2. Controlling Source Address Selection With Router 96 Advertisements . . . . . . . . . . . . . . . . . . . 34 98 5.3.3. Controlling Source Address Selection With ICMPv6 . . 35 99 5.3.4. Summary Of Methods For Controlling Source Address 100 Selection On The Failure Of An Uplink . . . . . . . . 35 101 5.4. Selecting Source Address Upon Failed Uplink Recovery . . 36 102 5.4.1. Controlling Source Address Selection With DHCPv6 . . 36 103 5.4.2. Controlling Source Address Selection With Router 104 Advertisements . . . . . . . . . . . . . . . . . . . 36 105 5.4.3. Controlling Source Address Selection With ICMP . . . 37 106 5.4.4. Summary Of Methods For Controlling Source Address 107 Selection Upon Failed Uplink Recovery . . . . . . . . 37 108 5.5. Selecting Source Address When All Uplinks Failed . . . . 38 109 5.5.1. Controlling Source Address Selection With DHCPv6 . . 38 110 5.5.2. Controlling Source Address Selection With Router 111 Advertisements . . . . . . . . . . . . . . . . . . . 38 112 5.5.3. Controlling Source Address Selection With ICMPv6 . . 39 113 5.5.4. Summary Of Methods For Controlling Source Address 114 Selection When All Uplinks Failed . . . . . . . . . . 39 115 5.6. Summary Of Methods For Controlling Source Address 116 Selection . . . . . . . . . . . . . . . . . . . . . . . . 39 117 5.7. Other Configuration Parameters . . . . . . . . . . . . . 41 118 5.7.1. DNS Configuration . . . . . . . . . . . . . . . . . . 41 119 6. Deployment Considerations . . . . . . . . . . . . . . . . . . 42 120 7. Other Solutions . . . . . . . . . . . . . . . . . . . . . . . 43 121 7.1. Shim6 . . . . . . . . . . . . . . . . . . . . . . . . . . 43 122 7.2. IPv6-to-IPv6 Network Prefix Translation . . . . . . . . . 43 123 7.3. Multipath Transport . . . . . . . . . . . . . . . . . . . 43 124 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 44 125 9. Security Considerations . . . . . . . . . . . . . . . . . . . 44 126 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 44 127 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 44 128 11.1. Normative References . . . . . . . . . . . . . . . . . . 44 129 11.2. Informative References . . . . . . . . . . . . . . . . . 46 130 Appendix A. Change Log . . . . . . . . . . . . . . . . . . . . . 49 131 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 49 133 1. Introduction 135 Site multihoming, the connection of a subscriber network to multiple 136 upstream networks using redundant uplinks, is a common enterprise 137 architecture for improving the reliability of its Internet 138 connectivity. If the site uses provider-independent (PI) addresses, 139 all traffic originating from the enterprise can use source addresses 140 from the PI address space. Site multihoming with PI addresses is 141 commonly used with both IPv4 and IPv6, and does not present any new 142 technical challenges. 144 It may be desirable for an enterprise site to connect to multiple 145 ISPs using provider-assigned (PA) addresses, instead of PI addresses. 147 Multihoming with provider-assigned addresses is typically less 148 expensive for the enterprise relative to using provider-independent 149 addresses. PA multihoming is also a practice that should be 150 facilitated and encouraged because it does not add to the size of the 151 Internet routing table, whereas PI multihoming does. Note that PA is 152 also used to mean "provider-aggregatable". In this document we 153 assume that provider-assigned addresses are always provider- 154 aggregatable. 156 With PA multihoming, for each ISP connection, the site is assigned a 157 prefix from within an address block allocated to that ISP by its 158 National or Regional Internet Registry. In the simple case of two 159 ISPs (ISP-A and ISP-B), the site will have two different prefixes 160 assigned to it (prefix-A and prefix-B). This arrangement is 161 problematic. First, packets with the "wrong" source address may be 162 dropped by one of the ISPs. In order to limit denial of service 163 attacks using spoofed source addresses, BCP38 [RFC2827] recommends 164 that ISPs filter traffic from customer sites to only allow traffic 165 with a source address that has been assigned by that ISP. So a 166 packet sent from a multihomed site on the uplink to ISP-B with a 167 source address in prefix-A may be dropped by ISP-B. 169 However, even if ISP-B does not implement BCP38 or ISP-B adds 170 prefix-A to its list of allowed source addresses on the uplink from 171 the multihomed site, two-way communication may still fail. If the 172 packet with source address in prefix-A was sent to ISP-B because the 173 uplink to ISP-A failed, then if ISP-B does not drop the packet and 174 the packet reaches its destination somewhere on the Internet, the 175 return packet will be sent back with a destination address in prefix- 176 A. The return packet will be routed over the Internet to ISP-A, but 177 it will not be delivered to the multihomed site because its link with 178 ISP-A has failed. Two-way communication would require some 179 arrangement for ISP-B to advertise prefix-A when the uplink to ISP-A 180 fails. 182 Note that the same may be true with a provider that does not 183 implement BCP 38, if his upstream provider does, or has no 184 corresponding route. The issue is not that the immediate provider 185 implements ingress filtering; it is that someone upstream does, or 186 lacks a route. 188 With IPv4, this problem is commonly solved by using [RFC1918] private 189 address space within the multi-homed site and Network Address 190 Translation (NAT) or Network Address/Port Translation (NAPT) on the 191 uplinks to the ISPs. However, one of the goals of IPv6 is to 192 eliminate the need for and the use of NAT or NAPT. Therefore, 193 requiring the use of NAT or NAPT for an enterprise site to multihome 194 with provider-assigned addresses is not an attractive solution. 196 [RFC6296] describes a translation solution specifically tailored to 197 meet the requirements of multi-homing with provider-assigned IPv6 198 addresses. With the IPv6-to-IPv6 Network Prefix Translation (NPTv6) 199 solution, within the site an enterprise can use Unique Local 200 Addresses [RFC4193] or the prefix assigned by one of the ISPs. As 201 traffic leaves the site on an uplink to an ISP, the source address 202 gets translated to an address within the prefix assigned by the ISP 203 on that uplink in a predictable and reversible manner. [RFC6296] is 204 currently classified as Experimental, and it has been implemented by 205 several vendors. See Section 7.2, for more discussion of NPTv6. 207 This document defines routing requirements for enterprise multihoming 208 using provider-assigned IPv6 addresses. We have made no attempt to 209 write these requirements in a manner that is agnostic to potential 210 solutions. Instead, this document focuses on the following general 211 class of solutions. 213 Each host at the enterprise has multiple addresses, at least one from 214 each ISP-assigned prefix. Each host, as discussed in Section 5.1 and 215 [RFC6724], is responsible for choosing the source address applied to 216 each packet it sends. A host SHOULD be able respond dynamically to 217 the failure of an uplink to a given ISP by no longer sending packets 218 with the source address corresponding to that ISP. Potential 219 mechanisms for the communication of changes in the network to the 220 host are Neighbor Discovery Router Advertisements, DHCPv6, and 221 ICMPv6. 223 The routers in the enterprise network are responsible for ensuring 224 that packets are delivered to the "correct" ISP uplink based on 225 source address. This requires that at least some routers in the site 226 network are able to take into account the source address of a packet 227 when deciding how to route it. That is, some routers must be capable 228 of some form of Source Address Dependent Routing (SADR), if only as 229 described in [RFC3704]. At a minimum, the routers connected to the 230 ISP uplinks (the site exit routers or SERs) must be capable of Source 231 Address Dependent Routing. Expanding the connected domain of routers 232 capable of SADR from the site exit routers deeper into the site 233 network will generally result in more efficient routing of traffic 234 with external destinations. 236 The document first looks in more detail at the enterprise networking 237 environments in which this solution is expected to operate. It then 238 discusses existing and proposed mechanisms for hosts to select the 239 source address applied to packets. Finally, it looks at the 240 requirements for routing that are needed to support these enterprise 241 network scenarios and the mechanisms by which hosts are expected to 242 select source addresses dynamically based on network state. 244 2. Requirements Language 246 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 247 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 248 "OPTIONAL" in this document are to be interpreted as described in BCP 249 14 [RFC2119] [RFC8174] when, and only when, they appear in all 250 capitals, as shown here. 252 3. Enterprise Multihoming Requirements 254 3.1. Simple ISP Connectivity with Connected SERs 256 We start by looking at a scenario in which a site has connections to 257 two ISPs, as shown in Figure 1. The site is assigned the prefix 258 2001:db8:0:a000::/52 by ISP-A and prefix 2001:db8:0:b000::/52 by ISP- 259 B. We consider three hosts in the site. H31 and H32 are on a LAN 260 that has been assigned subnets 2001:db8:0:a010::/64 and 261 2001:db8:0:b010::/64. H31 has been assigned the addresses 262 2001:db8:0:a010::31 and 2001:db8:0:b010::31. H32 has been assigned 263 2001:db8:0:a010::32 and 2001:db8:0:b010::32. H41 is on a different 264 subnet that has been assigned 2001:db8:0:a020::/64 and 265 2001:db8:0:b020::/64. 267 2001:db8:0:1234::101 H101 268 | 269 | 270 2001:db8:0:a010::31 -------- 271 2001:db8:0:b010::31 ,-----. / \ 272 +--+ +--+ +----+ ,' `. : : 273 +---|R1|---|R4|---+---|SERa|-+ ISP-A +--+-- : 274 H31--+ +--+ +--+ | +----+ `. ,' : : 275 | | `-----' : Internet : 276 | | : : 277 | | : : 278 | | : : 279 | | ,-----. : : 280 H32--+ +--+ | +----+ ,' `. : : 281 +---|R2|----------+---|SERb|-+ ISP-B +--+-- : 282 +--+ | +----+ `. ,' : : 283 | `-----' : : 284 | : : 285 +--+ +--+ +--+ \ / 286 H41------|R3|--|R5|--|R6| -------- 287 +--+ +--+ +--+ 289 2001:db8:0:a020::41 290 2001:db8:0:b020::41 292 Figure 1: Simple ISP Connectivity With Connected SERs 294 We refer to a router that connects the site to an ISP as a site edge 295 router(SER). Several other routers provide connectivity among the 296 internal hosts (H31, H32, and H41), as well as connecting the 297 internal hosts to the Internet through SERa and SERb. In this 298 example SERa and SERb share a direct connection to each other. In 299 Section 3.2, we consider a scenario where this is not the case. 301 For the moment, we assume that the hosts are able to make good 302 choices about which source addresses through some mechanism that 303 doesn't involve the routers in the site network. Here, we focus on 304 primary task of the routed site network, which is to get packets 305 efficiently to their destinations, while sending a packet to the ISP 306 that assigned the prefix that matches the source address of the 307 packet. In Section 5, we examine what role the routed network may 308 play in helping hosts make good choices about source addresses for 309 packets. 311 With this solution, routers will need form of Source Address 312 Dependent Routing, which will be new functionality. It would be 313 useful if an enterprise site does not need to upgrade all routers to 314 support the new SADR functionality in order to support PA multi- 315 homing. We consider if this is possible and what are the tradeoffs 316 of not having all routers in the site support SADR functionality. 318 In the topology in Figure 1, it is possible to support PA multihoming 319 with only SERa and SERb being capable of SADR. The other routers can 320 continue to forward based only on destination address, and exchange 321 routes that only consider destination address. In this scenario, 322 SERa and SERb communicate source-scoped routing information across 323 their shared connection. When SERa receives a packet with a source 324 address matching prefix 2001:db8:0:b000::/52 , it forwards the packet 325 to SERb, which forwards it on the uplink to ISP-B. The analogous 326 behaviour holds for traffic that SERb receives with a source address 327 matching prefix 2001:db8:0:a000::/52. 329 In Figure 1, when only SERa and SERb are capable of source address 330 dependent routing, PA multi-homing will work. However, the paths 331 over which the packets are sent will generally not be the shortest 332 paths. The forwarding paths will generally be more efficient as more 333 routers are capable of SADR. For example, if R4, R2, and R6 are 334 upgraded to support SADR, then can exchange source-scoped routes with 335 SERa and SERb. They will then know to send traffic with a source 336 address matching prefix 2001:db8:0:b000::/52 directly to SERb, 337 without sending it to SERa first. 339 3.2. Simple ISP Connectivity Where SERs Are Not Directly Connected 341 In Figure 2, we modify the topology slightly by inserting R7, so that 342 SERa and SERb are no longer directly connected. With this topology, 343 it is not enough to just enable SADR routing on SERa and SERb to 344 support PA multi-homing. There are two solutions to ways to enable 345 PA multihoming in this topology. 347 2001:db8:0:1234::101 H101 348 | 349 | 350 2001:db8:0:a010::31 -------- 351 2001:db8:0:b010::31 ,-----. / \ 352 +--+ +--+ +----+ ,' `. : : 353 +---|R1|---|R4|---+---|SERa|-+ ISP-A +--+-- : 354 H31--+ +--+ +--+ | +----+ `. ,' : : 355 | | `-----' : Internet : 356 | +--+ : : 357 | |R7| : : 358 | +--+ : : 359 | | ,-----. : : 360 H32--+ +--+ | +----+ ,' `. : : 361 +---|R2|----------+---|SERb|-+ ISP-B +--+-- : 362 +--+ | +----+ `. ,' : : 363 | `-----' : : 364 | : : 365 +--+ +--+ +--+ \ / 366 H41------|R3|--|R5|--|R6| -------- 367 +--+ +--+ +--+ | 368 | 369 2001:db8:0:a020::41 2001:db8:0:5678::501 H501 370 2001:db8:0:b020::41 372 Figure 2: Simple ISP Connectivity Where SERs Are Not Directly 373 Connected 375 One option is to effectively modify the topology by creating a 376 logical tunnel between SERa and SERb, using GRE for example. 377 Although SERa and SERb are not directly connected physically in this 378 topology, they can be directly connected logically by a tunnel. 380 The other option is to enable SADR functionality on R7. In this way, 381 R7 will exchange source-scoped routes with SERa and SERb, making the 382 three routers act as a single SADR domain. This illustrates the 383 basic principle that the minimum requirement for the routed site 384 network to support PA multi-homing is having all of the site exit 385 routers be part of a connected SADR domain. Extending the connected 386 SADR domain beyond that point can produce more efficient forwarding 387 paths. 389 3.3. Enterprise Network Operator Expectations 391 Before considering a more complex scenario, let's look in more detail 392 at the reasonably simple multihoming scenario in Figure 2 to 393 understand what can reasonably be expected from this solution. As a 394 general guiding principle, we assume an enterprise network operator 395 will expect a multihomed network to behave as close as to a single- 396 homed network as possible. So a solution that meets those 397 expectations where possible is a good thing. 399 For traffic between internal hosts and traffic from outside the site 400 to internal hosts, an enterprise network operator would expect there 401 to be no visible change in the path taken by this traffic, since this 402 traffic does not need to be routed in a way that depends on source 403 address. It is also reasonable to expect that internal hosts should 404 be able to communicate with each other using either of their source 405 addresses without restriction. For example, H31 should be able to 406 communicate with H41 using a packet with S=2001:db8:0:a010::31, 407 D=2001:db8:0:b010::41, regardless of the state of uplink to ISP-B. 409 These goals can be accomplished by having all of the routers in the 410 network continue to originate normal unscoped destination routes for 411 their connected networks. If we can arrange so that these unscoped 412 destination routes get used for forwarding this traffic, then we will 413 have accomplished the goal of keeping forwarding of traffic destined 414 for internal hosts, unaffected by the multihoming solution. 416 For traffic destined for external hosts, it is reasonable to expect 417 that traffic with an source address from the prefix assigned by ISP-A 418 to follow the path to that the traffic would follow if there is no 419 connection to ISP-B. This can be accomplished by having SERa 420 originate a source-scoped route of the form (S=2001:db8:0:a000::/52, 421 D=::/0) . If all of the routers in the site support SADR, then the 422 path of traffic exiting via ISP-A can match that expectation. If 423 some routers don't support SADR, then it is reasonable to expect that 424 the path for traffic exiting via ISP-A may be different within the 425 site. This is a tradeoff that the enterprise network operator may 426 decide to make. 428 It is important to understand how this multihoming solution behaves 429 when an uplink to one of the ISPs fails. To simplify this 430 discussion, we assume that all routers in the site support SADR. We 431 first start by looking at how the network operates when the uplinks 432 to both ISP-A and ISP-B are functioning properly. SERa originates a 433 source-scoped route of the form (S=2001:db8:0:a000::/52, D=::/0), and 434 SERb is originates a source-scoped route of the form 435 (S=2001:db8:0:b000::/52, D=::/0). These routes are distributed 436 through the routers in the site, and they establish within the 437 routers two set of forwarding paths for traffic leaving the site. 438 One set of forwarding paths is for packets with source address in 439 2001:db8:0:a000::/52. The other set of forwarding paths is for 440 packets with source address in 2001:db8:0:b000::/52. The normal 441 destination routes which are not scoped to these two source prefixes 442 play no role in the forwarding. Whether a packet exits the site via 443 SERa or via SERb is completely determined by the source address 444 applied to the packet by the host. So for example, when host H31 445 sends a packet to host H101 with (S=2001:db8:0:a010::31, 446 D=2001:db8:0:1234::101), the packet will only be sent out the link 447 from SERa to ISP-A. 449 Now consider what happens when the uplink from SERa to ISP-A fails. 450 The only way for the packets from H31 to reach H101 is for H31 to 451 start using the source address for ISP-B. H31 needs to send the 452 following packet: (S=2001:db8:0:b010::31, D=2001:db8:0:1234::101). 454 This behavior is very different from the behavior that occurs with 455 site multihoming using PI addresses or with PA addresses using NAT. 456 In these other multi-homing solutions, hosts do not need to react to 457 network failures several hops away in order to regain Internet 458 access. Instead, a host can be largely unaware of the failure of an 459 uplink to an ISP. When multihoming with PA addresses and NAT, 460 existing sessions generally need to be re-established after a failure 461 since the external host will receive packets from the internal host 462 with a new source address. However, new sessions can be established 463 without any action on the part of the hosts. 465 Another example where the behavior of this multihoming solution 466 differs significantly from that of multihoming with PI address or 467 with PA addresses using NAT is in the ability of the enterprise 468 network operator to route traffic over different ISPs based on 469 destination address. We still consider the fairly simple network of 470 Figure 2 and assume that uplinks to both ISPs are functioning. 471 Assume that the site is multihomed using PA addresses and NAT, and 472 that SERa and SERb each originate a normal destination route for 473 D=::/0, with the route origination dependent on the state of the 474 uplink to the respective ISP. 476 Now suppose it is observed that an important application running 477 between internal hosts and external host H101 experience much better 478 performance when the traffic passes through ISP-A (perhaps because 479 ISP-A provides lower latency to H101.) When multihoming this site 480 with PI addresses or with PA addresses and NAT, the enterprise 481 network operator can configure SERa to originate into the site 482 network a normal destination route for D=2001:db8:0:1234::/64 (the 483 destination prefix to reach H101) that depends on the state of the 484 uplink to ISP-A. When the link to ISP-A is functioning, the 485 destination route D=2001:db8:0:1234::/64 will be originated by SERa, 486 so traffic from all hosts will use ISP-A to reach H101 based on the 487 longest destination prefix match in the route lookup. 489 Implementing the same routing policy is more difficult with the PA 490 multihoming solution described in this document since it doesn't use 491 NAT. By design, the only way to control where a packet exits this 492 network is by setting the source address of the packet. Since the 493 network cannot modify the source address without NAT, the host must 494 set it. To implement this routing policy, each host needs to use the 495 source address from the prefix assigned by ISP-A to send traffic 496 destined for H101. Mechanisms have been proposed to allow hosts to 497 choose the source address for packets in a fine grained manner. We 498 will discuss these proposals in Section 5. However, interacting with 499 host operating systems in some manner to ensure a particular source 500 address is chosen for a particular destination prefix is not what an 501 enterprise network administrator would expect to have to do to 502 implement this routing policy. 504 3.4. More complex ISP connectivity 506 The previous sections considered two variations of a simple 507 multihoming scenario where the site is connected to two ISPs offering 508 only Internet connectivity. It is likely that many actual enterprise 509 multihoming scenarios will be similar to this simple example. 510 However, there are more complex multihoming scenarios that we would 511 like this solution to address as well. 513 It is fairly common for an ISP to offer a service in addition to 514 Internet access over the same uplink. Two variation of this are 515 reflected in Figure 3. In addition to Internet access, ISP-A offers 516 a service which requires the site to access host H51 at 517 2001:db8:0:5555::51. The site has a single physical and logical 518 connection with ISP-A, and ISP-A only allows access to H51 over that 519 connection. So when H32 needs to access the service at H51 it needs 520 to send packets with (S=2001:db8:0:a010::32, D=2001:db8:0:5555::51) 521 and those packets need to be forward out the link from SERa to ISP-A. 523 2001:db8:0:1234::101 H101 524 | 525 | 526 2001:db8:0:a010::31 -------- 527 2001:db8:0:b010::31 ,-----. / \ 528 +--+ +--+ +----+ ,' `. : : 529 +---|R1|---|R4|---+---|SERa|-+ ISP-A +--+-- : 530 H31--+ +--+ +--+ | +----+ `. ,' : : 531 | | `-----' : Internet : 532 | | | : : 533 | | H51 : : 534 | | 2001:db8:0:5555::51 : : 535 | +--+ : : 536 | |R7| : : 537 | +--+ : : 538 | | : : 539 | | ,-----. : : 540 H32--+ +--+ | +-----+ ,' `. : : 541 +---|R2|-----+----+--|SERb1|-+ ISP-B +--+-- : 542 +--+ | +-----+ `. ,' : : 543 +--+ `--|--' : : 544 2001:db8:0:a010::32 |R8| | \ / 545 +--+ ,--|--. -------- 546 | +-----+ ,' `. | 547 +-------|SERb2|-+ ISP-B | | 548 | +-----+ `. ,' H501 549 | `-----' 2001:db8:0:5678 550 | | ::501 551 +--+ +--+ H61 552 H41------|R3|--|R5| 2001:db8:0:6666::61 553 +--+ +--+ 555 2001:db8:0:a020::41 556 2001:db8:0:b020::41 558 Figure 3: Internet access and services offered by ISP-A and ISP-B 560 ISP-B illustrates a variation on this scenario. In addition to 561 Internet access, ISP-B also offers a service which requires the site 562 to access host H61. The site has two connections to two different 563 parts of ISP-B (shown as SERb1 and SERb2 in Figure 3). ISP-B expects 564 Internet traffic to use the uplink from SERb1, while it expects it 565 expects traffic destined for the service at H61 to use the uplink 566 from SERb2. For either uplink, ISP-B expects the ingress traffic to 567 have a source address matching the prefix it assigned to the site, 568 2001:db8:0:b000::/52. 570 As discussed before, we rely completely on the internal host to set 571 the source address of the packet properly. In the case of a packet 572 sent by H31 to access the service in ISP-B at H61, we expect the 573 packet to have the following addresses: (S=2001:db8:0:b010::31, 574 D=2001:db8:0:6666::61). The routed network has two potential ways of 575 distributing routes so that this packet exits the site on the uplink 576 at SERb2. 578 We could just rely on normal destination routes, without using 579 source-prefix scoped routes. If we have SERb2 originate a normal 580 unscoped destination route for D=2001:db8:0:6666::/64, the packets 581 from H31 to H61 will exit the site at SERb2 as desired. We should 582 not have to worry about SERa needing to originate the same route, 583 because ISP-B should choose a globally unique prefix for the service 584 at H61. 586 The alternative is to have SERb2 originate a source-prefix-scoped 587 destination route of the form (S=2001:db8:0:b000::/52, 588 D=2001:db8:0:6666::/64). From a forwarding point of view, the use of 589 the source-prefix-scoped destination route would result in traffic 590 with source addresses corresponding only to ISP-B being sent to 591 SERb2. Instead, the use of the unscoped destination route would 592 result in traffic with source addresses corresponding to ISP-A and 593 ISP-B being sent to SERb2, as long as the destination address matches 594 the destination prefix. It seems like either forwarding behavior 595 would be acceptable. 597 However, from the point of view of the enterprise network 598 administrator trying to configure, maintain, and trouble-shoot this 599 multihoming solution, it seems much clearer to have SERb2 originate 600 the source-prefix-scoped destination route correspond to the service 601 offered by ISP-B. In this way, all of the traffic leaving the site 602 is determined by the source-prefix-scoped routes, and all of the 603 traffic within the site or arriving from external hosts is determined 604 by the unscoped destination routes. Therefore, for this multihoming 605 solution we choose to originate source-prefix-scoped routes for all 606 traffic leaving the site. 608 3.5. ISPs and Provider-Assigned Prefixes 610 While we expect that most site multihoming involves connecting to 611 only two ISPs, this solution allows for connections to an arbitrary 612 number of ISPs to be supported. However, when evaluating scalable 613 implementations of the solution, it would be reasonable to assume 614 that the maximum number of ISPs that a site would connect to is five. 616 It is also useful to note that the prefixes assigned to the site by 617 different ISPs will not overlap. This must be the case , since the 618 provider-assigned addresses have to be globally unique. 620 3.6. Simplified Topologies 622 The topologies of many enterprise sites using this multihoming 623 solution may in practice be simpler than the examples that we have 624 used. The topology in Figure 1 could be further simplified by having 625 all hosts directly connected to the LAN connecting the two site exit 626 routers, SERa and SERb. The topology could also be simplified by 627 having the uplinks to ISP-A and ISP-B both connected to the same site 628 exit router. However, it is the aim of this draft to provide a 629 solution that applies to a broad a range of enterprise site network 630 topologies, so this draft focuses on providing a solution to the more 631 general case. The simplified cases will also be supported by this 632 solution, and there may even be optimizations that can be made for 633 simplified cases. This solution however needs to support more 634 complex topologies. 636 We are starting with the basic assumption that enterprise site 637 networks can be quite complex from a routing perspective. However, 638 even a complex site network can be multihomed to different ISPs with 639 PA addresses using IPv4 and NAT. It is not reasonable to expect an 640 enterprise network operator to change the routing topology of the 641 site in order to deploy IPv6. 643 4. Generating Source-Prefix-Scoped Forwarding Tables 645 So far we have described in general terms how the routers in this 646 solution that are capable of Source Address Dependent Routing will 647 forward traffic using both normal unscoped destination routes and 648 source-prefix-scoped destination routes. Here we give a precise 649 method for generating a source-prefix-scoped forwarding table on a 650 router that supports SADR. 652 1. Compute the next-hops for the source-prefix-scoped destination 653 prefixes using only routers in the connected SADR domain. These 654 are the initial source-prefix-scoped forwarding table entries. 656 2. Compute the next-hops for the unscoped destination prefixes using 657 all routers in the IGP. This is the unscoped forwarding table. 659 3. Augment each less specific source-prefix-scoped forwarding table 660 with all more specific source-prefix-scoped forwarding tables 661 entries based on the following rule. If the destination prefix 662 of the less specific source-prefix-scoped forwarding entry 663 exactly matches the destination prefix of an existing more 664 specific source-prefix-scoped forwarding entry (including 665 destination prefix length), then do not add the less specific 666 source-prefix-scoped forwarding entry. If the destination prefix 667 does NOT match an existing entry, then add the entry to the more 668 source-prefix-scoped forwarding table. As the unscoped 669 forwarding table is considered to be scoped to ::/0 this process 670 starts with propagating routes from the unscoped forwarding table 671 to source-prefix-scoped forwarding tables and then continues with 672 propagating routes to more-specific-source-prefix-scoped 673 forwarding tables should they exist. 675 The forward tables produced by this process are used in the following 676 way to forward packets. 678 1. Select the most specific (longerst prefix match) source-prefix- 679 scoped forwarding table that matches the source address of the 680 packet (again, the unscoped forwarding table is considered to be 681 scoped to ::/0). 683 2. Look up the destination address of the packet in the selected 684 forwarding table to determine the next-hop for the packet. 686 The following example illustrates how this process is used to create 687 a forwarding table for each provider-assigned source prefix. We 688 consider the multihomed site network in Figure 3. Initially we 689 assume that all of the routers in the site network support SADR. 690 Figure 4 shows the routes that are originated by the routers in the 691 site network. 693 Routes originated by SERa: 694 (S=2001:db8:0:a000::/52, D=2001:db8:0:5555/64) 695 (S=2001:db8:0:a000::/52, D=::/0) 696 (D=2001:db8:0:5555::/64) 697 (D=::/0) 699 Routes originated by SERb1: 700 (S=2001:db8:0:b000::/52, D=::/0) 701 (D=::/0) 703 Routes originated by SERb2: 704 (S=2001:db8:0:b000::/52, D=2001:db8:0:6666::/64) 705 (D=2001:db8:0:6666::/64) 707 Routes originated by R1: 708 (D=2001:db8:0:a010::/64) 709 (D=2001:db8:0:b010::/64) 711 Routes originated by R2: 712 (D=2001:db8:0:a010::/64) 713 (D=2001:db8:0:b010::/64) 715 Routes originated by R3: 716 (D=2001:db8:0:a020::/64) 717 (D=2001:db8:0:b020::/64) 719 Figure 4: Routes Originated by Routers in the Site Network 721 Each SER originates destination routes which are scoped to the source 722 prefix assigned by the ISP that the SER connects to. Note that the 723 SERs also originate the corresponding unscoped destination route. 724 This is not needed when all of the routers in the site support SADR. 725 However, it is required when some routers do not support SADR. This 726 will be discussed in more detail later. 728 We focus on how R8 constructs its source-prefix-scoped forwarding 729 tables from these route advertisements. R8 computes the next hops 730 for destination routes which are scoped to the source prefix 731 2001:db8:0:a000::/52. The results are shown in the first table in 732 Figure 5. (In this example, the next hops are computed assuming that 733 all links have the same metric.) Then, R8 computes the next hops for 734 destination routes which are scoped to the source prefix 735 2001:db8:0:b000::/52. The results are shown in the second table in 736 Figure 5 . Finally, R8 computes the next hops for the unscoped 737 destination prefixes. The results are shown in the third table in 738 Figure 5. 740 forwarding entries scoped to 741 source prefix = 2001:db8:0:a000::/52 742 ============================================ 743 D=2001:db8:0:5555/64 NH=R7 744 D=::/0 NH=R7 746 forwarding entries scoped to 747 source prefix = 2001:db8:0:b000::/52 748 ============================================ 749 D=2001:db8:0:6666/64 NH=SERb2 750 D=::/0 NH=SERb1 752 unscoped forwarding entries 753 ============================================ 754 D=2001:db8:0:a010::/64 NH=R2 755 D=2001:db8:0:b010::/64 NH=R2 756 D=2001:db8:0:a020::/64 NH=R5 757 D=2001:db8:0:b020::/64 NH=R5 758 D=2001:db8:0:5555::/64 NH=R7 759 D=2001:db8:0:6666::/64 NH=SERb2 760 D=::/0 NH=SERb1 762 Figure 5: Forwarding Entries Computed at R8 764 The final step is for R8 to augment the less specific source-prefix- 765 scoped forwarding entries with more specific source-prefix-scoped 766 forwarding entries. As unscoped forwarding table is considered being 767 scoped to ::/0 and both 2001:db8:0:a000::/52 and 2001:db8:0:b000::/52 768 are more specific prefixes of ::/0, the unscoped (scoped to ::/0) 769 forwarding table needs to be augmented with both more specific 770 source-prefix-scoped tables. If an less specific scoped forwarding 771 entry has the exact same destination prefix as an more specific 772 source-prefix-scoped forwarding entry (including destination prefix 773 length), then the more specific source-prefix-scoped forwarding entry 774 wins. 776 As as an example of how the source scoped forwarding entries are 777 augmented, we consider how the two entries in the first table in 778 Figure 5 (the table for source prefix = 2001:db8:0:a000::/52) are 779 augmented with entries from the third table in Figure 5 (the table of 780 unscoped or scoped for ::/0 forwarding entries). The first four 781 unscoped forwarding entries (D=2001:db8:0:a010::/64, 782 D=2001:db8:0:b010::/64, D=2001:db8:0:a020::/64, and 783 D=2001:db8:0:b020::/64) are not an exact match for any of the 784 existing entries in the forwarding table for source prefix 785 2001:db8:0:a000::/52. Therefore, these four entries are added to the 786 final forwarding table for source prefix 2001:db8:0:a000::/52. The 787 result of adding these entries is reflected in first four entries the 788 first table in Figure 6. 790 The next less specific scoped (scope is ::/0) forwarding table entry 791 is for D=2001:db8:0:5555::/64. This entry is an exact match for the 792 existing entry in the forwarding table for the more specific source 793 prefix 2001:db8:0:a000::/52. Therefore, we do not replace the 794 existing entry with the entry from the unscoped forwarding table. 795 This is reflected in the fifth entry in the first table in Figure 6. 796 (Note that since both scoped and unscoped entries have R7 as the next 797 hop, the result of applying this rule is not visible.) 799 The next less specific prefix scoped (scope is ::/0) forwarding table 800 entry is for D=2001:db8:0:6666::/64. This entry is not an exact 801 match for any existing entries in the forwarding table for source 802 prefix 2001:db8:0:a000::/52. Therefore, we add this entry. This is 803 reflected in the sixth entry in the first table in Figure 6. 805 The next less specific prefix scoped (scope is ::/0) forwarding table 806 entry is for D=::/0. This entry is an exact match for the existing 807 entry in the forwarding table for more specific source prefix 808 2001:db8:0:a000::/52. Therefore, we do not overwrite the existing 809 source-prefix-scoped entry, as can be seen in the last entry in the 810 first table in Figure 6. 812 if source address matches 2001:db8:0:a000::/52 813 then use this forwarding table 814 ============================================ 815 D=2001:db8:0:a010::/64 NH=R2 816 D=2001:db8:0:b010::/64 NH=R2 817 D=2001:db8:0:a020::/64 NH=R5 818 D=2001:db8:0:b020::/64 NH=R5 819 D=2001:db8:0:5555::/64 NH=R7 820 D=2001:db8:0:6666::/64 NH=SERb2 821 D=::/0 NH=R7 823 else if source address matches 2001:db8:0:b000::/52 824 then use this forwarding table 825 ============================================ 826 D=2001:db8:0:a010::/64 NH=R2 827 D=2001:db8:0:b010::/64 NH=R2 828 D=2001:db8:0:a020::/64 NH=R5 829 D=2001:db8:0:b020::/64 NH=R5 830 D=2001:db8:0:5555::/64 NH=R7 831 D=2001:db8:0:6666::/64 NH=SERb2 832 D=::/0 NH=SERb1 834 else if source address matches ::/0 use this forwarding table 835 ============================================ 836 D=2001:db8:0:a010::/64 NH=R2 837 D=2001:db8:0:b010::/64 NH=R2 838 D=2001:db8:0:a020::/64 NH=R5 839 D=2001:db8:0:b020::/64 NH=R5 840 D=2001:db8:0:5555::/64 NH=R7 841 D=2001:db8:0:6666::/64 NH=SERb2 842 D=::/0 NH=SERb1 844 Figure 6: Complete Forwarding Tables Computed at R8 846 The forwarding tables produced by this process at R8 have the desired 847 properties. A packet with a source address in 2001:db8:0:a000::/52 848 will be forwarded based on the first table in Figure 6. If the 849 packet is destined for the Internet at large or the service at 850 D=2001:db8:0:5555/64, it will be sent to R7 in the direction of SERa. 851 If the packet is destined for an internal host, then the first four 852 entries will send it to R2 or R5 as expected. Note that if this 853 packet has a destination address corresponding to the service offered 854 by ISP-B (D=2001:db8:0:5555::/64), then it will get forwarded to 855 SERb2. It will be dropped by SERb2 or by ISP-B, since it the packet 856 has a source address that was not assigned by ISP-B. However, this 857 is expected behavior. In order to use the service offered by ISP-B, 858 the host needs to originate the packet with a source address assigned 859 by ISP-B. 861 In this example, a packet with a source address that doesn't match 862 2001:db8:0:a000::/52 or 2001:db8:0:b000::/52 must have originated 863 from an external host. Such a packet will use the unscoped 864 forwarding table (the last table in Figure 6). These packets will 865 flow exactly as they would in absence of multihoming. 867 We can also modify this example to illustrate how it supports 868 deployments where not all routers in the site support SADR. 869 Continuing with the topology shown in Figure 3, suppose that R3 and 870 R5 do not support SADR. Instead they are only capable of 871 understanding unscoped route advertisements. The SADR routers in the 872 network will still originate the routes shown in Figure 4. However, 873 R3 and R5 will only understand the unscoped routes as shown in 874 Figure 7. 876 Routes originated by SERa: 877 (D=2001:db8:0:5555::/64) 878 (D=::/0) 880 Routes originated by SERb1: 881 (D=::/0) 883 Routes originated by SERb2: 884 (D=2001:db8:0:6666::/64) 886 Routes originated by R1: 887 (D=2001:db8:0:a010::/64) 888 (D=2001:db8:0:b010::/64) 890 Routes originated by R2: 891 (D=2001:db8:0:a010::/64) 892 (D=2001:db8:0:b010::/64) 894 Routes originated by R3: 895 (D=2001:db8:0:a020::/64) 896 (D=2001:db8:0:b020::/64) 898 Figure 7: Routes Advertisements Understood by Routers that do no 899 Support SADR 901 With these unscoped route advertisements, R5 will produce the 902 forwarding table shown in Figure 8. 904 forwarding table 905 ============================================ 906 D=2001:db8:0:a010::/64 NH=R8 907 D=2001:db8:0:b010::/64 NH=R8 908 D=2001:db8:0:a020::/64 NH=R3 909 D=2001:db8:0:b020::/64 NH=R3 910 D=2001:db8:0:5555::/64 NH=R8 911 D=2001:db8:0:6666::/64 NH=SERb2 912 D=::/0 NH=R8 914 Figure 8: Forwarding Table For R5, Which Doesn't Understand Source- 915 Prefix-Scoped Routes 917 Any traffic that needs to exit the site will eventually hit a SADR- 918 capable router. Once that traffic enters the SADR-capable domain, 919 then it will not leave that domain until it exits the site. This 920 property is required in order to guarantee that there will not be 921 routing loops involving SADR-capable and non-SADR-capable routers. 923 Note that the mechanism described here for converting source-prefix- 924 scoped destination prefix routing advertisements into forwarding 925 state is somewhat different from that proposed in 926 [I-D.ietf-rtgwg-dst-src-routing]. The method described in this 927 document is intended to be easy to understand for network enterprise 928 operators while at the same time being functionally correct. Another 929 difference is that the method in this document assumes that source 930 prefix will not overlap. Other differences between the two 931 approaches still need to be understood and reconciled. 933 An interesting side-effect of deploying SADR is if all routers in a 934 given network support SADR and have a scoped forwarding table, then 935 the unscoped forwarding table can be eliminated which ensures that 936 packets with legitimate source addresses only can leave the network 937 (as there are no scoped forwarding tables for spoofed/bogon source 938 addresses). It would prevent accidental leaks of ULA/reserved/link- 939 local sources to the Internet as well as ensures that no spoofing is 940 possible from the SADR-enabled network. 942 5. Mechanisms For Hosts To Choose Good Source Addresses In A Multihomed 943 Site 945 Until this point, we have made the assumption that hosts are able to 946 choose the correct source address using some unspecified mechanism. 947 This has allowed us to just focus on what the routers in a multihomed 948 site network need to do in order to forward packets to the correct 949 ISP based on source address. Now we look at possible mechanisms for 950 hosts to choose the correct source address. We also look at what 951 role, if any, the routers may play in providing information that 952 helps hosts to choose source addresses. 954 Any host that needs to be able to send traffic using the uplinks to a 955 given ISP is expected to be configured with an address from the 956 prefix assigned by that ISP. The host will control which ISP is used 957 for its traffic by selecting one of the addresses configured on the 958 host as the source address for outgoing traffic. It is the 959 responsibility of the site network to ensure that a packet with the 960 source address from an ISP is now sent on an uplink to that ISP. 962 If all of the ISP uplinks are working, the choice of source address 963 by the host may be driven by the desire to load share across ISP 964 uplinks, or it may be driven by the desire to take advantage of 965 certain properties of a particular uplink or ISP. If any of the ISP 966 uplinks is not working, then the choice of source address by the host 967 can determine if packets get dropped. 969 How a host should make good decisions about source address selection 970 in a multihomed site is not a solved problem. We do not attempt to 971 solve this problem in this document. Instead we discuss the current 972 state of affairs with respect to standardized solutions and 973 implementation of those solutions. We also look at proposed 974 solutions for this problem. 976 An external host initiating communication with a host internal to a 977 PA multihomed site will need to know multiple addresses for that host 978 in order to communicate with it using different ISPs to the 979 multihomed site. These addresses are typically learned through DNS. 980 (For simplicity, we assume that the external host is single-homed.) 981 The external host chooses the ISP that will be used at the remote 982 multihomed site by setting the destination address on the packets it 983 transmits. For a sessions originated from an external host to an 984 internal host, the choice of source address used by the internal host 985 is simple. The internal host has no choice but to use the 986 destination address in the received packet as the source address of 987 the transmitted packet. 989 For a session originated by a host internal to the multi-homed site, 990 the decision of what source address to select is more complicated. 991 We consider three main methods for hosts to get information about the 992 network. The two proactive methods are Neighbor Discovery Router 993 Advertisements and DHCPv6. The one reactive method we consider is 994 ICMPv6. Note that we are explicitly excluding the possibility of 995 having hosts participate in or even listen directly to routing 996 protocol advertisements. 998 First we look at how a host is currently expected to select the 999 source and destination address with which it sends a packet. 1001 5.1. Source Address Selection Algorithm on Hosts 1003 [RFC6724] defines the algorithms that hosts are expected to use to 1004 select source and destination addresses for packets. It defines an 1005 algorithm for selecting a source address and a separate algorithm for 1006 selecting a destination address. Both of these algorithms depend on 1007 a policy table. [RFC6724] defines a default policy which produces 1008 certain behavior. 1010 The rules in the two algorithms in [RFC6724] depend on many different 1011 properties of addresses. While these are needed for understanding 1012 how a host should choose addresses in an arbitrary environment, most 1013 of the rules are not relevant for understanding how a host should 1014 choose among multiple source addresses in mulitihomed envinronment 1015 when sending a packet to a remote host. Returning to the example in 1016 Figure 3, we look at what the default algorithms in [RFC6724] say 1017 about the source address that internal host H31 should use to send 1018 traffic to external host H101, somewhere on the Internet. Let's look 1019 at what rules in [RFC6724] are actually used by H31 in this case. 1021 There is no choice to be made with respect to destination address. 1022 H31 needs to send a packet with D=2001:db8:0:1234::101 in order to 1023 reach H101. So H31 have to choose between using 1024 S=2001:db8:0:a010::31 or S=2001:db8:0:b010::31 as the source address 1025 for this packet. We go through the rules for source address 1026 selection in Section 5 of [RFC6724]. Rule 1 (Prefer same address) is 1027 not useful to break the tie between source addresses, because neither 1028 the candidate source addresses equals the destination address. Rule 1029 2 (Prefer appropriate scope) is also not used in this scenario, 1030 because both source addresses and the destination address have global 1031 scope. 1033 Rule 3 (Avoid deprecated addresses) applies to an address that has 1034 been autoconfigured by a host using stateless address 1035 autoconfiguration as defined in [RFC4862]. An address autoconfigured 1036 by a host has a preferred lifetime and a valid lifetime. The address 1037 is preferred until the preferred lifetime expires, after which it 1038 becomes deprecated. A deprecated address is not used if there is a 1039 preferred address of the appropriate scope available. When the valid 1040 lifetime expires, the address cannot be used at all. The preferred 1041 and valid lifetimes for an autoconfigured address are set based on 1042 the corresponding lifetimes in the Prefix Information Option in 1043 Neighbor Discovery Router Advertisements. So a possible tool to 1044 control source address selection in this scenario would be for a host 1045 to make an address deprecated by having routers on that link, R1 and 1046 R2 in Figure 3, send a Router Advertisement message contaning a 1047 Prefix Information Option for the source prefix to be discouraged (or 1048 prohibited) with the preferred lifetime set to zero. This is a 1049 rather blunt tool, because it discourages or prohibits the use of 1050 that source prefix for all destinations. However, it may be useful 1051 in some scenarios. For example, if all uplinks to a particular ISP 1052 fail, it is desirable to prevent hosts from using source addresses 1053 from that ISP address space. 1055 Rule 4 (Avoid home addresses) does not apply here because we are not 1056 considering Mobile IP. 1058 Rule 5 (Prefer outgoing interface) is not useful in this scenario, 1059 because both source addresses are assigned to the same interface. 1061 Rule 5.5 (Prefer addresses in a prefix advertised by the next-hop) is 1062 not useful in the scenario when both R1 and R2 will advertise both 1063 source prefixes. However potentially this rule may allow a host to 1064 select the correct source prefix by selecting a next-hop. The most 1065 obvious way would be to make R1 to advertise itself as a default 1066 router and send PIO for 2001:db8:0:a010::/64, while R2 is advertising 1067 itself as a default router and sending PIO for 2001:db8:0:b010::/64. 1068 We'll discuss later how Rule 5.5 can be used to influence a source 1069 address selection in single-router topologies (e.g. when H41 is 1070 sending traffic using R3 as a default gateway). 1072 Rule 6 (Prefer matching label) refers to the Label value determined 1073 for each source and destination prefix as a result of applying the 1074 policy table to the prefix. With the default policy table defined in 1075 Section 2.1 of [RFC6724], Label(2001:db8:0:a010::31) = 5, 1076 Label(2001:db8:0:b010::31) = 5, and Label(2001:db8:0:1234::101) = 5. 1077 So with the default policy, Rule 6 does not break the tie. However, 1078 the algorithms in [RFC6724] are defined in such as way that non- 1079 default address selection policy tables can be used. [RFC7078] 1080 defines a way to distribute a non-default address selection policy 1081 table to hosts using DHCPv6. So even though the application of rule 1082 6 to this scenario using the default policy table is not useful, rule 1083 6 may still be a useful tool. 1085 Rule 7 (Prefer temporary addresses) has to do with the technique 1086 described in [RFC4941] to periodically randomize the interface 1087 portion of an IPv6 address that has been generated using stateless 1088 address autoconfiguration. In general, if H31 were using this 1089 technique, it would use it for both source addresses, for example 1090 creating temporary addresses 2001:db8:0:a010:2839:9938:ab58:830f and 1091 2001:db8:0:b010:4838:f483:8384:3208, in addition to 1092 2001:db8:0:a010::31 and 2001:db8:0:b010::31. So this rule would 1093 prefer the two temporary addresses, but it would not break the tie 1094 between the two source prefixes from ISP-A and ISP-B. 1096 Rule 8 (Use longest matching prefix) dictates that between two 1097 candidate source addresses the one which has longest common prefix 1098 length with the destination address. For example, if H31 were 1099 selecting the source address for sending packets to H101, this rule 1100 would not be a tie breaker as for both candidate source addresses 1101 2001:db8:0:a101::31 and 2001:db8:0:b101::31 the common prefix length 1102 with the destination is 48. However if H31 were selecting the source 1103 address for sending packets H41 address 2001:db8:0:a020::41, then 1104 this rule would result in using 2001:db8:0:a101::31 as a source 1105 (2001:db8:0:a101::31 and 2001:db8:0:a020::41 share the common prefix 1106 2001:db8:0:a000::/58, while for `2001:db8:0:b101::31 and 1107 2001:db8:0:a020::41 the common prefix is 2001:db8:0:a000::/51). 1108 Therefore rule 8 might be useful for selecting the correct source 1109 address in some but not all scenarios (for example if ISP-B services 1110 belong to 2001:db8:0:b000::/59 then H31 would always use 1111 2001:db8:0:b010::31 to access those destinations). 1113 So we can see that of the 8 source selection address rules from 1114 [RFC6724], five actually apply to our basic site multihoming 1115 scenario. The rules that are relevant to this scenario are 1116 summarized below. 1118 o Rule 3: Avoid deprecated addresses. 1120 o Rule 5.5: Prefer addresses in a prefix advertised by the next-hop. 1122 o Rule 6: Prefer matching label. 1124 o Rule 8: Prefer longest matching prefix. 1126 The two methods that we discuss for controlling the source address 1127 selection through the four relevant rules above are SLAAC Router 1128 Advertisement messages and DHCPv6. 1130 We also consider a possible role for ICMPv6 for getting traffic- 1131 driven feedback from the network. With the source address selection 1132 algorithm discussed above, the goal is to choose the correct source 1133 address on the first try, before any traffic is sent. However, 1134 another strategy is to choose a source address, send the packet, get 1135 feedback from the network about whether or not the source address is 1136 correct, and try another source address if it is not. 1138 We consider four scenarios where a host needs to select the correct 1139 source address. The first is when both uplinks are working. The 1140 second is when one uplink has failed. The third one is a situation 1141 when one failed uplink has recovered. The last one is failure of 1142 both (all) uplinks. 1144 5.2. Selecting Source Address When Both Uplinks Are Working 1146 Again we return to the topology in Figure 3. Suppose that the site 1147 administrator wants to implement a policy by which all hosts need to 1148 use ISP-A to reach H01 at D=2001:db8:0:1234::101. So for example, 1149 H31 needs to select S=2001:db8:0:a010::31. 1151 5.2.1. Distributing Address Selection Policy Table with DHCPv6 1153 This policy can be implemented by using DHCPv6 to distribute an 1154 address selection policy table that assigns the same label to 1155 destination address that match 2001:db8:0:1234::/64 as it does to 1156 source addresses that match 2001:db8:0:a000::/52. The following two 1157 entries accomplish this. 1159 Prefix Precedence Label 1160 2001:db8:0:1234::/64 50 33 1161 2001:db8:0:a000::/52 50 33 1163 Figure 9: Policy table entries to implement a routing policy 1165 This requires that the hosts implement [RFC6724], the basic source 1166 and destination address framework, along with [RFC7078], the DHCPv6 1167 extension for distributing a non-default policy table. Note that it 1168 does NOT require that the hosts use DHCPv6 for address assignment. 1169 The hosts could still use stateless address autoconfiguration for 1170 address configuration, while using DHCPv6 only for policy table 1171 distribution (see [RFC3736]). However this method has a number of 1172 disadvantages: 1174 o DHCPv6 support is not a mandatory requirement for IPv6 hosts, so 1175 this method might not work for all devices. 1177 o Network administrators are required to explicitly configure the 1178 desired network access policies on DHCPv6 servers. While it might 1179 be feasible in the scenarion of a single multihomed network, such 1180 approach might have some scalability issues, especially if the 1181 centralized DHCPv6 solution is deployed to serve a large number of 1182 multiomed sites. 1184 5.2.2. Controlling Source Address Selection With Router Advertisements 1186 Neighbor Discovery currently has two mechanisms to communicate prefix 1187 information to hosts. The base specification for Neighbor Discovery 1188 (see [RFC4861]) defines the Prefix Information Option (PIO) in the 1189 Router Advertisement (RA) message. When a host receives a PIO with 1190 the A-flag set, it can use the prefix in the PIO as source prefix 1191 from which it assigns itself an IP address using stateless address 1192 autoconfiguration (SLAAC) procedures described in [RFC4862]. In the 1193 example of Figure 3, if the site network is using SLAAC, we would 1194 expect both R1 and R2 to send RA messages with PIOs for both source 1195 prefixes 2001:db8:0:a010::/64 and 2001:db8:0:b010::/64 with the 1196 A-flag set. H31 would then use the SLAAC procedure to configure 1197 itself with the 2001:db8:0:a010::31 and 2001:db8:0:b010::31. 1199 Whereas a host learns about source prefixes from PIO messages, hosts 1200 can learn about a destination prefix from a Router Advertisement 1201 containing Route Information Option (RIO), as specified in [RFC4191]. 1202 The destination prefixes in RIOs are intended to allow a host to 1203 choose the router that it uses as its first hop to reach a particular 1204 destination prefix. 1206 As currently standardized, neither PIO nor RIO options contained in 1207 Neighbor Discovery Router Advertisements can communicate the 1208 information needed to implement the desired routing policy. PIO's 1209 communicate source prefixes, and RIO communicate destination 1210 prefixes. However, there is currently no standardized way to 1211 directly associate a particular destination prefix with a particular 1212 source prefix. 1214 [I-D.pfister-6man-sadr-ra] proposes a Source Address Dependent Route 1215 Information option for Neighbor Discovery Router Advertisements which 1216 would associate a source prefix and with a destination prefix. The 1217 details of [I-D.pfister-6man-sadr-ra] might need tweaking to address 1218 this use case. However, in order to be able to use Neighbor 1219 Discovery Router Advertisements to implement this routing policy, an 1220 extension that allows a R1 and R2 to explicitly communicate to H31 an 1221 association between S=2001:db8:0:a000::/52 D=2001:db8:0:1234::/64 1222 would be needed. 1224 However, Rule 5.5 of the source address selection algorithm 1225 (discussed in Section 5.1 above), together with default router 1226 preference (specified in [RFC4191]) and RIO can be used to influence 1227 a source address selection on a host as described below. Let's look 1228 at source address selection on the host H41. It receives RAs from R3 1229 with PIOs for 2001:db8:0:a020::/64 and 2001:db8:0:b020::/64. At that 1230 point all traffic would use the same next-hop (R3 link-local address) 1231 so Rule 5.5 does not apply. Now let's assume that R3 supports SADR 1232 and has two scoped forwarding tables, one scoped to 1233 S=2001:db8:0:a000::/52 and another scoped to S=2001:db8:0:b000::/52. 1234 If R3 generates two different link-local addresses for its interface 1235 facing H41 (one for each scoped forwarding table, LLA_A and LLA_B) 1236 and starts sending two different RAs: one is sent from LLA_A and 1237 includes PIO for 2001:db8:0:a020::/64, another us sent from LLA_B and 1238 includes PIO for 2001:db8:0:b020::/64. Now it is possible to 1239 influence H41 source address selection for destinations which follow 1240 the default route by setting default router preference in RAs. If it 1241 is desired that H41 reaches H101 (or any destinations in the 1242 Internet) via ISP-A, then RAs sent from LLA_A should have default 1243 router preference set to 01 (high priority), while RAs sent from 1244 LLA_B should have preference set to 11 (low). Then LLA_A would be 1245 chosen as a next-hop for H101 and therefore (as per rule 5.5) 1246 2001:db8:0:a020::41 would be selected as the source address. If, at 1247 the same time, it is desired that H61 is accessible via ISP-B then R3 1248 should include a RIO for 2001:db8:0:6666::/64 to its RA sent from 1249 LLA_B. H41 would chose LLA_B as a next-hop for all traffic to H61 1250 and then as per Rule 5.5, 2001:db8:0:b020::41 would be selected as a 1251 source address. 1253 If in the above mentioned scenario it is desirable that all Internet 1254 traffic leaves the network via ISP-A and the link to ISP-B is used 1255 for accessing ISP-B services only (not as ISP-A link backup), then 1256 RAs sent by R3 from LLA_B should have Router Lifetime set to 0 and 1257 should include RIOs for ISP-B address space. It would instruct H41 1258 to use LLA_A for all Internet traffic but use LLA_B as a next-hop 1259 while sending traffic to ISP-B addresses. 1261 The description of the mechanism above assumes SADR support by the 1262 first-hop routers as well as SERs. However, a first-hop router can 1263 still provide a less flexible version of this mechanism even without 1264 implementing SADR. This could be done by providing configuration 1265 knobs on the first-hop router that allow it to generate different 1266 link-local addresses and to send individual RAs for each prefix. 1268 The mechanism described above relies on Rule 5.5 of the default 1269 source address selection algorithm defined in [RFC6724]. [RFC8028] 1270 recommends that a host SHOULD select default routers for each prefix 1271 in which it is assigned an address. It also recommends that hosts 1272 SHOULD implement Rule 5.5. of [RFC6724]. Hosts following the 1273 recommendations specified in [RFC8028] therefore should be able to 1274 benefit from the solution described in this document. No standards 1275 need to be updated in regards to host behavior. 1277 5.2.3. Controlling Source Address Selection With ICMPv6 1279 We now discuss how one might use ICMPv6 to implement the routing 1280 policy to send traffic destined for H101 out the uplink to ISP-A, 1281 even when uplinks to both ISPs are working. If H31 started sending 1282 traffic to H101 with S=2001:db8:0:b010::31 and 1283 D=2001:db8:0:1234::101, it would be routed through SER-b1 and out the 1284 uplink to ISP-B. SERb1 could recognize that this is traffic is not 1285 following the desired routing policy and react by sending an ICMPv6 1286 message back to H31. 1288 In this example, we could arrange things so that SERb1 drops the 1289 packet with S=2001:db8:0:b010::31 and D=2001:db8:0:1234::101, and 1290 then sends to H31 an ICMPv6 Destination Unreachable message with Code 1291 5 (Source address failed ingress/egress policy). When H31 receives 1292 this packet, it would then be expected to try another source address 1293 to reach the destination. In this example, H31 would then send a 1294 packet with S=2001:db8:0:a010::31 and D=2001:db8:0:1234::101, which 1295 will reach SERa and be forwarded out the uplink to ISP-A. 1297 However, we would also want it to be the case that SERb1 does not 1298 enforce this routing policy when the uplink from SERa to ISP-A has 1299 failed. This could be accomplished by having SERa originate a 1300 source-prefix-scoped route for (S=2001:db8:0:a000::/52, 1301 D=2001:db8:0:1234::/64) and have SERb1 monitor the presence of that 1302 route. If that route is not present (because SERa has stopped 1303 originating it), then SERb1 will not enforce the routing policy, and 1304 it will forward packets with S=2001:db8:0:b010::31 and 1305 D=2001:db8:0:1234::101 out its uplink to ISP-B. 1307 We can also use this source-prefix-scoped route originated by SERa to 1308 communicate the desired routing policy to SERb1. We can define an 1309 EXCLUSIVE flag to be advertised together with the IGP route for 1310 (S=2001:db8:0:a000::/52, D=2001:db8:0:1234::/64). This would allow 1311 SERa to communicate to SERb that SERb should reject traffic for 1312 D=2001:db8:0:1234::/64 and respond with an ICMPv6 Destination 1313 Unreachable Code 5 message, as long as the route for 1314 (S=2001:db8:0:a000::/52, D=2001:db8:0:1234::/64) is present. 1316 Finally, if we are willing to extend ICMPv6 to support this solution, 1317 then we could create a mechanism for SERb1 to tell the host what 1318 source address it should be using to successfully forward packets 1319 that meet the policy. In its current form, when SERb1 sends an 1320 ICMPv6 Destination Unreachable Code 5 message, it is basically 1321 saying, "This source address is wrong. Try another source address." 1322 In the absence of a clear indication which address to try next, the 1323 host will iterate over all addresses assigned to the interface (e.g. 1324 various privacy addresses) which would lead to significant delays and 1325 degraded user experience. It would be better is if the ICMPv6 1326 message could say, "This source address is wrong. Instead use a 1327 source address in S=2001:db8:0:a000::/52.". 1329 However using ICMPv6 for signalling source address information back 1330 to hosts introduces new challenges. Most routers currently have 1331 software or hardware limits on generating ICMP messages. An site 1332 administrator deploying a solution that relies on the SERs generating 1333 ICMP messages could try to improve the performance of SERs for 1334 generating ICMP messages. However, in a large network, it is still 1335 likely that ICMP message generation limits will be reached. As a 1336 result hosts would not receive ICMPv6 back which in turn leads to 1337 traffic blackholing and poor user experience. To improve the 1338 scalability of ICMPv6-based signalling hosts SHOULD cache the 1339 preferred source address (or prefix) for the given destination (which 1340 in turn might cause issues in case of the corresponding ISP uplinks 1341 failure - see Section 5.3). In addition, the same source prefix 1342 SHOULD be used for other destinations in the same /64 as the original 1343 destination address. The source prefix SHOULD have a specific 1344 lifetime. Expiration of the lifetime SHOULD trigger the source 1345 address selection algorithm again. 1347 Using ICMPv6 Code 5 message for influencing source address selection 1348 allows an attacker to exhaust the list of candidate source addresses 1349 on the host by sending spoofed ICMPv6 Code 5 for all prefixes known 1350 on the network (therefore preventing a victim from establishing a 1351 communication with the destination host). To protect from such 1352 attack hosts SHOULD verify that the original packet header included 1353 into ICMPv6 error message was actually sent by the host. 1355 As currently standardized in [RFC4443], the ICMPv6 Destination 1356 Unreachable Message with Code 5 would allow for the iterative 1357 approach to retransmitting packets using different source addresses. 1358 As currently defined, the ICMPv6 message does not provide a mechanism 1359 to communication information about which source prefix should be used 1360 for a retransmitted packet. The current document does not define 1361 such a mechanism. However, we note that this might be a useful 1362 extension to define in a different document. 1364 5.2.4. Summary of Methods For Controlling Source Address Selection To 1365 Implement Routing Policy 1367 So to summarize this section, we have looked at three methods for 1368 implementing a simple routing policy where all traffic for a given 1369 destination on the Internet needs to use a particular ISP, even when 1370 the uplinks to both ISPs are working. 1372 The default source address selection policy cannot distinguish 1373 between the source addresses needed to enforce this policy, so a non- 1374 default policy table using associating source and destination 1375 prefixes using Label values would need to be installed on each host. 1376 A mechanism exists for DHCPv6 to distribute a non-default policy 1377 table but such solution would heavily rely on DHCPv6 support by host 1378 operating system. Moreover there is no mechanism to translate 1379 desired routing/traffic engineering policies into policy tables on 1380 DHCPv6 servers. Therefore using DHCPv6 for controlling address 1381 selection policy table is not recommended and SHOULD NOT be used. 1383 At the same time Router Advertisements provide a reliable mechanism 1384 to influence source address selection process via PIO, RIO and 1385 default router preferences. As all those options have been 1386 standardized by IETF and are supported by various operating systems, 1387 no changes are required on hosts. First-hop routers in the 1388 enterprise network need to be able of sending different RAs for 1389 different SLAAC prefixes (either based on scoped forwarding tables or 1390 based on pre-configured policies). 1392 SERs can enforce the routing policy by sending ICMPv6 Destination 1393 Unreachable messages with Code 5 (Source address failed ingress/ 1394 egress policy) for traffic that is being sent with the wrong source 1395 address. The policy distribution can be automated by defining an 1396 EXCLUSIVE flag for the source-prefix-scoped route which can be set on 1397 the SER that originates the route. As ICMPv6 message generation can 1398 be rate-limited on routers, it SHOULD NOT be used as the only 1399 mechanism to influence source address selection on hosts. While 1400 hosts SHOULD select the correct source address for a given 1401 destination the network SHOULD signal any source address issues back 1402 to hosts using ICMPv6 error messages. 1404 5.3. Selecting Source Address When One Uplink Has Failed 1406 Now we discuss if DHCPv6, Neighbor Discovery Router Advertisements, 1407 and ICMPv6 can help a host choose the right source address when an 1408 uplink to one of the ISPs has failed. Again we look at the scenario 1409 in Figure 3. This time we look at traffic from H31 destined for 1410 external host H501 at D=2001:db8:0:5678::501. We initially assume 1411 that the uplink from SERa to ISP-A is working and that the uplink 1412 from SERb1 to ISP-B is working. 1414 We assume there is no particular routing policy desired, so H31 is 1415 free to send packets with S=2001:db8:0:a010::31 or 1416 S=2001:db8:0:b010::31 and have them delivered to H501. For this 1417 example, we assume that H31 has chosen S=2001:db8:0:b010::31 so that 1418 the packets exit via SERb to ISP-B. Now we see what happens when the 1419 link from SERb1 to ISP-B fails. How should H31 learn that it needs 1420 to start sending the packet to H501 with S=2001:db8:0:a010::31 in 1421 order to start using the uplink to ISP-A? We need to do this in a 1422 way that doesn't prevent H31 from still sending packets with 1423 S=2001:db8:0:b010::31 in order to reach H61 at D=2001:db8:0:6666::61. 1425 5.3.1. Controlling Source Address Selection With DHCPv6 1427 For this example we assume that the site network in Figure 3 has a 1428 centralized DHCP server and all routers act as DHCP relay agents. We 1429 assume that both of the addresses assigned to H31 were assigned via 1430 DHCP. 1432 We could try to have the DHCP server monitor the state of the uplink 1433 from SERb1 to ISP-B in some manner and then tell H31 that it can no 1434 longer use S=2001:db8:0:b010::31 by settings its valid lifetime to 1435 zero. The DHCP server could initiate this process by sending a 1436 Reconfigure Message to H31 as described in Section 19 of [RFC3315]. 1437 Or the DHCP server can assign addresses with short lifetimes in order 1438 to force clients to renew them often. 1440 This approach would prevent H31 from using S=2001:db8:0:b010::31 to 1441 reach the a host on the Internet. However, it would also prevent H31 1442 from using S=2001:db8:0:b010::31 to reach H61 at 1443 D=2001:db8:0:6666::61, which is not desirable. 1445 Another potential approach is to have the DHCP server monitor the 1446 uplink from SERb1 to ISP-B and control the choice of source address 1447 on H31 by updating its address selection policy table via the 1448 mechanism in [RFC7078]. The DHCP server could initiate this process 1449 by sending a Reconfigure Message to H31. Note that [RFC3315] 1450 requires that Reconfigure Message use DHCP authentication. DHCP 1451 authentication could be avoided by using short address lifetimes to 1452 force clients to send Renew messages to the server often. If the 1453 host is not obtaining its IP addresses from the DHCP server, then it 1454 would need to use the Information Refresh Time option defined in 1455 [RFC4242]. 1457 If the following policy table can be installed on H31 after the 1458 failure of the uplink from SERb1, then the desired routing behavior 1459 should be achieved based on source and destination prefix being 1460 matched with label values. 1462 Prefix Precedence Label 1463 ::/0 50 44 1464 2001:db8:0:a000::/52 50 44 1465 2001:db8:0:6666::/64 50 55 1466 2001:db8:0:b000::/52 50 55 1468 Figure 10: Policy Table Needed On Failure Of Uplink From SERb1 1470 The described solution has a number of significant drawbacks, some of 1471 them already discussed in Section 5.2.1. 1473 o DHCPv6 support is not required for an IPv6 host and there are 1474 operating systems which do not support DHCPv6. Besides that, it 1475 does not appear that [RFC7078] has been widely implemented on host 1476 operating systems. 1478 o [RFC7078] does not clearly specify this kind of a dynamic use case 1479 where address selection policy needs to be updated quickly in 1480 response to the failure of a link. In a large network it would 1481 present scalability issues as many hosts need to be reconfigured 1482 in very short period of time. 1484 o Updating DHCPv6 server configuration each time an ISP uplink 1485 changes its state introduces some scalability issues, especially 1486 for mid/large distributed scale enterprise networks. In addition 1487 to that, the policy table needs to be manually configured by 1488 administrators which makes that solution prone to human error. 1490 o No mechanism exists for making DHCPv6 servers aware of network 1491 topology/routing changes in the network. In general DHCPv6 1492 servers monitoring network-related events sounds like a bad idea 1493 as completely new functionality beyond the scope of DHCPv6 role is 1494 required. 1496 5.3.2. Controlling Source Address Selection With Router Advertisements 1498 The same mechanism as discussed in Section 5.2.2 can be used to 1499 control the source address selection in the case of an uplink 1500 failure. If a particular prefix should not be used as a source for 1501 any destinations, then the router needs to send RA with Preferred 1502 Lifetime field for that prefix set to 0. 1504 Let's consider a scenario when all uplinks are operational and H41 1505 receives two different RAs from R3: one from LLA_A with PIO for 1506 2001:db8:0:a020::/64, default router preference set to 11 (low) and 1507 another one from LLA_B with PIO for 2001:db8:0:a020::/64, default 1508 router preference set to 01 (high) and RIO for 2001:db8:0:6666::/64. 1509 As a result H41 is using 2001:db8:0:b020::41 as a source address for 1510 all Internet traffic and those packets are sent by SERs to ISP-B. If 1511 SERb1 uplink to ISP-B failed, the desired behavior is that H41 stops 1512 using 2001:db8:0:b020::41 as a source address for all destinations 1513 but H61. To achieve that R3 should react to SERb1 uplink failure 1514 (which could be detected as the scoped route (S=2001:db8:0:b000::/52, 1515 D=::/0) disappearance) by withdrawing itself as a default router. R3 1516 sends a new RA from LLA_B with Router Lifetime value set to 0 (which 1517 means that it should not be used as default router). That RA still 1518 contains PIO for 2001:db8:0:b020::/64 (for SLAAC purposes) and RIO 1519 for 2001:db8:0:6666::/64 so H41 can reach H61 using LLA_B as a next- 1520 hop and 2001:db8:0:b020::41 as a source address. For all traffic 1521 following the default route, LLA_A will be used as a next-hop and 1522 2001:db8:0:a020::41 as a source address. 1524 If all uplinks to ISP-B have failed and therefore source addresses 1525 from ISP-B address space should not be used at all, the forwarding 1526 table scoped S=2001:db8:0:b000::/52 contains no entries. Hosts can 1527 be instructed to stop using source addresses from that block by 1528 sending RAs containing PIO with Preferred Lifetime set to 0. 1530 5.3.3. Controlling Source Address Selection With ICMPv6 1532 Now we look at how ICMPv6 messages can provide information back to 1533 H31. We assume again that at the time of the failure H31 is sending 1534 packets to H501 using (S=2001:db8:0:b010::31, 1535 D=2001:db8:0:5678::501). When the uplink from SERb1 to ISP-B fails, 1536 SERb1 would stop originating its source-prefix-scoped route for the 1537 default destination (S=2001:db8:0:b000::/52, D=::/0) as well as its 1538 unscoped default destination route. With these routes no longer in 1539 the IGP, traffic with (S=2001:db8:0:b010::31, D=2001:db8:0:5678::501) 1540 would end up at SERa based on the unscoped default destination route 1541 being originated by SERa. Since that traffic has the wrong source 1542 address to be forwarded to ISP-A, SERa would drop it and send a 1543 Destination Unreachable message with Code 5 (Source address failed 1544 ingress/egress policy) back to H31. H31 would then know to use 1545 another source address for that destination and would try with 1546 (S=2001:db8:0:a010::31, D=2001:db8:0:5678::501). This would be 1547 forwarded to SERa based on the source-prefix-scoped default 1548 destination route still being originated by SERa, and SERa would 1549 forward it to ISP-A. As discussed above, if we are willing to extend 1550 ICMPv6, SERa can even tell H31 what source address it should use to 1551 reach that destination. The expected host behaviour has been 1552 discussed in Section 5.2.3. Potential issue with using ICMPv6 for 1553 signalling source address issues back to hosts is that uplink to an 1554 ISP-B failure immediately invalidates source addresses from 1555 2001:db8:0:b000::/52 for all hosts which triggers a large number of 1556 ICMPv6 being sent back to hosts - the same scalability/rate limiting 1557 issues discussed in Section 5.2.3 would apply. 1559 5.3.4. Summary Of Methods For Controlling Source Address Selection On 1560 The Failure Of An Uplink 1562 It appears that DHCPv6 is not particularly well suited to quickly 1563 changing the source address used by a host in the event of the 1564 failure of an uplink, which eliminates DHCPv6 from the list of 1565 potential solutions. On the other hand Router Advertisements 1566 provides a reliable mechanism to dynamically provide hosts with a 1567 list of valid prefixes to use as source addresses as well as prevent 1568 particular prefixes to be used. While no additional new features are 1569 required to be implemented on hosts, routers need to be able to send 1570 RAs based on the state of scoped forwarding tables entries and to 1571 react to network topology changes by sending RAs with particular 1572 parameters set. 1574 The use of ICMPv6 Destination Unreachable messages generated by the 1575 SER (or any SADR-capable) routers seem like they have the potential 1576 to provide a support mechanism together with RAs to signal source 1577 address selection errors back to hosts, however scalability issues 1578 may arise in large networks in case of sudden topology change. 1579 Therefore it is highly desirable that hosts are able to select the 1580 correct source address in case of uplinks failure with ICMPv6 being 1581 an additional mechanism to signal unexpected failures back to hosts. 1583 The current behavior of different host operating system when 1584 receiving ICMPv6 Destination Unreachable message with code 5 (Source 1585 address failed ingress/egress policy) is not clear to the authors. 1586 Information from implementers, users, and testing would be quite 1587 helpful in evaluating this approach. 1589 5.4. Selecting Source Address Upon Failed Uplink Recovery 1591 The next logical step is to look at the scenario when a failed uplink 1592 on SERb1 to ISP-B is coming back up, so hosts can start using source 1593 addresses belonging to 2001:db8:0:b000::/52 again. 1595 5.4.1. Controlling Source Address Selection With DHCPv6 1597 The mechanism to use DHCPv6 to instruct the hosts (H31 in our 1598 example) to start using prefixes from ISP-B space (e.g. 1599 S=2001:db8:0:b010::31 for H31) to reach hosts on the Internet is 1600 quite similar to one discussed in Section 5.3.1 and shares the same 1601 drawbacks. 1603 5.4.2. Controlling Source Address Selection With Router Advertisements 1605 Let's look at the scenario discussed in Section 5.3.2. If the 1606 uplink(s) failure caused the complete withdrawal of prefixes from 1607 2001:db8:0:b000::/52 address space by setting Preferred Lifetime 1608 value to 0, then the recovery of the link should just trigger new RA 1609 being sent with non-zero Preferred Lifetime. In another scenario 1610 discussed in Section 5.3.2, the SERb1 uplink to ISP-B failure leads 1611 to disappearance of the (S=2001:db8:0:b000::/52, D=::/0) entry from 1612 the forwarding table scoped to S=2001:db8:0:b000::/52 and, in turn, 1613 caused R3 to send RAs from LLA_B with Router Lifetime set to 0. The 1614 recovery of the SERb1 uplink to ISP-B leads to 1615 (S=2001:db8:0:b000::/52, D=::/0) scoped forwarding entry re- 1616 appearance and instructs R3 that it should advertise itself as a 1617 default router for ISP-B address space domain (send RAs from LLA_B 1618 with non-zero Router Lifetime). 1620 5.4.3. Controlling Source Address Selection With ICMP 1622 It looks like ICMPv6 provides a rather limited functionality to 1623 signal back to hosts that particular source addresses have become 1624 valid again. Unless the changes in the uplink state a particular 1625 (S,D) pair, hosts can keep using the same source address even after 1626 an ISP uplink has come back up. For example, after the uplink from 1627 SERb1 to ISP-B had failed, H31 received ICMPv6 Code 5 message (as 1628 described in Section 5.3.3) and allegedly started using 1629 (S=2001:db8:0:a010::31, D=2001:db8:0:5678::501) to reach H501. Now 1630 when the SERb1 uplink comes back up, the packets with that (S,D) pair 1631 are still routed to SERa1 and sent to the Internet. Therefore H31 is 1632 not informed that it should stop using 2001:db8:0:a010::31 and start 1633 using 2001:db8:0:b010::31 again. Unless SERa has a policy configured 1634 to drop packets (S=2001:db8:0:a010::31, D=2001:db8:0:5678::501) and 1635 send ICMPv6 back if SERb1 uplink to ISP-B is up, H31 will be unaware 1636 of the network topology change and keep using S=2001:db8:0:a010::31 1637 for Internet destinations, including H51. 1639 One of the possible option may be using a scoped route with EXCLUSIVE 1640 flag as described in Section 5.2.3. SERa1 uplink recovery would 1641 cause (S=2001:db8:0:a000::/52, D=2001:db8:0:1234::/64) route to 1642 reappear in the routing table. In the absence of that route packets 1643 to H101 which were sent to ISP-B (as ISP-A uplink was down) with 1644 source addresses from 2001:db8:0:b000::/52. When the route re- 1645 appears SERb1 would reject those packets and sends ICMPv6 back as 1646 discussed in Section 5.2.3. Practically it might lead to scalability 1647 issues which have been already discussed in Section 5.2.3 and 1648 Section 5.4.3. 1650 5.4.4. Summary Of Methods For Controlling Source Address Selection Upon 1651 Failed Uplink Recovery 1653 Once again DHCPv6 does not look like reasonable choice to manipulate 1654 source address selection process on a host in the case of network 1655 topology changes. Using Router Advertisement provides the flexible 1656 mechanism to dynamically react to network topology changes (if 1657 routers are able to use routing changes as a trigger for sending out 1658 RAs with specific parameters). ICMPv6 could be considered as a 1659 supporting mechanism to signal incorrect source address back to hosts 1660 but should not be considered as the only mechanism to control the 1661 address selection in multihomed environments. 1663 5.5. Selecting Source Address When All Uplinks Failed 1665 One particular tricky case is a scenario when all uplinks have 1666 failed. In that case there is no valid source address to be used for 1667 any external destinations while it might be desirable to have intra- 1668 site connectivity. 1670 5.5.1. Controlling Source Address Selection With DHCPv6 1672 From DHCPv6 perspective uplinks failure should be treated as two 1673 independent failures and processed as described in Section 5.3.1. At 1674 this stage it is quite obvious that it would result in quite 1675 complicated policy table which needs to be explicitly configured by 1676 administrators and therefore seems to be impractical. 1678 5.5.2. Controlling Source Address Selection With Router Advertisements 1680 As discussed in Section 5.3.2 an uplink failure causes the scoped 1681 default entry to disappear from the scoped forwarding table and 1682 triggers RAs with zero Router Lifetime. Complete disappearance of 1683 all scoped entries for a given source prefix would cause the prefix 1684 being withdrawn from hosts by setting Preferred Lifetime value to 1685 zero in PIO. If all uplinks (SERa, SERb1 and SERb2) failed, hosts 1686 either lost their default routers and/or have no global IPv6 1687 addresses to use as a source. (Note that 'uplink failure' might mean 1688 'IPv6 connectivity failure with IPv4 still being reachable', in which 1689 case hosts might fall back to IPv4 if there is IPv4 connectivity to 1690 destinations). As a results intra-site connectivity is broken. One 1691 of the possible way to solve it is to use ULAs. 1693 All hosts have ULA addresses assigned in addition to GUAs and used 1694 for intra-site communication even if there is no GUA assigned to a 1695 host. To avoid accidental leaking of packets with ULA sources SADR- 1696 capable routers SHOULD have a scoped forwarding table for ULA source 1697 for internal routes but MUST NOT have an entry for D=::/0 in that 1698 table. In the absence of (S=ULA_Prefix; D=::/0) first-hop routers 1699 will send dedicated RAs from a unique link-local source LLA_ULA with 1700 PIO from ULA address space, RIO for the ULA prefix and Router 1701 Lifetime set to zero. The behaviour is consistent with the situation 1702 when SERb1 lost the uplink to ISP-B (so there is no Internet 1703 connectivity from 2001:db8:0:b000::/52 sources) but those sources can 1704 be used to reach some specific destinations. In the case of ULA 1705 there is no Internet connectivity from ULA sources but they can be 1706 used to reach another ULA destinations. Note that ULA usage could be 1707 particularly useful if all ISPs assign prefixes via DHCP-PD. In the 1708 absence of ULAs uplinks failure hosts would lost all their GUAs upon 1709 prefix lifetime expiration which again makes intra-site communication 1710 impossible. 1712 It should be noted that the Rule 5.5 (prefer a prefix advertised by 1713 the selected next-hop) takes precedence over the Rule 6 (prefer 1714 matching label, which ensures that GUA source addresses are preferred 1715 over ULAs for GUA destinations). Therefore if ULAs are used, the 1716 network adminstrator needs to ensure that while the site has an 1717 Internet connectivity, hosts do not select a roter which advertises 1718 ULA prefixes as their default router. 1720 5.5.3. Controlling Source Address Selection With ICMPv6 1722 In case of all uplinks failure all SERs will drop outgoing IPv6 1723 traffic and respond with ICMPv6 error message. In the large network 1724 when many hosts are trying to reach Internet destinations it means 1725 that SERs need to generate an ICMPv6 error to every packet they 1726 receive from hosts which presents the same scalability issues 1727 discussed in Section 5.3.3 1729 5.5.4. Summary Of Methods For Controlling Source Address Selection When 1730 All Uplinks Failed 1732 Again, combining SADR with Router Advertisements seems to be the most 1733 flexible and scalable way to control the source address selection on 1734 hosts. 1736 5.6. Summary Of Methods For Controlling Source Address Selection 1738 To summarize the scenarios and options discussed above: 1740 While DHCPv6 allows administrators to manipulate source address 1741 selection policy tables, this method has a number of significant 1742 disadvantages which eliminates DHCPv6 from a list of potential 1743 solutions: 1745 1. It required hosts to support DHCPv6 and its extension (RFC7078); 1747 2. DHCPv6 server needs to monitor network state and detect routing 1748 changes. 1750 3. The use of policy tables requires manual configuration and might 1751 be extremely complicated, especially in the case of distributed 1752 network when large number of remote sites are being served by 1753 centralized DHCPv6 servers. 1755 4. Network topology/routing policy changes could trigger 1756 simultaneous re-configuration of large number of hosts which 1757 present serious scalability issues. 1759 The use of Router Advertisements to influence the source address 1760 selection on hosts seem to be the most reliable, flexible and 1761 scalable solution. It has the following benefits: 1763 1. no new (non-standard) functionality needs to be implemented on 1764 hosts (except for [RFC4191] support); 1766 2. no changes in RA format; 1768 3. routers can react to routing table changes by sending RAs which 1769 would minimize the failover time in the case of network topology 1770 changes; 1772 4. information required for source address selection is broadcast to 1773 all affected hosts in case of topology change event which 1774 improves the scalability of the solution (comparing to DHCPv6 1775 reconfiguration or ICMPv6 error messages). 1777 To fully benefit from the RA-based solution, first-hop routers need 1778 to implement SADR and be able to send dedicated RAs per scoped 1779 forwarding table as discussed above, reacting to network changes with 1780 sending new RAs. It should be noted that the proposed solution would 1781 work even if first-hop routers are not SADR-capable but still able to 1782 send individual RAs for each ISP prefix and react to topology changes 1783 as discussed above (e.g. via configuration knobs). 1785 The RA-based solution relies heavily on hosts correctly implementing 1786 default address selection algorith as defined in [RFC6724]. While 1787 the basic (and most common) multihoming scenario (two or more 1788 Internet uplinks, no 'wall gardens') would work for any host 1789 supporting the minimal implementation of [RFC6724], more complex use 1790 cases (such as "wall garden" and other scenarios when some ISP 1791 resources can only be reached from that ISP address space) require 1792 that hosts support Rule 5.5 of the default address selection 1793 algorithm. There is some evidence that not all host OSes have that 1794 rule implemented currently. However it should be noted that 1795 [RFC8028] states that Rule 5.5 SHOULD be implemented. 1797 ICMPv6 Code 5 error message SHOULD be used to complement RA-based 1798 solution to signal incorrect source address selection back to hosts, 1799 but it SHOULD NOT be considered as the stand-alone solution. To 1800 prevent scenarios when hosts in multihomed envinronments incorrectly 1801 identify onlink/offlink destinations, hosts should treat ICMPv6 1802 Redirects as discussed in [RFC8028]. 1804 5.7. Other Configuration Parameters 1806 5.7.1. DNS Configuration 1808 In mutihomed envinronment each ISP might provide their own list of 1809 DNS servers. E.g. in the topology show on Figure 3, ISP-A might 1810 provide recursive DNS server H51 2001:db8:0:5555::51, while ISP-B 1811 might provide H61 2001:db8:0:6666::61 as a recursive DNS server. 1812 [RFC8106] defines IPv6 Router Advertisement options to allow IPv6 1813 routers to advertise a list of DNS recursive server addresses and a 1814 DNS Search List to IPv6 hosts. Using RDNSS together with 'scoped' 1815 RAs as described above would allow a first-hop router (R3 in the 1816 Figure 3) to send DNS server addresses and search lists provided by 1817 each ISP (or the corporate DNS servers addresses if the enterprise is 1818 running its own DNS servers). 1820 As discussed in Section 5.5.2, failure of all ISP uplinks would cause 1821 deprecaction of all addresses assigned to a host from the address 1822 space if all ISPs. If any intra-site IPv6 connectivity is still 1823 desirable (most likely to be the case for any mid/large scare 1824 network), then ULAs should be used as discussed in Section 5.5.2. In 1825 such a scenario, the enterprise network should run its own recursive 1826 DNS server(s) and provide its ULA addresses to hosts via RDNSS in RAs 1827 send for ULA-scoped forwarding table as described in Section 5.5.2. 1829 There are some scenarios when the final outcome of the name 1830 resolution might be different depending on: 1832 o which DNS server is used; 1834 o which source address the client uses to send a DNS query to the 1835 server (DNS split horizon). 1837 There is no way currently to instruct a host to use a particular DNS 1838 server out of the configured servers list for resolving a particular 1839 name. Therefore it does not seem feasible to solve the problem of 1840 DNS server selection on the host (it should be noted that this 1841 particular issue is protocol-agnostic and happens for IPv4 as well). 1842 In such a scenario it is recommended that the enterprise run its own 1843 local recursive DNS server. 1845 To influence host source address selection for packets sent to a 1846 particular DNS server the following requirements must be met: 1848 o the host supports RIO as defined in [RFC4191]; 1850 o the routers send RIO for routes to DNS server addresses. 1852 For example, if it is desirable that host H31 reaches the ISP-A DNS 1853 server H51 2001:db8:0:5555::51 using its source address 1854 2001:db8:0:a010::31, then both R1 and R2 should send the RIO 1855 containing the route to 2001:db8:0:5555::51 (or covering route) in 1856 their 'scoped' RAs, containing LLA_A as the default router address 1857 and the PO for SLAAC prefix 2001:db8:0:a010::/64. In that case the 1858 host H31 (if it supports the Rule 5.5) would select LLA_A as a next- 1859 hop and then chose 2001:db8:0:a010::31 as the source address for 1860 packets to the DNS server. 1862 It should be noted that [RFC8106] explicitly prohibits using DNS 1863 information if the RA router Lifetime expired: "An RDNSS address or a 1864 DNSSL domain name MUST be used only as long as both the RA router 1865 Lifetime (advertised by a Router Advertisement message) and the 1866 corresponding option Lifetime have not expired.". Therefore hosts 1867 might ignore RDNSS information provided in ULA-scoped RAs as those 1868 RAs would have router lifetime set to 0. However the updated version 1869 of RFC6106 ([RFC8106]) has that requirement removed. 1871 6. Deployment Considerations 1873 The solution described in this document requires certain mechanisms 1874 to be supported by the network infrastructure and hosts. It requires 1875 some routers in the enterprise site to support some form of Source 1876 Address Dependent Routing (SADR). It also requires hosts to be able 1877 to learn when the uplink to an ISP chages its state so the 1878 corresponding source addresses should (or should not) be used. 1879 Ongoing work to create mechanisms to accomplish this are discussed in 1880 this document, but they are still a work in progress. 1882 The solution discussed in this document relies on the default address 1883 selection algorithm ([RFC6724]) Rule 5.5. While [RFC6724] consideres 1884 this rule as optional, the recent [RFC8028] recommends that a host 1885 SHOULD select default routers for each prefix in which it is assigned 1886 an address. It also recommends that hosts SHOULD implement Rule 5.5. 1887 of [RFC6724]. Therefore while RFC8028-compliant hosts already have 1888 mechanism to learn about ISP uplinks state changes and selecting the 1889 source addresses accordingly, many hosts do not have such mechanism 1890 supported yet. 1892 It should be noted that multihomed enterprise network utilizing 1893 multipe ISP prefixes can be considered as a typical mutiple 1894 provisioning domain (mPVD) scenario, as desribed in [RFC7556]. This 1895 document defines a way for network to provide the PVD information to 1896 hosts indirectly, using the existing mechanisms. At the same time 1897 [I-D.ietf-intarea-provisioning-domains]takes one step further and 1898 describes a comprehensive mechanism for hosts to discover the whole 1899 set of configuration information associated with different PVD/ISPs. 1901 [I-D.ietf-intarea-provisioning-domains] complements this document in 1902 terms of making hosts being able to learn about ISP uplink states and 1903 selecting the corresponding source addresses. 1905 7. Other Solutions 1907 7.1. Shim6 1909 The Shim6 working group specified the Shim6 protocol [RFC5533] which 1910 allows a host at a multihomed site to communicate with an external 1911 host and exchange information about possible source and destination 1912 address pairs that they can use to communicate. It also specified 1913 the REAP protocol [RFC5534] to detect failures in the path between 1914 working address pairs and find new working address pairs. A 1915 fundamental requirement for Shim6 is that both internal and external 1916 hosts need to support Shim6. That is, both the host internal to the 1917 multihomed site and the host external to the multihomed site need to 1918 support Shim6 in order for there to be any benefit for the internal 1919 host to run Shim6. The Shim6 protocol specification was published in 1920 2009, but it has not been widely implemented. Therefore Shim6 is not 1921 considered as a viable solution for enterprise multihoming. 1923 7.2. IPv6-to-IPv6 Network Prefix Translation 1925 IPv6-to-IPv6 Network Prefix Translation (NPTv6) [RFC6296] is not the 1926 focus of this document. NPTv6 suffers from the same fundamental 1927 issue as any other address translation approaches: it breaks end-to- 1928 end connectivity. Therefore NPTv6 is not considered as desirable 1929 solution and this document intentionally focuses on solving 1930 enterprise multihoming problem without any form of address 1931 translations. 1933 With increasing interest and ongoing work in bringing path awareness 1934 to transport and application layer protocols hosts migth be able to 1935 determine the properties of the various network paths and choose 1936 among paths available to them. As selecting the correct source 1937 address is one of the possible mechanisms path-aware hosts may 1938 utilize, address translation negatively affects hosts path-awareness 1939 which makes NTPv6 even more undesirable solution. 1941 7.3. Multipath Transport 1943 Using multipath transport might solve the problems discussed in 1944 Section 5 it would allow hosts to use multiple source addresses for a 1945 single connection and switch between source addresses when a 1946 particular address becomes unavailable or a new address gets assigned 1947 to the host interface. Therefore if all hosts in the enterprise 1948 network are only using multipath transport for all connections, the 1949 signalling solution described in Section 5 migth not be needed (it 1950 should be noted that the Source Address Dependent Routing would still 1951 be required to delver packets to the correct uplinks). Unfortunatelt 1952 when this document was written, mutlipath transport alone can not be 1953 considered a solition for the problem of selecting the source address 1954 in a multihomed envinronments. There are significant number of hosts 1955 which do not use mulipath transport currently and it seems unlikely 1956 that the situation is going to change in any foreseeable future. As 1957 the solution for enterprise multihoming needs to work for the least 1958 common denominator: hosts without multipath transport support. In 1959 addition, not all protocols are using multipath transport. While 1960 multipath transport would complement the solution described in 1961 Section 5, it could not be considered as a sole solution to the 1962 problem of source address selection in multihomed envinronments. 1964 8. IANA Considerations 1966 This memo asks the IANA for no new parameters. 1968 9. Security Considerations 1970 This document introduces no new security or privacy considerations. 1971 Security considerations of using stateless address autoconfiguration 1972 is discussed in [RFC4862]. 1974 10. Acknowledgements 1976 The original outline was suggested by Ole Troan. 1978 The authors would like to thank the following people (in alphabetical 1979 order) for their review and feedback: Olivier Bonaventure, Brian E 1980 Carpenter, Lorenzo Colitti, David Lamparter, Acee Lindem, Philip 1981 Matthewsu, Robert Raszuk, Dave Thaler. 1983 11. References 1985 11.1. Normative References 1987 [RFC1122] Braden, R., Ed., "Requirements for Internet Hosts - 1988 Communication Layers", STD 3, RFC 1122, 1989 DOI 10.17487/RFC1122, October 1989, 1990 . 1992 [RFC1123] Braden, R., Ed., "Requirements for Internet Hosts - 1993 Application and Support", STD 3, RFC 1123, 1994 DOI 10.17487/RFC1123, October 1989, 1995 . 1997 [RFC1918] Rekhter, Y., Moskowitz, B., Karrenberg, D., de Groot, G., 1998 and E. Lear, "Address Allocation for Private Internets", 1999 BCP 5, RFC 1918, DOI 10.17487/RFC1918, February 1996, 2000 . 2002 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 2003 Requirement Levels", BCP 14, RFC 2119, 2004 DOI 10.17487/RFC2119, March 1997, 2005 . 2007 [RFC2827] Ferguson, P. and D. Senie, "Network Ingress Filtering: 2008 Defeating Denial of Service Attacks which employ IP Source 2009 Address Spoofing", BCP 38, RFC 2827, DOI 10.17487/RFC2827, 2010 May 2000, . 2012 [RFC3315] Droms, R., Ed., Bound, J., Volz, B., Lemon, T., Perkins, 2013 C., and M. Carney, "Dynamic Host Configuration Protocol 2014 for IPv6 (DHCPv6)", RFC 3315, DOI 10.17487/RFC3315, July 2015 2003, . 2017 [RFC3582] Abley, J., Black, B., and V. Gill, "Goals for IPv6 Site- 2018 Multihoming Architectures", RFC 3582, 2019 DOI 10.17487/RFC3582, August 2003, 2020 . 2022 [RFC4116] Abley, J., Lindqvist, K., Davies, E., Black, B., and V. 2023 Gill, "IPv4 Multihoming Practices and Limitations", 2024 RFC 4116, DOI 10.17487/RFC4116, July 2005, 2025 . 2027 [RFC4191] Draves, R. and D. Thaler, "Default Router Preferences and 2028 More-Specific Routes", RFC 4191, DOI 10.17487/RFC4191, 2029 November 2005, . 2031 [RFC4193] Hinden, R. and B. Haberman, "Unique Local IPv6 Unicast 2032 Addresses", RFC 4193, DOI 10.17487/RFC4193, October 2005, 2033 . 2035 [RFC4218] Nordmark, E. and T. Li, "Threats Relating to IPv6 2036 Multihoming Solutions", RFC 4218, DOI 10.17487/RFC4218, 2037 October 2005, . 2039 [RFC4219] Lear, E., "Things Multihoming in IPv6 (MULTI6) Developers 2040 Should Think About", RFC 4219, DOI 10.17487/RFC4219, 2041 October 2005, . 2043 [RFC4242] Venaas, S., Chown, T., and B. Volz, "Information Refresh 2044 Time Option for Dynamic Host Configuration Protocol for 2045 IPv6 (DHCPv6)", RFC 4242, DOI 10.17487/RFC4242, November 2046 2005, . 2048 [RFC6296] Wasserman, M. and F. Baker, "IPv6-to-IPv6 Network Prefix 2049 Translation", RFC 6296, DOI 10.17487/RFC6296, June 2011, 2050 . 2052 [RFC7157] Troan, O., Ed., Miles, D., Matsushima, S., Okimoto, T., 2053 and D. Wing, "IPv6 Multihoming without Network Address 2054 Translation", RFC 7157, DOI 10.17487/RFC7157, March 2014, 2055 . 2057 [RFC7556] Anipko, D., Ed., "Multiple Provisioning Domain 2058 Architecture", RFC 7556, DOI 10.17487/RFC7556, June 2015, 2059 . 2061 [RFC8028] Baker, F. and B. Carpenter, "First-Hop Router Selection by 2062 Hosts in a Multi-Prefix Network", RFC 8028, 2063 DOI 10.17487/RFC8028, November 2016, 2064 . 2066 [RFC8106] Jeong, J., Park, S., Beloeil, L., and S. Madanapalli, 2067 "IPv6 Router Advertisement Options for DNS Configuration", 2068 RFC 8106, DOI 10.17487/RFC8106, March 2017, 2069 . 2071 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2072 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 2073 May 2017, . 2075 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 2076 (IPv6) Specification", STD 86, RFC 8200, 2077 DOI 10.17487/RFC8200, July 2017, 2078 . 2080 11.2. Informative References 2082 [I-D.baker-ipv6-isis-dst-src-routing] 2083 Baker, F. and D. Lamparter, "IPv6 Source/Destination 2084 Routing using IS-IS", draft-baker-ipv6-isis-dst-src- 2085 routing-07 (work in progress), July 2017. 2087 [I-D.baker-rtgwg-src-dst-routing-use-cases] 2088 Baker, F., Xu, M., Yang, S., and J. Wu, "Requirements and 2089 Use Cases for Source/Destination Routing", draft-baker- 2090 rtgwg-src-dst-routing-use-cases-02 (work in progress), 2091 April 2016. 2093 [I-D.boutier-babel-source-specific] 2094 Boutier, M. and J. Chroboczek, "Source-Specific Routing in 2095 Babel", draft-boutier-babel-source-specific-03 (work in 2096 progress), July 2017. 2098 [I-D.huitema-shim6-ingress-filtering] 2099 Huitema, C., "Ingress filtering compatibility for IPv6 2100 multihomed sites", draft-huitema-shim6-ingress- 2101 filtering-00 (work in progress), September 2005. 2103 [I-D.ietf-intarea-provisioning-domains] 2104 Pfister, P., Vyncke, E., Pauly, T., Schinazi, D., and W. 2105 Shao, "Discovering Provisioning Domain Names and Data", 2106 draft-ietf-intarea-provisioning-domains-02 (work in 2107 progress), June 2018. 2109 [I-D.ietf-rtgwg-dst-src-routing] 2110 Lamparter, D. and A. Smirnov, "Destination/Source 2111 Routing", draft-ietf-rtgwg-dst-src-routing-06 (work in 2112 progress), October 2017. 2114 [I-D.pfister-6man-sadr-ra] 2115 Pfister, P., "Source Address Dependent Route Information 2116 Option for Router Advertisements", draft-pfister-6man- 2117 sadr-ra-01 (work in progress), June 2015. 2119 [I-D.xu-src-dst-bgp] 2120 Xu, M., Yang, S., and J. Wu, "Source/Destination Routing 2121 Using BGP-4", draft-xu-src-dst-bgp-00 (work in progress), 2122 March 2016. 2124 [PATRICIA] 2125 Morrison, D., "Practical Algorithm to Retrieve Information 2126 Coded in Alphanumeric", Journal of the ACM 15(4) 2127 pp514-534, October 1968. 2129 [RFC3704] Baker, F. and P. Savola, "Ingress Filtering for Multihomed 2130 Networks", BCP 84, RFC 3704, DOI 10.17487/RFC3704, March 2131 2004, . 2133 [RFC3736] Droms, R., "Stateless Dynamic Host Configuration Protocol 2134 (DHCP) Service for IPv6", RFC 3736, DOI 10.17487/RFC3736, 2135 April 2004, . 2137 [RFC4443] Conta, A., Deering, S., and M. Gupta, Ed., "Internet 2138 Control Message Protocol (ICMPv6) for the Internet 2139 Protocol Version 6 (IPv6) Specification", STD 89, 2140 RFC 4443, DOI 10.17487/RFC4443, March 2006, 2141 . 2143 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 2144 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 2145 DOI 10.17487/RFC4861, September 2007, 2146 . 2148 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 2149 Address Autoconfiguration", RFC 4862, 2150 DOI 10.17487/RFC4862, September 2007, 2151 . 2153 [RFC4941] Narten, T., Draves, R., and S. Krishnan, "Privacy 2154 Extensions for Stateless Address Autoconfiguration in 2155 IPv6", RFC 4941, DOI 10.17487/RFC4941, September 2007, 2156 . 2158 [RFC5533] Nordmark, E. and M. Bagnulo, "Shim6: Level 3 Multihoming 2159 Shim Protocol for IPv6", RFC 5533, DOI 10.17487/RFC5533, 2160 June 2009, . 2162 [RFC5534] Arkko, J. and I. van Beijnum, "Failure Detection and 2163 Locator Pair Exploration Protocol for IPv6 Multihoming", 2164 RFC 5534, DOI 10.17487/RFC5534, June 2009, 2165 . 2167 [RFC6724] Thaler, D., Ed., Draves, R., Matsumoto, A., and T. Chown, 2168 "Default Address Selection for Internet Protocol Version 6 2169 (IPv6)", RFC 6724, DOI 10.17487/RFC6724, September 2012, 2170 . 2172 [RFC7078] Matsumoto, A., Fujisaki, T., and T. Chown, "Distributing 2173 Address Selection Policy Using DHCPv6", RFC 7078, 2174 DOI 10.17487/RFC7078, January 2014, 2175 . 2177 [RFC7788] Stenberg, M., Barth, S., and P. Pfister, "Home Networking 2178 Control Protocol", RFC 7788, DOI 10.17487/RFC7788, April 2179 2016, . 2181 [RFC8041] Bonaventure, O., Paasch, C., and G. Detal, "Use Cases and 2182 Operational Experience with Multipath TCP", RFC 8041, 2183 DOI 10.17487/RFC8041, January 2017, 2184 . 2186 [RFC8305] Schinazi, D. and T. Pauly, "Happy Eyeballs Version 2: 2187 Better Connectivity Using Concurrency", RFC 8305, 2188 DOI 10.17487/RFC8305, December 2017, 2189 . 2191 Appendix A. Change Log 2193 Initial Version: July 2016 2195 Authors' Addresses 2197 Fred Baker 2198 Santa Barbara, California 93117 2199 USA 2201 Email: FredBaker.IETF@gmail.com 2203 Chris Bowers 2204 Juniper Networks 2205 Sunnyvale, California 94089 2206 USA 2208 Email: cbowers@juniper.net 2210 Jen Linkova 2211 Google 2212 Mountain View, California 94043 2213 USA 2215 Email: furry@google.com