idnits 2.17.1 draft-ietf-mboned-driad-amt-discovery-11.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 : ---------------------------------------------------------------------------- == There are 1 instance of lines with non-RFC6890-compliant IPv4 addresses in the document. If these are example addresses, they should be changed. == There are 2 instances of lines with multicast IPv4 addresses in the document. If these are generic example addresses, they should be changed to use the 233.252.0.x range defined in RFC 5771 Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC7450, updated by this document, for RFC5378 checks: 2001-02-27) -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (December 18, 2019) is 1590 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Obsolete informational reference (is this intentional?): RFC 2845 (Obsoleted by RFC 8945) -- Obsolete informational reference (is this intentional?): RFC 8499 (Obsoleted by RFC 9499) Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Mboned J. Holland 3 Internet-Draft Akamai Technologies, Inc. 4 Updates: 7450 (if approved) December 18, 2019 5 Intended status: Standards Track 6 Expires: June 20, 2020 8 DNS Reverse IP AMT (Automatic Multicast Tunneling) Discovery 9 draft-ietf-mboned-driad-amt-discovery-11 11 Abstract 13 This document updates RFC 7450 (Automatic Multicast Tunneling, or 14 AMT) by modifying the relay discovery process. A new DNS resource 15 record named AMTRELAY is defined for publishing AMT relays for 16 source-specific multicast channels. The reverse IP DNS zone for a 17 multicast sender's IP address is configured to use AMTRELAY resource 18 records to advertise a set of AMT relays that can receive and forward 19 multicast traffic from that sender over an AMT tunnel. Other 20 extensions and clarifications to the relay discovery process are also 21 defined. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at https://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on June 20, 2020. 40 Copyright Notice 42 Copyright (c) 2019 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (https://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 58 1.1. Background . . . . . . . . . . . . . . . . . . . . . . . 4 59 1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 60 1.2.1. Relays and Gateways . . . . . . . . . . . . . . . . . 4 61 1.2.2. Definitions . . . . . . . . . . . . . . . . . . . . . 4 62 2. Relay Discovery Operation . . . . . . . . . . . . . . . . . . 6 63 2.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 6 64 2.2. Signaling and Discovery . . . . . . . . . . . . . . . . . 6 65 2.3. Optimal Relay Selection . . . . . . . . . . . . . . . . . 9 66 2.3.1. Overview . . . . . . . . . . . . . . . . . . . . . . 9 67 2.3.2. Preference Ordering . . . . . . . . . . . . . . . . . 10 68 2.3.3. Connecting to Multiple Relays . . . . . . . . . . . . 13 69 2.4. Happy Eyeballs . . . . . . . . . . . . . . . . . . . . . 13 70 2.4.1. Overview . . . . . . . . . . . . . . . . . . . . . . 13 71 2.4.2. Algorithm Guidelines . . . . . . . . . . . . . . . . 13 72 2.4.3. Connection Definition . . . . . . . . . . . . . . . . 14 73 2.5. Guidelines for Restarting Discovery . . . . . . . . . . . 15 74 2.5.1. Overview . . . . . . . . . . . . . . . . . . . . . . 15 75 2.5.2. Updates to Restarting Events . . . . . . . . . . . . 16 76 2.5.3. Tunnel Stability . . . . . . . . . . . . . . . . . . 17 77 2.5.4. Traffic Health . . . . . . . . . . . . . . . . . . . 17 78 2.5.5. Relay Loaded or Shutting Down . . . . . . . . . . . . 19 79 2.5.6. Relay Discovery Messages vs. Restarting Discovery . . 19 80 2.5.7. Independent Discovery Per Traffic Source . . . . . . 20 81 2.6. DNS Configuration . . . . . . . . . . . . . . . . . . . . 20 82 2.7. Waiting for DNS resolution . . . . . . . . . . . . . . . 20 83 3. Example Deployments . . . . . . . . . . . . . . . . . . . . . 21 84 3.1. Example Receiving Networks . . . . . . . . . . . . . . . 21 85 3.1.1. Internet Service Provider . . . . . . . . . . . . . . 21 86 3.1.2. Small Office . . . . . . . . . . . . . . . . . . . . 22 87 3.2. Example Sending Networks . . . . . . . . . . . . . . . . 24 88 3.2.1. Sender-controlled Relays . . . . . . . . . . . . . . 24 89 3.2.2. Provider-controlled Relays . . . . . . . . . . . . . 25 90 4. AMTRELAY Resource Record Definition . . . . . . . . . . . . . 26 91 4.1. AMTRELAY RRType . . . . . . . . . . . . . . . . . . . . . 26 92 4.2. AMTRELAY RData Format . . . . . . . . . . . . . . . . . . 26 93 4.2.1. RData Format - Precedence . . . . . . . . . . . . . . 27 94 4.2.2. RData Format - Discovery Optional (D-bit) . . . . . . 27 95 4.2.3. RData Format - Type . . . . . . . . . . . . . . . . . 27 96 4.2.4. RData Format - Relay . . . . . . . . . . . . . . . . 28 98 4.3. AMTRELAY Record Presentation Format . . . . . . . . . . . 28 99 4.3.1. Representation of AMTRELAY RRs . . . . . . . . . . . 28 100 4.3.2. Examples . . . . . . . . . . . . . . . . . . . . . . 29 101 5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 29 102 6. Security Considerations . . . . . . . . . . . . . . . . . . . 30 103 6.1. Use of AMT . . . . . . . . . . . . . . . . . . . . . . . 30 104 6.2. Record-spoofing . . . . . . . . . . . . . . . . . . . . . 30 105 6.3. Congestion . . . . . . . . . . . . . . . . . . . . . . . 31 106 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 31 107 8. References . . . . . . . . . . . . . . . . . . . . . . . . . 31 108 8.1. Normative References . . . . . . . . . . . . . . . . . . 31 109 8.2. Informative References . . . . . . . . . . . . . . . . . 33 110 Appendix A. Unknown RRType construction . . . . . . . . . . . . 35 111 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 36 113 1. Introduction 115 This document defines DNS Reverse IP AMT Discovery (DRIAD), a 116 mechanism for AMT gateways to discover AMT relays that are capable of 117 forwarding multicast traffic from a known source IP address. 119 AMT (Automatic Multicast Tunneling) is defined in [RFC7450], and 120 provides a method to transport multicast traffic over a unicast 121 tunnel, in order to traverse non-multicast-capable network segments. 123 Section 4.1.5 of [RFC7450] explains that the relay selection process 124 for AMT is intended to be more flexible than the particular discovery 125 method described in that document, and further explains that the 126 selection process might need to depend on the source of the multicast 127 traffic in some deployments, since a relay must be able to receive 128 multicast traffic from the desired source in order to forward it. 130 That section goes on to suggest DNS-based queries as a possible 131 solution. DRIAD is a DNS-based solution, as suggested there. This 132 solution also addresses the relay discovery issues in the 133 "Disadvantages" lists in Section 3.3 of [RFC8313] and Section 3.4 of 134 [RFC8313]. 136 The goal for DRIAD is to enable multicast connectivity between 137 separate multicast-enabled networks when neither the sending nor the 138 receiving network is connected to a multicast-enabled backbone, 139 without pre-configuring any peering arrangement between the networks. 141 This document updates Section 5.2.3.4 of [RFC7450] by adding a new 142 extension to the relay discovery procedure. 144 1.1. Background 146 The reader is assumed to be familiar with the basic DNS concepts 147 described in [RFC1034], [RFC1035], and the subsequent documents that 148 update them, particularly [RFC2181]. 150 The reader is also assumed to be familiar with the concepts and 151 terminology regarding source-specific multicast as described in 152 [RFC4607] and the use of IGMPv3 [RFC3376] and MLDv2 [RFC3810] for 153 group management of source-specific multicast channels, as described 154 in [RFC4604]. 156 The reader should also be familiar with AMT, particularly the 157 terminology listed in Section 3.2 of [RFC7450] and Section 3.3 of 158 [RFC7450]. 160 1.2. Terminology 162 1.2.1. Relays and Gateways 164 When reading this document, it's especially helpful to recall that 165 once an AMT tunnel is established, the relay receives native 166 multicast traffic and sends unicast tunnel-encapsulated traffic to 167 the gateway, and the gateway receives the tunnel-encapsulated 168 packets, decapsulates them, and forwards them as native multicast 169 packets, as illustrated in Figure 1. 171 Multicast +-----------+ Unicast +-------------+ Multicast 172 >---------> | AMT relay | >=======> | AMT gateway | >---------> 173 +-----------+ +-------------+ 175 Figure 1: AMT Tunnel Illustration 177 1.2.2. Definitions 178 +------------+------------------------------------------------------+ 179 | Term | Definition | 180 +------------+------------------------------------------------------+ 181 | (S,G) | A source-specific multicast channel, as described in | 182 | | [RFC4607]. A pair of IP addresses with a source host | 183 | | IP and destination group IP. | 184 | | | 185 | discovery | A broker or load balancer for AMT relay discovery, | 186 | broker | as mentioned in section 4.2.1.1 of [RFC7450]. | 187 | | | 188 | downstream | Further from the source of traffic, as described in | 189 | | [RFC7450]. | 190 | | | 191 | FQDN | Fully Qualified Domain Name, as described in | 192 | | [RFC8499] | 193 | | | 194 | gateway | An AMT gateway, as described in [RFC7450] | 195 | | | 196 | L flag | The "Limit" flag described in Section 5.1.1.4 of | 197 | | [RFC7450] | 198 | | | 199 | relay | An AMT relay, as described in [RFC7450] | 200 | | | 201 | RPF | Reverse Path Forwarding, as described in [RFC5110] | 202 | | | 203 | RR | A DNS Resource Record, as described in [RFC1034] | 204 | | | 205 | RRType | A DNS Resource Record Type, as described in | 206 | | [RFC1034] | 207 | | | 208 | SSM | Source-specific multicast, as described in [RFC4607] | 209 | | | 210 | upstream | Closer to the source of traffic, as described in | 211 | | [RFC7450]. | 212 | | | 213 | CMTS | Cable Modem Termination System | 214 | | | 215 | OLT | Optical Line Terminal | 216 +------------+------------------------------------------------------+ 218 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 219 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 220 "OPTIONAL" in this document are to be interpreted as described in 221 [RFC2119] and [RFC8174] when, and only when, they appear in all 222 capitals, as shown here. 224 2. Relay Discovery Operation 226 2.1. Overview 228 The AMTRELAY resource record (RR) defined in this document is used to 229 publish the IP address or domain name of a set of AMT relays or 230 discovery brokers that can receive, encapsulate, and forward 231 multicast traffic from a particular sender. 233 The sender is the owner of the RR, and configures the zone so that it 234 contains a set of RRs that provide the addresses or domain names of 235 AMT relays (or discovery brokers that advertise relays) that can 236 receive multicast IP traffic from that sender. 238 This enables AMT gateways in remote networks to discover an AMT relay 239 that is capable of forwarding traffic from the sender. This in turn 240 enables those AMT gateways to receive the multicast traffic tunneled 241 over a unicast AMT tunnel from those relays, and then to pass the 242 multicast packets into networks or applications that are using the 243 gateway to subscribe to traffic from that sender. 245 This mechanism only works for source-specific multicast (SSM) 246 channels. The source address of the (S,G) is reversed and used as an 247 index into one of the reverse mapping trees (in-addr.arpa for IPv4, 248 as described in Section 3.5 of [RFC1035], or ip6.arpa for IPv6, as 249 described in Section 2.5 of [RFC3596]). 251 This mechanism should be treated as an extension of the AMT relay 252 discovery procedure described in Section 5.2.3.4 of [RFC7450]. A 253 gateway that supports this method of AMT relay discovery SHOULD use 254 this method whenever it's performing the relay discovery procedure, 255 and the source IP addresses for desired (S,G)s are known to the 256 gateway, and conditions match the requirements outlined in 257 Section 2.3. 259 Some detailed example use cases are provided in Section 3, and other 260 applicable example topologies appear in Section 3.3 of [RFC8313], 261 Section 3.4 of [RFC8313], and Section 3.5 of [RFC8313]. 263 2.2. Signaling and Discovery 265 This section describes a typical example of the end-to-end process 266 for signaling a receiver's join of an SSM channel that relies on an 267 AMTRELAY RR. 269 The example in Figure 2 contains 2 multicast-enabled networks that 270 are both connected to the internet with non-multicast-capable links, 271 and which have no direct association with each other. 273 A content provider operates a sender, which is a source of multicast 274 traffic inside a multicast-capable network. 276 An end user who is a customer of the content provider has a 277 multicast-capable internet service provider, which operates a 278 receiving network that uses an AMT gateway. The AMT gateway is 279 DRIAD-capable. 281 The content provider provides the user with a receiving application 282 that tries to subscribe to at least one (S,G). This receiving 283 application could for example be a file transfer system using FLUTE 284 [RFC6726] or a live video stream using RTP [RFC3550], or any other 285 application that might subscribe to an SSM channel. 287 +---------------+ 288 | Sender | 289 | | | 198.51.100.15 | 290 | | +---------------+ 291 |Data| | 292 |Flow| Multicast | 293 \| |/ Network | 294 \ / | 5: Propagate RPF for Join(S,G) 295 \ / +---------------+ 296 \/ | AMT Relay | 297 | 2001:db8::f | 298 +---------------+ 299 | 4: Gateway connects to Relay, 300 sends Join(S,G) over tunnel 301 | 302 Unicast 303 Tunnel | 305 ^ | 3: --> DNS Query: type=AMTRELAY, 306 | / 15.100.51.198.in-addr.arpa. 307 | | / <-- Response: 308 Join/Leave +-------------+ AMTRELAY=2001:db8::f 309 Signals | AMT gateway | 310 | +-------------+ 311 | | 2: Propagate RPF for Join(S,G) 312 | Multicast | 313 Network | 314 | 1: Join(S=198.51.100.15,G=232.252.0.2) 315 +-------------+ 316 | Receiver | 317 | (end user) | 318 +-------------+ 320 Figure 2: DRIAD Messaging 322 In this simple example, the sender IP is 198.51.100.15, it is sending 323 traffic to the group address 232.252.0.2, and the relay IP is 324 2001:db8::f. 326 The content provider has previously configured the DNS zone that 327 contains the domain name "15.100.51.198.in-addr.arpa.", which is the 328 reverse lookup domain name for his sender. The zone file contains an 329 AMTRELAY RR with the Relay's IP address. (See Section 4.3 for 330 details about the AMTRELAY RR format and semantics.) 332 The sequence of events depicted in Figure 2 is as follows: 334 1. The end user starts the app, which issues a join to the (S,G): 335 (198.51.100.15, 232.252.0.2). 337 2. The join propagates with RPF through the receiver's multicast- 338 enabled network with PIM [RFC7761] or another multicast routing 339 mechanism, until the AMT gateway receives a signal to join the 340 (S,G). 342 3. The AMT gateway performs a reverse DNS lookup for the AMTRELAY 343 RRType, by sending an AMTRELAY RRType query for the FQDN 344 "15.100.51.198.in-addr.arpa.", using the reverse IP domain name 345 for the sender's source IP address (the S from the (S,G)), as 346 described in Section 3.5 of [RFC1035]. 348 The DNS resolver for the AMT gateway uses ordinary DNS recursive 349 resolution until it has the authoritative result that the content 350 provider configured, which informs the AMT gateway that the relay 351 address is 2001:db8::f. 353 4. The AMT gateway performs AMT handshakes with the AMT relay as 354 described in Section 4 of [RFC7450], then forwards a Membership 355 report to the relay indicating subscription to the (S,G). 357 5. The relay propagates the join through its network toward the 358 sender, then forwards the appropriate AMT-encapsulated traffic to 359 the gateway, which decapsulates and forwards it as native 360 multicast through its downstream network to the end user. 362 In the case of an IPv6 (S,G), the only difference in the AMT relay 363 discovery process is the use of the ip6.arpa reverse IP domain name, 364 as described in Section 2.5 of [RFC3596]), instead of the in- 365 addr.arpa domain name. 367 For example, if the (S,G) is (2001:db8:c::f, ffe3::80fc:2), the 368 reverse domain name for the AMTRELAY query would be: 370 f.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.c.0.0.0.8.b.d.0.1.0.0.2.ip6. 371 arpa. 373 2.3. Optimal Relay Selection 375 2.3.1. Overview 377 The reverse source IP DNS query of an AMTRELAY RR is a good way for a 378 gateway to discover a relay that is known to the sender. 380 However, it is NOT necessarily a good way to discover the best relay 381 for that gateway to use, because the RR will only provide information 382 about relays known to the source. 384 If there is an upstream relay in a network that is topologically 385 closer to the gateway and able to receive and forward multicast 386 traffic from the sender, that relay is better for the gateway to use, 387 since more of the network path uses native multicast, allowing more 388 chances for packet replication. But since that relay is not known to 389 the sender, it won't be advertised in the sender's reverse IP DNS 390 record. An example network that illustrates this scenario is 391 outlined in Section 3.1.2. 393 It's only appropriate for an AMT gateway to discover an AMT relay by 394 querying an AMTRELAY RR owned by a sender when all of these 395 conditions are met: 397 1. The gateway needs to propagate a join of an (S,G) over AMT, 398 because in the gateway's network, no RPF next hop toward the 399 source can propagate a native multicast join of the (S,G); and 401 2. The gateway is not already connected to a relay that forwards 402 multicast traffic from the source of the (S,G); and 404 3. The gateway is not configured to use a particular IP address for 405 AMT discovery, or a relay discovered with that IP is not able to 406 forward traffic from the source of the (S,G); and 408 4. The gateway is not able to find an upstream AMT relay with DNS-SD 409 [RFC6763], using "_amt._udp" as the Service section of the 410 queries, or a relay discovered this way is not able to forward 411 traffic from the source of the (S,G) (as described in 412 Section 2.5.4.1 or Section 2.5.5); and 414 5. The gateway is not able to find an upstream AMT relay with the 415 well-known anycast addresses from Section 7 of [RFC7450]. 417 When the above conditions are met, the gateway has no path within its 418 local network that can receive multicast traffic from the source IP 419 of the (S,G). 421 In this situation, the best way to find a relay that can forward the 422 required traffic is to use information that comes from the operator 423 of the sender. When the sender has configured an AMTRELAY RR, 424 gateways can use the DRIAD mechanism defined in this document to 425 discover the relay information provided by the sender. 427 Note that the DNS-SD service is not source-specific, so even though 428 several methods of finding a more local source of multicast traffic 429 connectivity are preferred where available to using a relay provided 430 by an AMTRELAY RR, a gateway further upstream would still be using 431 the AMTRELAY RR unless the upstream network has a peering or direct 432 connectivity that provides an alternative end-to-end multicast 433 transport path for the (S,G)'s traffic. 435 2.3.2. Preference Ordering 437 This section defines a preference ordering for relay addresses during 438 the relay discovery process. Gateways are encouraged to implement a 439 Happy Eyeballs algorithm to try candidate relays concurrently, but 440 even gateways that do not implement a Happy Eyeballs algorithm SHOULD 441 use this ordering, except as noted. 443 When establishing an AMT tunnel to forward multicast data, it's very 444 important for the discovery process to prioritize the network 445 topology considerations ahead of address selection considerations, in 446 order to gain the packet replication benefits from using multicast 447 instead of unicast tunneling in the multicast-capable portions of the 448 network path. 450 The intent of the advice and requirements in this section is to 451 describe how a gateway should make use of the concurrency provided by 452 a Happy Eyeballs algorithm to reduce the join latency, while still 453 prioritizing network efficiency considerations over Address Selection 454 considerations. 456 Section 4 of [RFC8305] requires a Happy Eyeballs algorithm to sort 457 the addresses with the Destination Address Selection defined in 458 Section 6 of [RFC6724], but for the above reasons, that requirement 459 is superseded in the AMT discovery use case by the following 460 considerations: 462 o Prefer Local Relays 463 Figure 5 and Section 3.1.2 provide a motivating example to prefer 464 DNS-SD [RFC6763] for discovery strictly ahead of using the 465 AMTRELAY RR controlled by the sender for AMT discovery. 467 For this reason, it's RECOMMENDED that AMT gateways by default 468 perform service discovery using DNS Service Discovery (DNS-SD) 469 [RFC6763] for _amt._udp. (with chosen as 470 described in Section 11 of [RFC6763]) and use the AMT relays 471 discovered that way in preference to AMT relays discoverable via 472 the mechanism defined in this document (DRIAD). 474 o Prefer Relays Managed by the Containing Network 476 When no local relay is discoverable with DNS-SD, it still may be 477 the case that a relay local to the receiver is operated by the 478 network providing transit services to the receiver. 480 In this case, when the network cannot make the relay discoverable 481 via DNS-SD, the network SHOULD use the well-known anycast 482 addresses from Section 7 of [RFC7450] to route discovery traffic 483 to the relay most appropriate to the receiver's gateway. 485 Accordingly, the gateway SHOULD by default discover a relay with 486 the well-known AMT anycast addresses as the second preference 487 after DNS-SD when searching for a local relay. 489 o Let Sender Manage Relay Provisioning 491 A related motivating example in the sending-side network is 492 provided by considering a sender that needs to instruct the 493 gateways on how to select between connecting to Figure 6 or 494 Figure 7 (from Section 3.2), in order to manage load and failover 495 scenarios in a manner that operates well with the sender's 496 provisioning strategy for horizontal scaling of AMT relays. 498 In this example about the sending-side network, the precedence 499 field described in Section 4.2.1 is a critical method of control 500 so that senders can provide the appropriate guidance to gateways 501 during the discovery process. 503 Therefore, after DNS-SD, the precedence from the RR MUST be used 504 for sorting preference ahead of the Destination Address Selection 505 ordering from Section 6 of [RFC6724], so that only relay IPs with 506 the same precedence are directly compared according to the 507 Destination Address Selection ordering. 509 Accordingly, AMT gateways SHOULD by default prefer relays in this 510 order: 512 1. DNS-SD 513 2. Anycast addresses from Section 7 of [RFC7450] 514 3. DRIAD 516 This default behavior MAY be overridden by administrative 517 configuration where other behavior is more appropriate for the 518 gateway within its network. 520 Among relay addresses that have an equivalent preference as described 521 above, a Happy Eyeballs algorithm for AMT SHOULD use use the 522 Destination Address Selection defined in Section 6 of [RFC6724]. 524 Among relay addresses that still have an equivalent preference after 525 the above orderings, a gateway SHOULD make a non-deterministic choice 526 (such as a pseudorandom selection) for relay preference ordering, in 527 order to support load balancing by DNS configurations that provide 528 many relay options. 530 The gateway MAY introduce a bias in the non-deterministic choice 531 according to information obtained out of band or from a historical 532 record about network topology, timing information, or the response to 533 a probing mechanism, that indicates some expected benefits from 534 selecting some relays in preference to others. Details about the 535 structure and collection of this information are out of scope for 536 this document, but a gateway in possession of such information MAY 537 use it to prefer topologically closer relays. 539 Within the above constraints, gateways MAY make use of other 540 considerations from Section 4 of [RFC8305], such as the address 541 family interleaving strategies, to produce a final ordering of 542 candidate relay addresses. 544 Note also that certain relay addresses might be excluded from 545 consideration by the hold-down timers described in Section 2.5.4.1 or 546 Section 2.5.5. These relays constitute "unusable destinations" under 547 Rule 1 of the Destination Address Selection, and are also not part of 548 the superseding considerations described above. 550 The discovery and connection process for the relay addresses in the 551 above described ordering MAY operate in parallel, subject to delays 552 prescribed by the Happy Eyeballs requirements described in Section 5 553 of [RFC8305] for successively launched concurrent connection 554 attempts. 556 2.3.3. Connecting to Multiple Relays 558 In some deployments, it may be useful for a gateway to connect to 559 multiple upstream relays and subscribe to the same traffic, in order 560 to support an active/active failover model. A gateway SHOULD NOT be 561 configured to do so without guaranteeing that adequate bandwidth is 562 available. 564 A gateway configured to do this SHOULD still use the same preference 565 ordering logic from Section 2.3.2 for each connection. (Note that 566 this ordering allows for overriding by explicit administrative 567 configuration where required.) 569 2.4. Happy Eyeballs 571 2.4.1. Overview 573 Often, multiple choices of relay will exist for a gateway using DRIAD 574 for relay discovery. Happy Eyeballs [RFC8305] provides a widely 575 deployed and generalizable strategy for probing multiple possible 576 connections in parallel, therefore it is RECOMMENDED that DRIAD- 577 capable gateways implement a Happy Eyeballs algorithm to support fast 578 discovery of the most preferred available relay, by probing multiple 579 relays concurrently. 581 The parallel discovery logic of a Happy Eyeballs algorithm serves to 582 reduce join latency for the initial join of an SSM channel. This 583 section and the preference ordering of relays defined in 584 Section 2.3.2 taken together provide guidance on use of a Happy 585 Eyeballs algorithm for the case of establishing AMT connections. 587 Note that according to the definition in Section 2.4.3 of this 588 document, establishing the connection occurs before sending a 589 membership report. As described in Section 5 of [RFC8305], only one 590 of the successful connections will be used, and the others are all 591 canceled or ignored. In the context of an AMT connection, this means 592 the gateway will send the membership reports that subscribe to 593 traffic only for the chosen connection, after the Happy Eyeballs 594 algorithm resolves. 596 2.4.2. Algorithm Guidelines 598 During the "Initiation of asynchronous DNS queries" phase described 599 in Section 3 of [RFC8305]), a gateway attempts to resolve the domain 600 names listed in Section 2.3. This consists of resolving the SRV 601 queries for DNS-SD domains for the AMT service, as well as the 602 AMTRELAY query for the reverse IP domain defined in this document. 604 Each of the SRV and AMTRELAY responses might contain one or more IP 605 addresses, (as with type 1 or type 2 AMTRELAY responses, or when the 606 SRV Additional Data section of the SRV response contains the address 607 records for the target, as urged by [RFC2782]), or they might contain 608 only domain names (as with type 3 responses from Section 4.2.3, or an 609 SRV response without an additional data section). 611 When present, IP addresses in the initial response provide resolved 612 destination address candidates for the "Sorting of resolved 613 destination addresses" phase described in Section 4 of [RFC8305]), 614 whereas domain names without IP addresses in the initial response 615 result in another set of queries for AAAA and A records, whose 616 responses provide the candidate resolved destination addresses. 618 Since the SRV or AMTRELAY responses don't have a bound on the count 619 of queries that might be generated aside from the bounds imposed by 620 the DNS resolver, it's important for the gateway to provide a rate 621 limit on the DNS queries. The DNS query functionality is expected to 622 follow ordinary standards and best practices for DNS clients. A 623 gateway MAY use an existing DNS client implementation that does so, 624 and MAY rely on that client's rate limiting logic to avoid issuing 625 excessive queries. Otherwise, a gateway MUST provide a rate limit 626 for the DNS queries, and its default settings SHOULD NOT permit more 627 than 10 queries for any 100-millisecond period (though this MAY be 628 overridable by administrative configuration). 630 As the resolved IP addresses arrive, the Happy Eyeballs algorithm 631 sorts them according to the requirements and recommendations given in 632 Section 2.3.2, and attempts connections with the corresponding relays 633 under the algorithm restrictions and guidelines given in [RFC8305] 634 for the "Establishment of one connection, which cancels all other 635 attempts" phase. As described in Section 3 of [RFC8305], DNS 636 resolution is treated as asynchronous, and connection initiation does 637 not wait for lagging DNS responses. 639 2.4.3. Connection Definition 641 Section 5 of [RFC8305] non-normatively describes success at a 642 connection attempt as "generally when the TCP handshake completes". 644 There is no normative definition of a connection in the AMT 645 specification [RFC7450], and there is no TCP connection involved in 646 an AMT tunnel. 648 However, the concept of an AMT connection in the context of a Happy 649 Eyeballs algorithm is a useful one, and so this section provides the 650 following normative definition: 652 o An AMT connection is established successfully when the gateway 653 receives from a newly discovered relay a valid Membership Query 654 message (Section 5.1.4 of [RFC7450]) that does not have the L flag 655 set. 657 See Section 2.5.5 of this document for further information about the 658 relevance of the L flag to the establishment of a Happy Eyeballs 659 connection. See Section 2.5.4 for an overview of how to respond if 660 the connection does not provide multicast connectivity to the source. 662 To "cancel" this kind of AMT connection for the Happy Eyeballs 663 algorithm, a gateway that has not sent a membership report with a 664 subscription would simply stop sending AMT packets for that 665 connection. A gateway only sends a membership report to a connection 666 it has chosen as the most preferred available connection. 668 2.5. Guidelines for Restarting Discovery 670 2.5.1. Overview 672 It's expected that gateways deployed in different environments will 673 use a variety of heuristics to decide when it's appropriate to 674 restart the relay discovery process, in order to meet different 675 performance goals (for example, to fulfill different kinds of service 676 level agreements). 678 In general, restarting the discovery process is always safe for the 679 gateway and relay during any of the events listed in this section, 680 but may cause a disruption in the forwarded traffic if the discovery 681 process results in choosing a different relay, because this changes 682 the RPF forwarding tree for the multicast traffic upstream of the 683 gateway. This is likely to result in some dropped or duplicated 684 packets from channels actively being tunneled from the old relay to 685 the gateway. 687 The degree of impact on the traffic from choosing a different relay 688 may depend on network conditions between the gateway and the new 689 relay, as well as the network conditions and topology between the 690 sender and the new relay, as this may cause the relay to propagate a 691 new RPF join toward the sender. 693 Balancing the expected impact on the tunneled traffic against likely 694 or observed problems with an existing connection to the relay is the 695 goal of the heuristics that gateways use to determine when to restart 696 the discovery process. 698 The non-normative advice in this section should be treated as 699 guidelines to operators and implementors working with AMT systems 700 that can use DRIAD as part of the relay discovery process. 702 2.5.2. Updates to Restarting Events 704 Section 5.2.3.4.1 of [RFC7450] lists several events that may cause a 705 gateway to start or restart the discovery procedure. 707 This document provides some updates and recommendations regarding the 708 handling of these and similar events. The first 5 events are copied 709 here and numbered for easier reference, and the remaining 4 events 710 are newly added for consideration in this document: 712 1. When a gateway pseudo-interface is started (enabled). 714 2. When the gateway wishes to report a group subscription when none 715 currently exist. 717 3. Before sending the next Request message in a membership update 718 cycle. 720 4. After the gateway fails to receive a response to a Request 721 message. 723 5. After the gateway receives a Membership Query message with the L 724 flag set to 1. 726 6. When the gateway wishes to report an (S,G) subscription with a 727 source address that does not currently have other group 728 subscriptions. 730 7. When there is a network change detected, for example when a 731 gateway is operating inside an end user device or application, 732 and the device joins a different network, or when the domain 733 portion of a DNS-SD domain name changes in response to a DHCP 734 message or administrative configuration. 736 8. When substantial loss, persistent congestion, or network overload 737 is detected in the stream of AMT packets from a relay. 739 9. When the gateway has reported one or more (S,G) subscriptions, 740 but no traffic is received from the source for some timeout. 741 (See Section 2.5.4.1). 743 This list is not exhaustive, nor are any of the listed events 744 strictly required to always force a restart of the discovery process. 746 Note that during event #1, a gateway may use DNS-SD, but does not 747 have sufficient information to use DRIAD, since no source is known. 749 2.5.3. Tunnel Stability 751 In general, subscribers to active traffic flows that are being 752 forwarded by an AMT gateway are less likely to experience a 753 degradation in service (for example, from missing or duplicated 754 packets) when the gateway continues using the same relay, as long as 755 the relay is not overloaded and the network conditions remain stable. 757 Therefore, gateways SHOULD avoid performing a full restart of the 758 discovery process during routine cases of event #3 (sending a new 759 Request message), since it occurs frequently in normal operation. 761 However, see Section 2.5.4, Section 2.5.6, and Section 2.5.4.3 for 762 more information about exceptional cases when it may be appropriate 763 to use event #3. 765 2.5.4. Traffic Health 767 2.5.4.1. Absence of Traffic 769 If a gateway indicates one or more (S,G) subscriptions in a 770 Membership Update message, but no traffic for any of the (S,G)s is 771 received in a reasonable time, it's appropriate for the gateway to 772 restart the discovery process. 774 If the gateway restarts the discovery process multiple times 775 consecutively for this reason, the timeout period SHOULD be adjusted 776 to provide a random exponential back-off. 778 The RECOMMENDED timeout is a random value in the range 779 [initial_timeout, MIN(initial_timeout * 2^retry_count, 780 maximum_timeout)], with a RECOMMENDED initial_timeout of 4 seconds 781 and a RECOMMENDED maximum_timeout of 120 seconds (which is the 782 recommended minimum NAT mapping timeout described in [RFC4787]). 784 Note that the recommended initial_timeout is larger than the initial 785 timout recommended in the similar algorithm from Section 5.2.3.4.3 of 786 [RFC7450]. This is to provide time for RPF Join propagation in the 787 sending network. Although the timeout values may be administratively 788 adjusted to support performance requirements, operators are advised 789 to consider the possibility of join propagation delays between the 790 sender and the relay when choosing an appropriate timeout value. 792 Gateways restarting the discovery process because of an absence of 793 traffic MUST use a hold-down timer that removes this relay from 794 consideration during subsequent rounds of discovery while active. 795 The hold-down SHOULD last for no less than 3 minutes and no more than 796 10 minutes. 798 2.5.4.2. Loss and Congestion 800 In some gateway deployments, it is also feasible to monitor the 801 health of traffic flows through the gateway, for example by detecting 802 the rate of packet loss by communicating out of band with receivers, 803 or monitoring the packets of known protocols with sequence numbers. 804 Where feasible, it's encouraged for gateways to use such traffic 805 health information to trigger a restart of the discovery process 806 during event #3 (before sending a new Request message). 808 However, to avoid synchronized rediscovery by many gateways 809 simultaneously after a transient network event upstream of a relay 810 results in many receivers detecting poor flow health at the same 811 time, it's recommended to add a random delay before restarting the 812 discovery process in this case. 814 The span of the random portion of the delay should be no less than 10 815 seconds by default, but may be administratively configured to support 816 different performance requirements. 818 2.5.4.3. Ancient Discovery Information 820 In most cases, a gateway actively receiving healthy traffic from a 821 relay that has not indicated load with the L flag should prefer to 822 remain connected to the same relay, as described in Section 2.5.3. 824 However, a relay that appears healthy but has been forwarding traffic 825 for days or weeks may have an increased chance of becoming unstable. 826 Gateways may benefit from restarting the discovery process during 827 event #3 (before sending a Request message) after the expiration of a 828 long-term timeout, on the order of multiple hours, or even days in 829 some deployments. 831 It may be beneficial for such timers to consider the amount of 832 traffic currently being forwarded, and to give a higher probability 833 of restarting discovery during periods with an unusually low data 834 rate, to reduce the impact on active traffic while still avoiding 835 relying on the results of a very old discovery. 837 Other issues may also be worth considering as part of this heuristic; 838 for example, if the DNS expiry time of the record that was used to 839 discover the current relay has not passed, the long term timer might 840 be restarted without restarting the discovery process. 842 2.5.5. Relay Loaded or Shutting Down 844 The L flag (see Section 5.1.4.4 of [RFC7450]) is the preferred 845 mechanism for a relay to signal overloading or a graceful shutdown to 846 gateways. 848 A gateway that supports handling of the L flag should generally 849 restart the discovery process when it processes a Membership Query 850 packet with the L flag set. If an L flag is received while a 851 concurrent Happy Eyeballs discovery process is under way for multiple 852 candidate relays (Section 2.4), the relay sending the L flag SHOULD 853 NOT be considered for the relay selection. 855 It is also RECOMMENDED that gateways avoid choosing a relay that has 856 recently sent an L flag, with approximately a 10-minute hold-down. 857 Gateways SHOULD treat this hold-down timer in the same way as the 858 hold-down in Section 2.5.4.1, so that the relay is removed from 859 consideration for short-term subsequent rounds of discovery. 861 2.5.6. Relay Discovery Messages vs. Restarting Discovery 863 All AMT relays are required by [RFC7450] to support handling of Relay 864 Discovery messages (e.g. in Section 5.3.3.2 of [RFC7450]). 866 So a gateway with an existing connection to a relay can send a Relay 867 Discovery message to the unicast address of that AMT relay. Under 868 stable conditions with an unloaded relay, it's expected that the 869 relay will return its own unicast address in the Relay Advertisement, 870 in response to such a Relay Discovery message. Since this will not 871 result in the gateway changing to another relay unless the relay 872 directs the gateway away, this is a reasonable exception to the 873 advice against handling event #3 described in Section 2.5.3. 875 This behavior is discouraged for gateways that do support the L flag, 876 to avoid sending unnecessary packets over the network. 878 However, gateways that do not support the L flag may be able to avoid 879 a disruption in the forwarded traffic by sending such Relay Discovery 880 messages regularly. When a relay is under load or has started a 881 graceful shutdown, it may respond with a different relay address, 882 which the gateway can use to connect to a different relay. This kind 883 of coordinated handoff will likely result in a smaller disruption to 884 the traffic than if the relay simply stops responding to Request 885 messages, and stops forwarding traffic. 887 This style of Relay Discovery message (one sent to the unicast 888 address of a relay that's already forwarding traffic to this gateway) 889 SHOULD NOT be considered a full restart of the relay discovery 890 process. It is RECOMMENDED for gateways to support the L flag, but 891 for gateways that do not support the L flag, sending this message 892 during event #3 may help mitigate service degradation when relays 893 become unstable. 895 2.5.7. Independent Discovery Per Traffic Source 897 Relays discovered via the AMTRELAY RR are source-specific relay 898 addresses, and may use different pseudo-interfaces from each other 899 and from relays discovered via DNS-SD or a non-source-specific 900 address, as described in Section 4.1.2.1 of [RFC7450]. 902 Restarting the discovery process for one pseudo-interface does not 903 require restarting the discovery process for other pseudo-interfaces. 904 Gateway heuristics about restarting the discovery process should 905 operate independently for different tunnels to relays, when 906 responding to events that are specific to the different tunnels. 908 2.6. DNS Configuration 910 Often an AMT gateway will only have access to the source and group IP 911 addresses of the desired traffic, and will not know any other name 912 for the source of the traffic. Because of this, typically the best 913 way of looking up AMTRELAY RRs will be by using the source IP address 914 as an index into one of the reverse mapping trees (in-addr.arpa for 915 IPv4, as described in Section 3.5 of [RFC1035], or ip6.arpa for IPv6, 916 as described in Section 2.5 of [RFC3596]). 918 Therefore, it is RECOMMENDED that AMTRELAY RRs be added to reverse IP 919 zones as appropriate. AMTRELAY records MAY also appear in other 920 zones, since this may be necessary to perform delegation from the 921 reverse zones (see for example Section 5.2 of [RFC2317]), but the use 922 case enabled by this document requires a reverse IP mapping for the 923 source from an (S,G) in order to be useful to most AMT gateways. 925 When performing the AMTRELAY RR lookup, any CNAMEs or DNAMEs found 926 MUST be followed. This is necessary to support zone delegation. 927 Some examples outlining this need are described in [RFC2317]. 929 See Section 4 and Section 4.3 for a detailed explanation of the 930 contents for a DNS Zone file. 932 2.7. Waiting for DNS resolution 934 The DNS query functionality is expected to follow ordinary standards 935 and best practices for DNS clients. A gateway MAY use an existing 936 DNS client implementation that does so, and MAY rely on that client's 937 retry logic to determine the timeouts between retries. 939 Otherwise, a gateway MAY re-send a DNS query if it does not receive 940 an appropriate DNS response within some timeout period. If the 941 gateway retries multiple times, the timeout period SHOULD be adjusted 942 to provide a random exponential back-off. 944 As with the waiting process for the Relay Advertisement message from 945 Section 5.2.3.4.3 of [RFC7450], the RECOMMENDED timeout is a random 946 value in the range [initial_timeout, MIN(initial_timeout * 947 2^retry_count, maximum_timeout)], with a RECOMMENDED initial_timeout 948 of 1 second and a RECOMMENDED maximum_timeout of 120 seconds. 950 3. Example Deployments 952 3.1. Example Receiving Networks 954 3.1.1. Internet Service Provider 956 One example of a receiving network is an Internet Service Provider 957 (ISP) that offers multicast ingest services to its subscribers, 958 illustrated in Figure 3. 960 In the example network below, subscribers can join (S,G)s with MLDv2 961 or IGMPv3 as described in [RFC4604], and the AMT gateway in this ISP 962 can receive and forward multicast traffic from one of the example 963 sending networks in Section 3.2 by discovering the appropriate AMT 964 relays with a DNS lookup for the AMTRELAY RR with the reverse IP of 965 the source in the (S,G). 967 Internet 968 ^ ^ Multicast-enabled 969 | | Receiving Network 970 +------|------------|-------------------------+ 971 | | | | 972 | +--------+ +--------+ +=========+ | 973 | | Border |---| Border | | AMT | | 974 | | Router | | Router | | gateway | | 975 | +--------+ +--------+ +=========+ | 976 | | | | | 977 | +-----+------+-----------+--+ | 978 | | | | 979 | +-------------+ +-------------+ | 980 | | Agg Routers | .. | Agg Routers | | 981 | +-------------+ +-------------+ | 982 | / \ \ / \ | 983 | +---------------+ +---------------+ | 984 | |Access Systems | ....... |Access Systems | | 985 | |(CMTS/OLT/etc.)| |(CMTS/OLT/etc.)| | 986 | +---------------+ +---------------+ | 987 | | | | 988 +--------|------------------------|-----------+ 989 | | 990 +---+-+-+---+---+ +---+-+-+---+---+ 991 | | | | | | | | | | 992 /-\ /-\ /-\ /-\ /-\ /-\ /-\ /-\ /-\ /-\ 993 |_| |_| |_| |_| |_| |_| |_| |_| |_| |_| 995 Subscribers 997 Figure 3: Receiving ISP Example 999 3.1.2. Small Office 1001 Another example receiving network is a small branch office that 1002 regularly accesses some multicast content, illustrated in Figure 4. 1004 This office has desktop devices that need to receive some multicast 1005 traffic, so an AMT gateway runs on a LAN with these devices, to pull 1006 traffic in through a non-multicast next-hop. 1008 The office also hosts some mobile devices that have AMT gateway 1009 instances embedded inside apps, in order to receive multicast traffic 1010 over their non-multicast wireless LAN. (Note that the "Legacy 1011 Router" is a simplification that's meant to describe a variety of 1012 possible conditions; for example it could be a device providing a 1013 split-tunnel VPN as described in [RFC7359], deliberately excluding 1014 multicast traffic for a VPN tunnel, rather than a device which is 1015 incapable of multicast forwarding.) 1017 Internet 1018 (non-multicast) 1019 ^ 1020 | Office Network 1021 +----------|----------------------------------+ 1022 | | | 1023 | +---------------+ (Wifi) Mobile apps | 1024 | | Modem+ | Wifi | - - - - w/ embedded | 1025 | | Router | AP | AMT gateways | 1026 | +---------------+ | 1027 | | | 1028 | | | 1029 | +----------------+ | 1030 | | Legacy Router | | 1031 | | (unicast) | | 1032 | +----------------+ | 1033 | / | \ | 1034 | / | \ | 1035 | +--------+ +--------+ +--------+=========+ | 1036 | | Phones | | ConfRm | | Desks | AMT | | 1037 | | subnet | | subnet | | subnet | gateway | | 1038 | +--------+ +--------+ +--------+=========+ | 1039 | | 1040 +---------------------------------------------+ 1042 Figure 4: Small Office (no multicast up) 1044 By adding an AMT relay to this office network as in Figure 5, it's 1045 possible to make use of multicast services from the example 1046 multicast-capable ISP in Section 3.1.1. 1048 Multicast-capable ISP 1049 ^ 1050 | Office Network 1051 +----------|----------------------------------+ 1052 | | | 1053 | +---------------+ (Wifi) Mobile apps | 1054 | | Modem+ | Wifi | - - - - w/ embedded | 1055 | | Router | AP | AMT gateways | 1056 | +---------------+ | 1057 | | +=======+ | 1058 | +---Wired LAN---| AMT | | 1059 | | | relay | | 1060 | +----------------+ +=======+ | 1061 | | Legacy Router | | 1062 | | (unicast) | | 1063 | +----------------+ | 1064 | / | \ | 1065 | / | \ | 1066 | +--------+ +--------+ +--------+=========+ | 1067 | | Phones | | ConfRm | | Desks | AMT | | 1068 | | subnet | | subnet | | subnet | gateway | | 1069 | +--------+ +--------+ +--------+=========+ | 1070 | | 1071 +---------------------------------------------+ 1073 Figure 5: Small Office Example 1075 When multicast-capable networks are chained like this, with a network 1076 like the one in Figure 5 receiving internet services from a 1077 multicast-capable network like the one in Figure 3, it's important 1078 for AMT gateways to reach the more local AMT relay, in order to avoid 1079 accidentally tunneling multicast traffic from a more distant AMT 1080 relay with unicast, and failing to utilize the multicast transport 1081 capabilities of the network in Figure 3. 1083 3.2. Example Sending Networks 1085 3.2.1. Sender-controlled Relays 1087 When a sender network is also operating AMT relays to distribute 1088 multicast traffic, as in Figure 6, each address could appear as an 1089 AMTRELAY RR for the reverse IP of the sender, or one or more domain 1090 names could appear in AMTRELAY RRs, and the AMT relay addresses can 1091 be discovered by finding A or AAAA records from those domain names. 1093 Sender Network 1094 +-----------------------------------+ 1095 | | 1096 | +--------+ +=======+ +=======+ | 1097 | | Sender | | AMT | | AMT | | 1098 | +--------+ | relay | | relay | | 1099 | | +=======+ +=======+ | 1100 | | | | | 1101 | +-----+------+----------+ | 1102 | | | 1103 +-----------|-----------------------+ 1104 v 1105 Internet 1106 (non-multicast) 1108 Figure 6: Small Office Example 1110 3.2.2. Provider-controlled Relays 1112 When an ISP offers a service to transmit outbound multicast traffic 1113 through a forwarding network, it might also offer AMT relays in order 1114 to reach receivers without multicast connectivity to the forwarding 1115 network, as in Figure 7. In this case it's recommended that the ISP 1116 also provide at least one domain name for the AMT relays for use with 1117 the AMTRELAY RR. 1119 When the sender wishes to use the relays provided by the ISP for 1120 forwarding multicast traffic, an AMTRELAY RR should be configured to 1121 use the domain name provided by the ISP, to allow for address 1122 reassignment of the relays without forcing the sender to reconfigure 1123 the corresponding AMTRELAY RRs. 1125 +--------+ 1126 | Sender | 1127 +---+----+ Multicast-enabled 1128 | Sending Network 1129 +-----------|-------------------------------+ 1130 | v | 1131 | +------------+ +=======+ +=======+ | 1132 | | Agg Router | | AMT | | AMT | | 1133 | +------------+ | relay | | relay | | 1134 | | +=======+ +=======+ | 1135 | | | | | 1136 | +-----+------+--------+---------+ | 1137 | | | | 1138 | +--------+ +--------+ | 1139 | | Border |---| Border | | 1140 | | Router | | Router | | 1141 | +--------+ +--------+ | 1142 +-----|------------|------------------------+ 1143 | | 1144 v v 1145 Internet 1146 (non-multicast) 1148 Figure 7: Sending ISP Example 1150 4. AMTRELAY Resource Record Definition 1152 4.1. AMTRELAY RRType 1154 The AMTRELAY RRType has the mnemonic AMTRELAY and type code 260 1155 (decimal). 1157 The AMTRELAY RR is class independent. 1159 4.2. AMTRELAY RData Format 1161 The AMTRELAY RData consists of a 8-bit precedence field, a 1-bit 1162 "Discovery Optional" field, a 7-bit type field, and a variable length 1163 relay field. 1165 0 1 2 3 1166 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 1167 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1168 | precedence |D| type | | 1169 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + 1170 ~ relay ~ 1171 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1173 4.2.1. RData Format - Precedence 1175 This is an 8-bit precedence for this record. It is interpreted in 1176 the same way as the PREFERENCE field described in Section 3.3.9 of 1177 [RFC1035]. 1179 Relays listed in AMTRELAY records with a lower value for precedence 1180 are to be attempted first. 1182 4.2.2. RData Format - Discovery Optional (D-bit) 1184 The D bit is a "Discovery Optional" flag. 1186 If the D bit is set to 0, a gateway using this RR MUST perform AMT 1187 relay discovery as described in Section 4.2.1.1 of [RFC7450], rather 1188 than directly sending an AMT Request message to the relay. 1190 That is, the gateway MUST receive an AMT Relay Advertisement message 1191 (Section 5.1.2 of [RFC7450]) for an address before sending an AMT 1192 Request message (Section 5.1.3 of [RFC7450]) to that address. Before 1193 receiving the Relay Advertisement message, this record has only 1194 indicated that the address can be used for AMT relay discovery, not 1195 for a Request message. This is necessary for devices that are not 1196 fully functional AMT relays, but rather load balancers or brokers, as 1197 mentioned in Section 4.2.1.1 of [RFC7450]. 1199 If the D bit is set to 1, the gateway MAY send an AMT Request message 1200 directly to the discovered relay address without first sending an AMT 1201 Discovery message. 1203 This bit should be set according to advice from the AMT relay 1204 operator. The D bit MUST be set to zero when no information is 1205 available from the AMT relay operator about its suitability. 1207 4.2.3. RData Format - Type 1209 The type field indicates the format of the information that is stored 1210 in the relay field. 1212 The following values are defined: 1214 o type = 0: The relay field is empty (0 bytes). 1216 o type = 1: The relay field contains a 4-octet IPv4 address. 1218 o type = 2: The relay field contains a 16-octet IPv6 address. 1220 o type = 3: The relay field contains a wire-encoded domain name. 1221 The wire-encoded format is self-describing, so the length is 1222 implicit. The domain name MUST NOT be compressed. (See 1223 Section 3.3 of [RFC1035] and Section 4 of [RFC3597].) 1225 4.2.4. RData Format - Relay 1227 The relay field is the address or domain name of the AMT relay. It 1228 is formatted according to the type field. 1230 When the type field is 0, the length of the relay field is 0, and it 1231 indicates that no AMT relay should be used for multicast traffic from 1232 this source. 1234 When the type field is 1, the length of the relay field is 4 octets, 1235 and a 32-bit IPv4 address is present. This is an IPv4 address as 1236 described in Section 3.4.1 of [RFC1035]. This is a 32-bit number in 1237 network byte order. 1239 When the type field is 2, the length of the relay field is 16 octets, 1240 and a 128-bit IPv6 address is present. This is an IPv6 address as 1241 described in Section 2.2 of [RFC3596]. This is a 128-bit number in 1242 network byte order. 1244 When the type field is 3, the relay field is a normal wire-encoded 1245 domain name, as described in Section 3.3 of [RFC1035]. Compression 1246 MUST NOT be used, for the reasons given in Section 4 of [RFC3597]. 1248 For a type 3 record, the D-bit and preference fields carry over to 1249 all A or AAAA records for the domain name. There is no difference in 1250 the result of the discovery process when it's obtained by type 1 or 1251 type 2 AMTRELAY records with identical D-bit and preference fields, 1252 vs. when the result is obtained by a type 3 AMTRELAY record that 1253 resolves to the same set of IPv4 and IPv6 addresses via A and AAAA 1254 lookups. 1256 4.3. AMTRELAY Record Presentation Format 1258 4.3.1. Representation of AMTRELAY RRs 1260 AMTRELAY RRs may appear in a zone data master file. The precedence, 1261 D-bit, relay type, and relay fields are REQUIRED. 1263 If the relay type field is 0, the relay field MUST be ".". 1265 The presentation for the record is as follows: 1267 IN AMTRELAY precedence D-bit type relay 1269 4.3.2. Examples 1271 In a DNS authoritative nameserver that understands the AMTRELAY type, 1272 the zone might contain a set of entries like this: 1274 $ORIGIN 100.51.198.in-addr.arpa. 1275 10 IN AMTRELAY 10 0 1 203.0.113.15 1276 10 IN AMTRELAY 10 0 2 2001:db8::15 1277 10 IN AMTRELAY 128 1 3 amtrelays.example.com. 1279 This configuration advertises an IPv4 discovery address, an IPv6 1280 discovery address, and a domain name for AMT relays which can receive 1281 traffic from the source 198.51.100.10. The IPv4 and IPv6 addresses 1282 are configured with a D-bit of 0 (meaning discovery is mandatory, as 1283 described in Section 4.2.2), and a precedence 10 (meaning they're 1284 preferred ahead of the last entry, which has precedence 128). 1286 For zone files in name servers that don't support the AMTRELAY RRType 1287 natively, it's possible to use the format for unknown RR types, as 1288 described in [RFC3597]. This approach would replace the AMTRELAY 1289 entries in the example above with the entries below: 1291 10 IN TYPE260 \# ( 1292 6 ; length 1293 0a ; precedence=10 1294 01 ; D=0, relay type=1, an IPv4 address 1295 cb00710f ) ; 203.0.113.15 1296 10 IN TYPE260 \# ( 1297 18 ; length 1298 0a ; precedence=10 1299 02 ; D=0, relay type=2, an IPv6 address 1300 20010db800000000000000000000000f ) ; 2001:db8::15 1301 10 IN TYPE260 \# ( 1302 24 ; length 1303 80 ; precedence=128 1304 83 ; D=1, relay type=3, a wire-encoded domain name 1305 09616d7472656c617973076578616d706c6503636f6d ) ; domain name 1307 See Appendix A for more details. 1309 5. IANA Considerations 1311 This document updates the IANA Registry for DNS Resource Record Types 1312 by assigning type 260 to the AMTRELAY record. 1314 This document creates a new registry named "AMTRELAY Resource Record 1315 Parameters", with a sub-registry for the "Relay Type Field". The 1316 initial values in the sub-registry are: 1318 +-------+---------------------------------------+ 1319 | Value | Description | 1320 +-------+---------------------------------------+ 1321 | 0 | No relay is present. | 1322 | 1 | A 4-byte IPv4 address is present | 1323 | 2 | A 16-byte IPv6 address is present | 1324 | 3 | A wire-encoded domain name is present | 1325 | 4-255 | Unassigned | 1326 +-------+---------------------------------------+ 1328 Values 0, 1, 2, and 3 are further explained in Section 4.2.3 and 1329 Section 4.2.4. Relay type numbers 4 through 255 can be assigned with 1330 a policy of Specification Required (as described in [RFC8126]). 1332 6. Security Considerations 1334 6.1. Use of AMT 1336 This document defines a mechanism that enables a more widespread and 1337 automated use of AMT, even without access to a multicast backbone. 1338 Operators of networks and applications that include a DRIAD-capable 1339 AMT gateway are advised to carefully consider the security 1340 considerations in Section 6 of [RFC7450]. 1342 AMT gateway operators also are encouraged to take appropriate steps 1343 to ensure the integrity of the data received via AMT, for example by 1344 the opportunistic use of IPSec [RFC4301] to secure traffic received 1345 from AMT relays, when IPSECKEY records [RFC4025] are available or 1346 when a trust relationship with the AMT relays can be otherwise 1347 established and secured. 1349 6.2. Record-spoofing 1351 The AMTRELAY resource record contains information that SHOULD be 1352 communicated to the DNS client without being modified. The method 1353 used to ensure the result was unmodified is up to the client. 1355 There must be a trust relationship between the end consumer of this 1356 resource record and the DNS server. This relationship may be end-to- 1357 end DNSSEC validation, or a secure connection to a trusted DNS server 1358 that provides end-to-end safety, to prevent record-spoofing of the 1359 response from the trusted server. The connection to the trusted 1360 server can use any secure channel, such as with a TSIG [RFC2845] or 1361 SIG(0) [RFC2931] channel, a secure local channel on the host, DNS 1362 over TLS [RFC7858], DNS over HTTPS [RFC8484], or some other mechanism 1363 that provides authentication of the RR. 1365 If an AMT gateway accepts a maliciously crafted AMTRELAY record, the 1366 result could be a Denial of Service, or receivers processing 1367 multicast traffic from a source under the attacker's control. 1369 6.3. Congestion 1371 Multicast traffic, particularly interdomain multicast traffic, 1372 carries some congestion risks, as described in Section 4 of 1373 [RFC8085]. 1375 Application implementors and network operators that use AMT gateways 1376 are advised to take precautions including monitoring of application 1377 traffic behavior, traffic authentication at ingest, rate-limiting of 1378 multicast traffic, and the use of circuit-breaker techniques such as 1379 those described in Section 3.1.10 of [RFC8085] and similar 1380 protections at the network level, in order to ensure network health 1381 in the event of misconfiguration, poorly written applications that 1382 don't follow UDP congestion control principles, or deliberate attack. 1384 Section 4.1.4.2 of [RFC7450] and Section 6.1 of [RFC7450] provide 1385 some further considerations and advice about mitigating congestion 1386 risk. 1388 7. Acknowledgements 1390 This specification was inspired by the previous work of Doug Nortz, 1391 Robert Sayko, David Segelstein, and Percy Tarapore, presented in the 1392 MBONED working group at IETF 93. 1394 Thanks to Jeff Goldsmith, Toerless Eckert, Mikael Abrahamsson, Lenny 1395 Giuliano, Mark Andrews, Sandy Zheng, Kyle Rose, Ben Kaduk, Bill 1396 Atwood, Tim Chown, Warren Kumari, Dan Romanescu, Bernard Aboba, 1397 Carlos Pignataro, Niclas Comstedt, Mirja Kuehlewind, Henning Rogge, 1398 Eric Vyncke, Barry Lieba, Roman Danyliw, Alissa Cooper, Suresh 1399 Krishnan, and Adam Roach for their very helpful reviews and comments. 1401 8. References 1403 8.1. Normative References 1405 [RFC1034] Mockapetris, P., "Domain names - concepts and facilities", 1406 STD 13, RFC 1034, DOI 10.17487/RFC1034, November 1987, 1407 . 1409 [RFC1035] Mockapetris, P., "Domain names - implementation and 1410 specification", STD 13, RFC 1035, DOI 10.17487/RFC1035, 1411 November 1987, . 1413 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1414 Requirement Levels", BCP 14, RFC 2119, 1415 DOI 10.17487/RFC2119, March 1997, 1416 . 1418 [RFC2181] Elz, R. and R. Bush, "Clarifications to the DNS 1419 Specification", RFC 2181, DOI 10.17487/RFC2181, July 1997, 1420 . 1422 [RFC2782] Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for 1423 specifying the location of services (DNS SRV)", RFC 2782, 1424 DOI 10.17487/RFC2782, February 2000, 1425 . 1427 [RFC3376] Cain, B., Deering, S., Kouvelas, I., Fenner, B., and A. 1428 Thyagarajan, "Internet Group Management Protocol, Version 1429 3", RFC 3376, DOI 10.17487/RFC3376, October 2002, 1430 . 1432 [RFC3596] Thomson, S., Huitema, C., Ksinant, V., and M. Souissi, 1433 "DNS Extensions to Support IP Version 6", STD 88, 1434 RFC 3596, DOI 10.17487/RFC3596, October 2003, 1435 . 1437 [RFC3597] Gustafsson, A., "Handling of Unknown DNS Resource Record 1438 (RR) Types", RFC 3597, DOI 10.17487/RFC3597, September 1439 2003, . 1441 [RFC3810] Vida, R., Ed. and L. Costa, Ed., "Multicast Listener 1442 Discovery Version 2 (MLDv2) for IPv6", RFC 3810, 1443 DOI 10.17487/RFC3810, June 2004, 1444 . 1446 [RFC4604] Holbrook, H., Cain, B., and B. Haberman, "Using Internet 1447 Group Management Protocol Version 3 (IGMPv3) and Multicast 1448 Listener Discovery Protocol Version 2 (MLDv2) for Source- 1449 Specific Multicast", RFC 4604, DOI 10.17487/RFC4604, 1450 August 2006, . 1452 [RFC4607] Holbrook, H. and B. Cain, "Source-Specific Multicast for 1453 IP", RFC 4607, DOI 10.17487/RFC4607, August 2006, 1454 . 1456 [RFC6724] Thaler, D., Ed., Draves, R., Matsumoto, A., and T. Chown, 1457 "Default Address Selection for Internet Protocol Version 6 1458 (IPv6)", RFC 6724, DOI 10.17487/RFC6724, September 2012, 1459 . 1461 [RFC6763] Cheshire, S. and M. Krochmal, "DNS-Based Service 1462 Discovery", RFC 6763, DOI 10.17487/RFC6763, February 2013, 1463 . 1465 [RFC7450] Bumgardner, G., "Automatic Multicast Tunneling", RFC 7450, 1466 DOI 10.17487/RFC7450, February 2015, 1467 . 1469 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1470 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1471 March 2017, . 1473 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 1474 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 1475 May 2017, . 1477 [RFC8305] Schinazi, D. and T. Pauly, "Happy Eyeballs Version 2: 1478 Better Connectivity Using Concurrency", RFC 8305, 1479 DOI 10.17487/RFC8305, December 2017, 1480 . 1482 8.2. Informative References 1484 [RFC2317] Eidnes, H., de Groot, G., and P. Vixie, "Classless IN- 1485 ADDR.ARPA delegation", BCP 20, RFC 2317, 1486 DOI 10.17487/RFC2317, March 1998, 1487 . 1489 [RFC2845] Vixie, P., Gudmundsson, O., Eastlake 3rd, D., and B. 1490 Wellington, "Secret Key Transaction Authentication for DNS 1491 (TSIG)", RFC 2845, DOI 10.17487/RFC2845, May 2000, 1492 . 1494 [RFC2931] Eastlake 3rd, D., "DNS Request and Transaction Signatures 1495 ( SIG(0)s )", RFC 2931, DOI 10.17487/RFC2931, September 1496 2000, . 1498 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1499 Jacobson, "RTP: A Transport Protocol for Real-Time 1500 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1501 July 2003, . 1503 [RFC4025] Richardson, M., "A Method for Storing IPsec Keying 1504 Material in DNS", RFC 4025, DOI 10.17487/RFC4025, March 1505 2005, . 1507 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1508 Internet Protocol", RFC 4301, DOI 10.17487/RFC4301, 1509 December 2005, . 1511 [RFC4787] Audet, F., Ed. and C. Jennings, "Network Address 1512 Translation (NAT) Behavioral Requirements for Unicast 1513 UDP", BCP 127, RFC 4787, DOI 10.17487/RFC4787, January 1514 2007, . 1516 [RFC5110] Savola, P., "Overview of the Internet Multicast Routing 1517 Architecture", RFC 5110, DOI 10.17487/RFC5110, January 1518 2008, . 1520 [RFC6726] Paila, T., Walsh, R., Luby, M., Roca, V., and R. Lehtonen, 1521 "FLUTE - File Delivery over Unidirectional Transport", 1522 RFC 6726, DOI 10.17487/RFC6726, November 2012, 1523 . 1525 [RFC7359] Gont, F., "Layer 3 Virtual Private Network (VPN) Tunnel 1526 Traffic Leakages in Dual-Stack Hosts/Networks", RFC 7359, 1527 DOI 10.17487/RFC7359, August 2014, 1528 . 1530 [RFC7761] Fenner, B., Handley, M., Holbrook, H., Kouvelas, I., 1531 Parekh, R., Zhang, Z., and L. Zheng, "Protocol Independent 1532 Multicast - Sparse Mode (PIM-SM): Protocol Specification 1533 (Revised)", STD 83, RFC 7761, DOI 10.17487/RFC7761, March 1534 2016, . 1536 [RFC7858] Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D., 1537 and P. Hoffman, "Specification for DNS over Transport 1538 Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858, May 1539 2016, . 1541 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 1542 Writing an IANA Considerations Section in RFCs", BCP 26, 1543 RFC 8126, DOI 10.17487/RFC8126, June 2017, 1544 . 1546 [RFC8313] Tarapore, P., Ed., Sayko, R., Shepherd, G., Eckert, T., 1547 Ed., and R. Krishnan, "Use of Multicast across Inter- 1548 domain Peering Points", BCP 213, RFC 8313, 1549 DOI 10.17487/RFC8313, January 2018, 1550 . 1552 [RFC8484] Hoffman, P. and P. McManus, "DNS Queries over HTTPS 1553 (DoH)", RFC 8484, DOI 10.17487/RFC8484, October 2018, 1554 . 1556 [RFC8499] Hoffman, P., Sullivan, A., and K. Fujiwara, "DNS 1557 Terminology", BCP 219, RFC 8499, DOI 10.17487/RFC8499, 1558 January 2019, . 1560 Appendix A. Unknown RRType construction 1562 In a DNS resolver that understands the AMTRELAY type, the zone file 1563 might contain this line: 1565 IN AMTRELAY 128 0 3 amtrelays.example.com. 1567 In order to translate this example to appear as an unknown RRType as 1568 defined in [RFC3597], one could run the following program: 1570 1571 $ cat translate.py 1572 #!/usr/bin/env python3 1573 import sys 1574 name=sys.argv[1] 1575 wire='' 1576 for dn in name.split('.'): 1577 if len(dn) > 0: 1578 wire += ('%02x' % len(dn)) 1579 wire += (''.join('%02x'%ord(x) for x in dn)) 1580 print(len(wire)//2) + 2 1581 print(wire) 1583 $ ./translate.py amtrelays.example.com 1584 24 1585 09616d7472656c617973076578616d706c6503636f6d 1586 1588 The length and the hex string for the domain name 1589 "amtrelays.example.com" are the outputs of this program, yielding a 1590 length of 22 and the above hex string. 1592 22 is the length of the wire-encoded domain name, so to this we add 2 1593 (1 for the precedence field and 1 for the combined D-bit and relay 1594 type fields) to get the full length of the RData, and encode the 1595 precedence, D-bit, and relay type fields as octets, as described in 1596 Section 4. 1598 This results in a zone file entry like this: 1600 IN TYPE260 \# ( 24 ; length 1601 80 ; precedence = 128 1602 03 ; D-bit=0, relay type=3 (wire-encoded domain name) 1603 09616d7472656c617973076578616d706c6503636f6d ) ; domain name 1605 Author's Address 1607 Jake Holland 1608 Akamai Technologies, Inc. 1609 150 Broadway 1610 Cambridge, MA 02144 1611 United States of America 1613 Email: jakeholland.net@gmail.com