idnits 2.17.1 draft-ietf-mboned-ieee802-mcast-problems-02.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 (August 17, 2018) is 2076 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'ICMPv6' is mentioned on line 549, but not defined == Outdated reference: A later version (-23) exists of draft-ietf-6lo-ap-nd-06 == Outdated reference: A later version (-20) exists of draft-ietf-6lo-backbone-router-06 == Outdated reference: A later version (-30) exists of draft-ietf-6tisch-architecture-14 Summary: 0 errors (**), 0 flaws (~~), 5 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Internet Area C. Perkins 3 Internet-Draft M. McBride 4 Intended status: Informational Futurewei 5 Expires: February 18, 2019 D. Stanley 6 HPE 7 W. Kumari 8 Google 9 JC. Zuniga 10 SIGFOX 11 August 17, 2018 13 Multicast Considerations over IEEE 802 Wireless Media 14 draft-ietf-mboned-ieee802-mcast-problems-02 16 Abstract 18 Well-known issues with multicast have prevented the deployment of 19 multicast in 802.11 [dot11], [mc-props], [mc-prob-stmt], and other 20 local-area wireless environments. IETF multicast experts have been 21 meeting together to discuss these issues and provide IEEE updates. 22 The mboned working group is chartered to receive regular reports on 23 the current state of the deployment of multicast technology, create 24 "practice and experience" documents that capture the experience of 25 those who have deployed and are deploying various multicast 26 technologies, and provide feedback to other relevant working groups. 27 This document offers guidance on known limitations and problems with 28 wireless multicast. Also described are various multicast enhancement 29 features that have been specified at IETF and IEEE 802 for wireless 30 media, as well as some operational chioces that can be taken to 31 improve the performace of the network. Finally, some recommendations 32 are provided about the usage and combination of these features and 33 operational choices. 35 Status of This Memo 37 This Internet-Draft is submitted in full conformance with the 38 provisions of BCP 78 and BCP 79. 40 Internet-Drafts are working documents of the Internet Engineering 41 Task Force (IETF). Note that other groups may also distribute 42 working documents as Internet-Drafts. The list of current Internet- 43 Drafts is at https://datatracker.ietf.org/drafts/current/. 45 Internet-Drafts are draft documents valid for a maximum of six months 46 and may be updated, replaced, or obsoleted by other documents at any 47 time. It is inappropriate to use Internet-Drafts as reference 48 material or to cite them other than as "work in progress." 49 This Internet-Draft will expire on February 18, 2019. 51 Copyright Notice 53 Copyright (c) 2018 IETF Trust and the persons identified as the 54 document authors. All rights reserved. 56 This document is subject to BCP 78 and the IETF Trust's Legal 57 Provisions Relating to IETF Documents 58 (https://trustee.ietf.org/license-info) in effect on the date of 59 publication of this document. Please review these documents 60 carefully, as they describe your rights and restrictions with respect 61 to this document. Code Components extracted from this document must 62 include Simplified BSD License text as described in Section 4.e of 63 the Trust Legal Provisions and are provided without warranty as 64 described in the Simplified BSD License. 66 Table of Contents 68 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 69 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 4 70 3. Identified mulitcast issues . . . . . . . . . . . . . . . . . 5 71 3.1. Issues at Layer 2 and Below . . . . . . . . . . . . . . . 5 72 3.1.1. Multicast reliability . . . . . . . . . . . . . . . . 5 73 3.1.2. Lower and Variable Data Rate . . . . . . . . . . . . 5 74 3.1.3. High Interference . . . . . . . . . . . . . . . . . . 6 75 3.1.4. Power-save Effects on Multicast . . . . . . . . . . . 6 76 3.2. Issues at Layer 3 and Above . . . . . . . . . . . . . . . 7 77 3.2.1. IPv4 issues . . . . . . . . . . . . . . . . . . . . . 7 78 3.2.2. IPv6 issues . . . . . . . . . . . . . . . . . . . . . 8 79 3.2.3. MLD issues . . . . . . . . . . . . . . . . . . . . . 8 80 3.2.4. Spurious Neighbor Discovery . . . . . . . . . . . . . 9 81 4. Multicast protocol optimizations . . . . . . . . . . . . . . 9 82 4.1. Proxy ARP in 802.11-2012 . . . . . . . . . . . . . . . . 10 83 4.2. IPv6 Address Registration and Proxy Neighbor Discovery . 10 84 4.3. Buffering to Improve Battery Life . . . . . . . . . . . . 11 85 4.4. IPv6 support in 802.11-2012 . . . . . . . . . . . . . . . 12 86 4.5. Conversion of multicast to unicast . . . . . . . . . . . 12 87 4.6. Directed Multicast Service (DMS) . . . . . . . . . . . . 13 88 4.7. GroupCast with Retries (GCR) . . . . . . . . . . . . . . 13 89 5. Operational optimizations . . . . . . . . . . . . . . . . . . 14 90 5.1. Mitigating Problems from Spurious Neighbor Discovery . . 14 91 6. Multicast Considerations for Other Wireless Media . . . . . . 16 92 7. Recommendations . . . . . . . . . . . . . . . . . . . . . . . 16 93 8. Discussion Items . . . . . . . . . . . . . . . . . . . . . . 16 94 9. Security Considerations . . . . . . . . . . . . . . . . . . . 17 95 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 17 96 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 17 97 12. Informative References . . . . . . . . . . . . . . . . . . . 17 98 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 19 100 1. Introduction 102 Performance issues have been observed when multicast packet 103 transmissions of IETF protocols are used over IEEE 802 wireless 104 media. Even though enhamcements for multicast transmissions have 105 been designed at both IETF and IEEE 802, incompatibilities still 106 exist between specifications, implementations and configuration 107 choices. 109 Many IETF protocols depend on multicast/broadcast for delivery of 110 control messages to multiple receivers. Multicast is used for 111 various purposes such as neighborhood discovery, network flooding, 112 address resolution, as well minimizing media occupancy for the 113 transmission of data that is intended for multiple receivers. In 114 addition to protocol use of broadcast/multicast for control messages, 115 more applications, such as push to talk in hospitals, video in 116 enterprises and lectures in Universities, are streaming over wifi. 117 Many types of end devices are increasingly using wifi for their 118 connectivity. 120 IETF protocols typically rely on network protocol layering in order 121 to reduce or eliminate any dependence of higher level protocols on 122 the specific nature of the MAC layer protocols or the physical media. 123 In the case of multicast transmissions, higher level protocols have 124 traditionally been designed as if transmitting a packet to an IP 125 address had the same cost in interference and network media access, 126 regardless of whether the destination IP address is a unicast address 127 or a multicast or broadcast address. This model was reasonable for 128 networks where the physical medium was wired, like Ethernet. 129 Unfortunately, for many wireless media, the costs to access the 130 medium can be quite different. Multicast over wifi has often been 131 plagued by such poor performance that it is disallowed. Some 132 enhancements have been designed in IETF protocols that are assumed to 133 work primarily over wireless media. However, these enhancements are 134 usually implemented in limited deployments and not widespread on most 135 wireless networks. 137 IEEE 802 wireless protocols have been designed with certain features 138 to support multicast traffic. For instance, lower modulations are 139 used to transmit multicast frames, so that these can be received by 140 all stations in the cell, regardless of the distance or path 141 attenuation from the base station or access point. However, these 142 lower modulation transmissions occupy the medium longer; they hamper 143 efficient transmission of traffic using higher order modulations to 144 nearby stations. For these and other reasons, IEEE 802 working 145 groups such as 802.11 have designed features to improve the 146 performance of multicast transmissions at Layer 2 [ietf_802-11]. In 147 addition to protocol design features, certain operational and 148 configuration enhancements can ameliorate the network performance 149 issues created by multicast traffic. as described in Section 5. 151 In discussing these issues over email, and in a side meeting at IETF 152 99, it has been generally agreed that these problems will not be 153 fixed anytime soon primarily because it's expensive to do so and 154 multicast is unreliable. A big problem is that multicast is somewhat 155 a second class citizen, to unicast, over wifi. There are many 156 protocols using multicast and there needs to be something provided in 157 order to make them more reliable. The problem of IPv6 neighbor 158 discovery saturating the wifi link is only part of the problem. Wifi 159 traffic classes may help. We need to determine what problem should 160 be solved by the IETF and what problem should be solved by the IEEE 161 (see Section 8). A "multicast over wifi" IETF mailing list has been 162 formed (mcast-wifi@ietf.org) for further discussion. This draft will 163 be updated according to the current state of discussion. 165 This document details various problems caused by multicast 166 transmission over wireless networks, including high packet error 167 rates, no acknowledgements, and low data rate. It also explains some 168 enhancements that have been designed at IETF and IEEE 802 to 169 ameliorate the effects of multicast traffic. Recommendations are 170 also provided to implementors about how to use and combine these 171 enhancements. Some advice about the operational choices that can be 172 taken is also included. It is likely that this document will also be 173 considered relevant to designers of future IEEE wireless 174 specifications. 176 2. Terminology 178 This document uses the following definitions: 180 AP 181 IEEE 802.11 Access Point. 183 basic rate 184 The "lowest common denominator" data rate at which multicast and 185 broadcast traffic is generally transmitted. 187 DTIM 188 Delivery Traffic Indication Map (DTIM): An information element 189 that advertises whether or not any associated stations have 190 buffered multicast or broadcast frames. 192 MCS 193 Modulation and Coding Scheme. 195 STA 196 802.11 station (e.g. handheld device). 198 TIM 199 Traffic Indication Map (TIM): An information element that 200 advertises whether or not any associated stations have buffered 201 unicast frames. 203 3. Identified mulitcast issues 205 3.1. Issues at Layer 2 and Below 207 In this section we describe some of the issues related to the use of 208 multicast transmissions over IEEE 802 wireless technologies. 210 3.1.1. Multicast reliability 212 Multicast traffic is typically much less reliable than unicast 213 traffic. Since multicast makes point-to-multipoint communications, 214 multiple acknowledgements would be needed to guarantee reception at 215 all recipients. Since typically there are no ACKs for multicast 216 packets, it is not possible for the Access Point (AP) to know whether 217 or not a retransmission is needed. Even in the wired Internet, this 218 characteristic often causes undesirably high error rates. This has 219 contributed to the relatively slow uptake of multicast applications 220 even though the protocols have long been available. The situation 221 for wireless links is much worse, and is quite sensitive to the 222 presence of background traffic. Consequently, there can be a high 223 packet error rate (PER) due to lack of retransmission, and because 224 the sender never backs off. It is not uncommon for there to be a 225 packet loss rate of 5% or more, which is particularly troublesome for 226 video and other environments where high data rates and high 227 reliability are required. 229 3.1.2. Lower and Variable Data Rate 231 One big difference between multicast over wired versus multicast over 232 wired is that transmission over wired links often occurs at a fixed 233 rate. Wifi, on the other hand, has a transmission rate which varies 234 depending upon the client's proximity to the AP. The throughput of 235 video flows, and the capacity of the broader wifi network, will 236 change and will impact the ability for QoS solutions to effectively 237 reserve bandwidth and provide admission control. 239 For wireless stations associated with an Access Points, the power 240 necessary for good reception can vary from station to station. For 241 unicast, the goal is to minimize power requirements while maximizing 242 the data rate to the destination. For multicast, the goal is simply 243 to maximize the number of receivers that will correctly receive the 244 multicast packet; generally the Access Point has to use a much lower 245 data rate at a power level high enough for even the farthest station 246 to receive the packet. Consequently, the data rate of a video 247 stream, for instance, would be constrained by the environmental 248 considerations of the least reliable receiver associated with the 249 Access Point. 251 Because more robust modulation and coding schemes (MCSs) have longer 252 range but also lower data rate, multicast / broadcast traffic is 253 generally transmitted at the lowest common denominator rate, also 254 known as the basic rate. The amount of additional interference 255 depends on the specific wireless technology. In fact backward 256 compatibility and multi-stream implementations mean that the maximum 257 unicast rates are currently up to a few Gb/s, so there can be a more 258 than 3 orders of magnitude difference in the transmission rate 259 between the basic rates to optimal unicast forwarding. Some 260 techinues employed to increase spectral efficiency, such as spatial 261 multiplexing in mimo systems, are not available with more than one 262 intended reciever; it is not the case that backwards compatibility is 263 the only factor responsible for lower multicast transmission rates. 265 Wired multicast also affects wireless LANs when the AP extends the 266 wired segment; in that case, multicast / broadcast frames on the 267 wired LAN side are copied to WLAN. Since broadcast messages are 268 transmitted at the most robust MCS, many large frames are sent at a 269 slow rate over the air. 271 3.1.3. High Interference 273 Transmissions at a lower rate require longer occupancy of the 274 wireless medium and thus take away from the airtime of other 275 communications and degrade the overall capacity. Furthermore, 276 transmission at higher power, as is required to reach all multicast 277 clients associated to the AP, proportionately increases the area of 278 interference. 280 3.1.4. Power-save Effects on Multicast 282 One of the characteristics of multicast transmission is that every 283 station has to be configured to wake up to receive the multicast, 284 even though the received packet may ultimately be discarded. This 285 process can have a large effect on the power consumption by the 286 multicast receiver station. 288 Multicast can work poorly with the power-save mechanisms defined in 289 IEEE 802.11e, for the following reasons. 291 o Clients may be unable to stay in sleep mode due to multicast 292 control packets frequently waking them up. 293 o Both unicast and multicast traffic can be delayed by power-saving 294 mechanisms. 295 o A unicast packet is delayed until a STA wakes up and requests it. 296 Unicast traffic may also be delayed to improve power save, 297 efficiency and increase probability of aggregation. 298 o Multicast traffic is delayed in a wireless network if any of the 299 STAs in that network are power savers. All STAs associated to the 300 AP have to be awake at a known time to receive multicast traffic. 301 o Packets can also be discarded due to buffer limitations in the AP 302 and non-AP STA. 304 3.2. Issues at Layer 3 and Above 306 This section identifies some representative IETF protocols, and 307 describes possible negative effects due to performance degradation 308 when using multicast transmissions for control messages. Common uses 309 of multicast include: 311 o Control plane for IPv4 and IPv6 312 o ARP and Neighbor Discovery 313 o Service discovery 314 o Applications (video delivery, stock data etc) 315 o Other L3 protocols (non-IP) 317 3.2.1. IPv4 issues 319 The following list contains a few representative IPv4 protocols using 320 multicast. 322 o ARP 323 o DHCP 324 o mDNS 326 After initial configuration, ARP and DHCP occur much less commonly. 327 But service discovery can occur at any time. Apple's Bonjour 328 protocol, for instance, provides service discovery (for printing) 329 that utilizes multicast. It's the first thing operators drop. Even 330 if multicast snooping is utilized, many devices register at once 331 using Bonjour, causing serious network degradation. 333 3.2.2. IPv6 issues 335 IPv6 makes much more extensive use of multicast, including the 336 following: 338 o DHCPv6 339 o IPv6 Neighbor Discovery Protocol (NDP) is not very tolerant of 340 packet losses. In particular, the Duplicate Address Detection 341 (DAD) process fails when the owner of an address does not receive 342 the multicast DAD message from another node that wishes to own 343 that same address. This can result in an address being duplicated 344 in the subnet, breaking a basic assumption of IPv6 connectivity. 345 o IPv6 NDP Neighbor Solicitation (NS) messages used in DAD and 346 Address Lookup make use of Link-Scope multicast. In contrast to 347 IPv4, an IPv6 Node will typically use multiple addresses, and may 348 change them often for privacy reasons. This multiplies the impact 349 of multicast messages that are associated to the mobility of a 350 Node. Router advertisement (RA) messages are also periodically 351 multicasted over the Link. 352 o Neighbors may be considered lost if several consecutive packets 353 fail. 355 Address Resolution 357 Service Discovery 359 Route Discovery 361 Decentralized Address Assignment 363 Geographic routing 365 3.2.3. MLD issues 367 Multicast Listener Discovery(MLD) [RFC4541] is often used to identify 368 members of a multicast group that are connected to the ports of a 369 switch. Forwarding multicast frames into a WiFi-enabled area can use 370 such switch support for hardware forwarding state information. 371 However, since IPv6 makes heavy use of multicast, each STA with an 372 IPv6 address will require state on the switch for several and 373 possibly many multicast solicited-node addresses. Multicast 374 addresses that do not have forwarding state installed (perhaps due to 375 hardware memory limitations on the switch) cause frames to be flooded 376 on all ports of the switch. 378 3.2.4. Spurious Neighbor Discovery 380 On the Internet there is a "background radiation" of scanning traffic 381 (people scanning for vulnerable machines) and backscatter (responses 382 from spoofed traffic, etc). This means that routers very often 383 receive packets destined for machines whose IP addresses may or may 384 not be in use. In the cases where the IP is assigned to a host, the 385 router broadcasts an ARP request, gets back an ARP reply, and caches 386 it; then traffic can be delivered to the host. When the IP address 387 is not in use, the router broadcasts one (or more) ARP requests, and 388 never gets a reply. This means that it does not populate the ARP 389 cache, and the next time there is traffic for that IP address the 390 router will rebroadcast the ARP requests. 392 The rate of these ARP requests is proportional to the size of the 393 subnets, the rate of scanning and backscatter, and how long the 394 router keeps state on non-responding ARPs. As it turns out, this 395 rate is inversely proportional to how occupied the subnet is (valid 396 ARPs end up in a cache, stopping the broadcasting; unused IPs never 397 respond, and so cause more broadcasts). Depending on the address 398 space in use, the time of day, how occupied the subnet is, and other 399 unknown factors, on the order of 2000 broadcasts per second have been 400 observed at the IETF NOCs. 402 On a wired network, there is not a huge difference between unicast, 403 multicast and broadcast traffic. Due to hardware filtering (see, 404 e.g., [Deri-2010]), inadvertently flooded traffic (or high amounts of 405 ethernet multicast) on wired networks can be quite a bit less costly, 406 compared to wireless cases where sleeping devices have to wake up to 407 process packets. Wired Ethernets tend to be switched networks, 408 further reducing interference from multicast. There is effectively 409 no collision / scheduling problem except at extremely high port 410 utilizations. 412 This is not true in the wireless realm; wireless equipment is often 413 unable to send high volumes of broadcast and multicast traffic. 414 Consequently, on the wireless networks, we observe a significant 415 amount of dropped broadcast and multicast packets. This, in turn, 416 means that when a host connects it is often not able to complete 417 DHCP, and IPv6 RAs get dropped, leading to users being unable to use 418 the network. 420 4. Multicast protocol optimizations 422 This section lists some optimizations that have been specified in 423 IEEE 802 and IETF that are aimed at reducing or eliminating the 424 issues discussed in Section 3. 426 4.1. Proxy ARP in 802.11-2012 428 The AP knows the MAC address and IP address for all associated STAs. 429 In this way, the AP acts as the central "manager" for all the 802.11 430 STAs in its BSS. Proxy ARP is easy to implement at the AP, and 431 offers the following advantages: 433 o Reduced broadcast traffic (transmitted at low MCS) on the wireless 434 medium 435 o STA benefits from extended power save in sleep mode, as ARP 436 requests for STA's IP address are handled instead by the AP. 437 o ARP frames are kept off the wireless medium. 438 o No changes are needed to STA implementation. 440 Here is the specification language as described in clause 10.23.13 of 441 [dot11-proxyarp]: 443 When the AP supports Proxy ARP "[...] the AP shall maintain a 444 Hardware Address to Internet Address mapping for each associated 445 station, and shall update the mapping when the Internet Address of 446 the associated station changes. When the IPv4 address being 447 resolved in the ARP request packet is used by a non-AP STA 448 currently associated to the BSS, the proxy ARP service shall 449 respond on behalf of the non-AP STA" 451 4.2. IPv6 Address Registration and Proxy Neighbor Discovery 453 As used in this section, a Low-Power Wireless Personal Area Network 454 (6LoWPAN) denotes a low power lossy network (LLN) that supports 455 6LoWPAN Header Compression (HC) [RFC6282]. A 6TiSCH network 456 [I-D.ietf-6tisch-architecture] is an example of a 6LowPAN. In order 457 to control the use of IPv6 multicast over 6LoWPANs, the 6LoWPAN 458 Neighbor Discovery (6LoWPAN ND) [RFC6775] standard defines an address 459 registration mechanism that relies on a central registry to assess 460 address uniqueness, as a substitute to the inefficient Duplicate 461 Address Detection (DAD) mechanism found in the mainstream IPv6 462 Neighbor Discovery Protocol (NDP) [RFC4861][RFC4862]. 464 The 6lo Working Group is now completing an update 465 [I-D.ietf-6lo-rfc6775-update] to RFC6775. The update enables the 466 registration to a Backbone Router [I-D.ietf-6lo-backbone-router], 467 which proxies for the registered addresses with the mainstream IPv6 468 NDP running on a high speed aggragating backbone. The update also 469 enables a proxy registration on behalf of the registered node, e.g. 470 by a 6LoWPAN router to which the mobile node is attached. 472 The general idea behind the backbone router concept is that in a 473 variety of Wireless Local Area Networks (WLANs) and Wireless Personal 474 Area Networks (WPANs), the broadcast/multicast domain should be 475 controlled, and connectivity to a particular link that provides the 476 subnet should be left to Layer-3. The model for the Backbone Router 477 operation is represented in Figure 1. 479 | 480 +-----+ 481 | | Gateway (default) router 482 | | 483 +-----+ 484 | 485 | Backbone Link 486 +--------------------+------------------+ 487 | | | 488 +-----+ +-----+ +-----+ 489 | | Backbone | | Backbone | | Backbone 490 | | router | | router | | router 491 +-----+ +-----+ +-----+ 492 o o o o o o 493 o o o o o o o o o o o o o o 494 o o o o o o o o o o o o o o o 495 o o o o o o o o o o 496 o o o o o o o 498 LLN LLN LLN 500 Figure 1: Backbone Link and Backbone Routers 502 LLN nodes can move freely from an LLN anchored at one IPv6 Backbone 503 Router to an LLN anchored at another Backbone Router on the same 504 backbone, keeping any of the IPv6 addresses they have configured. 505 The Backbone Routers maintain a Binding Table of their Registered 506 Nodes, which serves as a distributed database of all the LLN Nodes. 507 An extension to the Neighbor Discovery Protocol is introduced to 508 exchange that information across the Backbone Link in the reactive 509 fashion of mainstream IPv6 Neighbor Discovery. 511 RFC6775 and follow-on work (e.g., [I-D.ietf-6lo-ap-nd], are designed 512 to address the needs of LLNs, but the techniques are likely to be 513 valuable on any type of link where sleeping devices are attached, or 514 where the use of broadcast and multicast operations should be 515 limited. 517 4.3. Buffering to Improve Battery Life 519 Methods have been developed to help save battery life; for example, a 520 device might not wake up when the AP receives a multicast packet. 521 The AP acts on behalf of STAs in various ways. To enable use of the 522 power-saving feature for STAs in its BSS, the AP buffers frames for 523 delivery to the STA at the time when the STA is scheduled for 524 reception. If an AP, for instance, expresses a DTIM (Delivery 525 Traffic Indication Message) of 3 then the AP will send a multicast 526 packet every 3 packets. In fact, when any single wireless client 527 associated with an access point has 802.11 power-save mode enabled, 528 the access point buffers all multicast frames and sends them only 529 after the next DTIM beacon. 531 But in practice, most AP's will send a multicast every 30 packets. 532 For unicast there's a TIM (Traffic Indication Message); but since 533 multicast is going to everyone, the AP sends a broadcast to everyone. 534 DTIM does power management but clients can choose whether or not to 535 wake up or not and whether or not to drop the packet. Unfortunately, 536 without proper administrative control, such clients may no longer be 537 able to determine why their multicast operations do not work. 539 4.4. IPv6 support in 802.11-2012 541 IPv6 uses Neighbor Discovery Protocol (NDP) instead of ARP. Every 542 IPv6 node subscribes to a special multicast address for this purpose. 544 Here is the specification language from clause 10.23.13 of 545 [dot11-proxyarp]: 547 "When an IPv6 address is being resolved, the Proxy Neighbor 548 Discovery service shall respond with a Neighbor Advertisement 549 message [...] on behalf of an associated STA to an [ICMPv6] 550 Neighbor Solicitation message [...]. When MAC address mappings 551 change, the AP may send unsolicited Neighbor Advertisement 552 Messages on behalf of a STA." 554 NDP may be used to request additional information 556 o Maximum Transmission Unit 557 o Router Solicitation 558 o Router Advertisement, etc. 560 NDP messages are sent as group addressed (broadcast) frames in 561 802.11. Using the proxy operation helps to keep NDP messages off the 562 wireless medium. 564 4.5. Conversion of multicast to unicast 566 It is often possible to transmit multicast control and data messages 567 by using unicast transmissions to each station individually. 569 4.6. Directed Multicast Service (DMS) 571 There are situations where more is needed than simply converting 572 multicast to unicast. For these purposes, DMS enables a client to 573 request that the AP transmit multicast group addressed frames 574 destined to the requesting clients as individually addressed frames 575 [i.e., convert multicast to unicast]. Here are some characteristics 576 of DMS: 578 o Requires 802.11n A-MSDUs 579 o Individually addressed frames are acknowledged and are buffered 580 for power save clients 581 o The requesting STA may specify traffic characteristics for DMS 582 traffic 583 o DMS was defined in IEEE Std 802.11v-2011 584 o DMS requires changes to both AP and STA implementation. 586 DMS is not currently implemented in products. See [Tramarin2017] and 587 [Oliva2013] for more information. 589 4.7. GroupCast with Retries (GCR) 591 GCR (defined in [dot11aa]) provides greater reliability by using 592 either unsolicited retries or a block acknowledgement mechanism. GCR 593 increases probability of broadcast frame reception success, but still 594 does not guarantee success. 596 For the block acknowledgement mechanism, the AP transmits each group 597 addressed frame as conventional group addressed transmission. 598 Retransmissions are group addressed, but hidden from non-11aa 599 clients. A directed block acknowledgement scheme is used to harvest 600 reception status from receivers; retransmissions are based upon these 601 responses. 603 GCR is suitable for all group sizes including medium to large groups. 604 As the number of devices in the group increases, GCR can send block 605 acknowledgement requests to only a small subset of the group. GCR 606 does require changes to both AP and STA implementation. 608 GCR may introduce unacceptable latency. After sending a group of 609 data frames to the group, the AP has do the following: 611 o unicast a Block Ack Request (BAR) to a subset of members. 612 o wait for the corresponding Block Ack (BA). 613 o retransmit any missed frames. 614 o resume other operations which may have been delayed. 616 This latency may not be acceptable for some traffic. 618 There are ongoing extensions in 802.11 to improve GCR performance. 620 o BAR is sent using downlink MU-MIMO (note that downlink MU-MIMO is 621 already specified in 802.11-REVmc 4.3). 622 o BA is sent using uplink MU-MIMO (which is a .11ax feature). 623 o Additional 802.11ax extensions are under consideration; see 624 [mc-ack-mux] 625 o Latency may also be reduced by simultaneously receiving BA 626 information from multiple clients. 628 5. Operational optimizations 630 This section lists some operational optimizations that can be 631 implemented when deploying wireless IEEE 802 networks to mitigate the 632 issues discussed in Section 3. 634 5.1. Mitigating Problems from Spurious Neighbor Discovery 636 ARP Sponges 638 An ARP Sponge sits on a network and learn which IPs addresses 639 are actually in use. It also listen for ARP requests, and, if 640 it sees an ARP for an IP address which it believes is not used, 641 it will reply with its own MAC address. This means that the 642 router now has an IP to MAC mapping, which it caches. If that 643 IP is later assigned to an machine (e.g using DHCP), the ARP 644 sponge will see this, and will stop replying for that address. 645 Gratuitous ARPs (or the machine ARPing for its gateway) will 646 replace the sponged address in the router ARP table. This 647 technique is quite effective; but, unfortunately, the ARP 648 sponge daemons were not really designed for this use (the 649 standard one [arpsponge], was designed to deal with the 650 disappearance of participants from an IXP) and so are not 651 optimized for this purpose. We have to run one daemon per 652 subnet, the tuning is tricky (the scanning rate versus the 653 population rate versus retires, etc.) and sometimes the daemons 654 just seem to stop, requiring a restart of the daemon and 655 causing disruption. 657 Router mitigations 659 Some routers (often those based on Linux) implement a "negative 660 ARP cache" daemon. Simply put, if the router does not see a 661 reply to an ARP it can be configured to cache this information 662 for some interval. Unfortunately, the core routers which we 663 are using do not support this. When a host connects to network 664 and gets an IP address, it will ARP for its default gateway 665 (the router). The router will update its cache with the IP to 666 host MAC mapping learnt from the request (passive ARP 667 learning). 669 Firewall unused space 671 The distribution of users on wireless networks / subnets 672 changes from meeting to meeting (e.g the "IETF-secure" SSID was 673 renamed to "IETF", fewer users use "IETF-legacy", etc). This 674 utilization is difficult to predict ahead of time, but we can 675 monitor the usage as attendees use the different networks. By 676 configuring multiple DHCP pools per subnet, and enabling them 677 sequentially, we can have a large subnet, but only assign 678 addresses from the lower portions of it. This means that we 679 can apply input IP access lists, which deny traffic to the 680 upper, unused portions. This means that the router does not 681 attempt to forward packets to the unused portions of the 682 subnets, and so does not ARP for it. This method has proven to 683 be very effective, but is somewhat of a blunt axe, is fairly 684 labor intensive, and requires coordination. 686 Disabling/filtering ARP requests 688 In general, the router does not need to ARP for hosts; when a 689 host connects, the router can learn the IP to MAC mapping from 690 the ARP request sent by that host. This means that we should 691 be able to disable and / or filter ARP requests from the 692 router. Unfortunately, ARP is a very low level / fundamental 693 part of the IP stack, and is often offloaded from the normal 694 control plane. While many routers can filter layer-2 traffic, 695 this is usually implemented as an input filter and / or has 696 limited ability to filter output broadcast traffic. This means 697 that the simple "just disable ARP or filter it outbound" seems 698 like a really simple (and obvious) solution, but 699 implementations / architectural issues make this difficult or 700 awkward in practice. 702 NAT 704 The broadcasts are overwhelmingly being caused by outside 705 scanning / backscatter traffic. This means that, if we were to 706 NAT the entire (or a large portion) of the attendee networks, 707 there would be no NAT translation entries for unused addresses, 708 and so the router would never ARP for them. The IETF NOC has 709 discussed NATing the entire (or large portions) attendee 710 address space, but a: elegance and b: flaming torches and 711 pitchfork concerns means we have not attempted this yet. 713 Stateful firewalls 714 Another obvious solution would be to put a stateful firewall 715 between the wireless network and the Internet. This firewall 716 would block incoming traffic not associated with an outbound 717 request. The IETF philosophy has been to have the network as 718 open as possible / honor the end-to-end principle. An attendee 719 on the meeting network should be an Internet host, and should 720 be able to receive unsolicited requests. Unfortunately, 721 keeping the network working and stable is the first priority 722 and a stateful firewall may be required in order to achieve 723 this. 725 6. Multicast Considerations for Other Wireless Media 727 Many of the causes of performance degradation described in earlier 728 sections are also observable for wireless media other than 802.11. 730 For instance, problems with power save, excess media occupancy, and 731 poor reliability will also affect 802.15.3 and 802.15.4. However, 732 802.15 media specifications do not include mechanisms similar to 733 those developed for 802.11. In fact, the design philosophy for 734 802.15 is oriented towards minimality, with the result that many such 735 functions would more likely be relegated to operation within higher 736 layer protocols. This leads to a patchwork of non-interoperable and 737 vendor-specific solutions. See [uli] for some additional discussion, 738 and a proposal for a task group to resolve similar issues, in which 739 the multicast problems might be considered for mitigation. 741 7. Recommendations 743 This section will provide some recommendations about the usage and 744 combinations of the multicast enhancements described in Section 4 and 745 Section 5. 747 (FFS) 749 8. Discussion Items 751 This section will suggest some discussion items for further 752 resolution. 754 The IETF may need to decide that broadcast is more expensive so 755 multicast needs to be sent wired. For example, 802.1ak works on 756 ethernet and wifi. 802.1ak has been pulled into 802.1Q as of 802.1Q- 757 2011. 802.1Q-2014 can be looked at here: http://www.ieee802.org/1/ 758 pages/802.1Q-2014.html. If a generic solution is not found, 759 guidelines for multicast over wifi should be established. 761 To provide an idea going forward, perhaps a reliable registration to 762 Layer-2 multicast groups and a reliable multicast operation at 763 Layer-2 could provide a generic solution. There is no need to 764 support 2^24 groups to get solicited node multicast working: it is 765 possible to simply select a number of trailing bits that make sense 766 for a given network size to limit the amount of unwanted deliveries 767 to reasonable levels. IEEE 802.1, 802.11, and 802.15 should be 768 encouraged to revisit L2 multicast issues. In particular, Wi-Fi 769 provides a broadcast service, not a multicast one; at the PHY level, 770 all frames are broadcast except in very unusual cases in which 771 special beamforming transmitters are used. Unicast offers the 772 advantage of being much faster (2 orders of magnitude) and much more 773 reliable (L2 ARQ). 775 9. Security Considerations 777 This document does not introduce any security mechanisms, and does 778 not have affect existing security mechanisms. 780 10. IANA Considerations 782 This document does not specify any IANA actions. 784 11. Acknowledgements 786 This document has benefitted from discussions with the following 787 people, in alphabetical order: Pascal Thubert 789 12. Informative References 791 [arpsponge] 792 Arien Vijn, Steven Bakker, "Arp Sponge", March 2015. 794 [Deri-2010] 795 Deri, L. and J. Gasparakis, "10 Gbit Hardware Packet 796 Filtering Using Commodity Network Adapters", RIPE 61, 797 2010, . 800 [dot11] P802.11, "Part 11: Wireless LAN Medium Access Control 801 (MAC) and Physical Layer (PHY) Specifications", March 802 2012. 804 [dot11-proxyarp] 805 P802.11, "Proxy ARP in 802.11ax", September 2015. 807 [dot11aa] P802.11, "Part 11: Wireless LAN Medium Access Control 808 (MAC) and Physical Layer (PHY) Specifications Amendment 2: 809 MAC Enhancements for Robust Audio Video Streaming", March 810 2012. 812 [I-D.ietf-6lo-ap-nd] 813 Thubert, P., Sarikaya, B., and M. Sethi, "Address 814 Protected Neighbor Discovery for Low-power and Lossy 815 Networks", draft-ietf-6lo-ap-nd-06 (work in progress), 816 February 2018. 818 [I-D.ietf-6lo-backbone-router] 819 Thubert, P., "IPv6 Backbone Router", draft-ietf-6lo- 820 backbone-router-06 (work in progress), February 2018. 822 [I-D.ietf-6lo-rfc6775-update] 823 Thubert, P., Nordmark, E., Chakrabarti, S., and C. 824 Perkins, "Registration Extensions for 6LoWPAN Neighbor 825 Discovery", draft-ietf-6lo-rfc6775-update-21 (work in 826 progress), June 2018. 828 [I-D.ietf-6tisch-architecture] 829 Thubert, P., "An Architecture for IPv6 over the TSCH mode 830 of IEEE 802.15.4", draft-ietf-6tisch-architecture-14 (work 831 in progress), April 2018. 833 [ietf_802-11] 834 Dorothy Stanley, "IEEE 802.11 multicast capabilities", Nov 835 2015. 837 [mc-ack-mux] 838 Yusuke Tanaka et al., "Multiplexing of Acknowledgements 839 for Multicast Transmission", July 2015. 841 [mc-prob-stmt] 842 Mikael Abrahamsson and Adrian Stephens, "Multicast on 843 802.11", March 2015. 845 [mc-props] 846 Adrian Stephens, "IEEE 802.11 multicast properties", March 847 2015. 849 [Oliva2013] 850 de la Oliva, A., Serrano, P., Salvador, P., and A. Banchs, 851 "Performance evaluation of the IEEE 802.11aa multicast 852 mechanisms for video streaming", 2013 IEEE 14th 853 International Symposium on "A World of Wireless, Mobile 854 and Multimedia Networks" (WoWMoM) pp. 1-9, June 2013. 856 [RFC4541] Christensen, M., Kimball, K., and F. Solensky, 857 "Considerations for Internet Group Management Protocol 858 (IGMP) and Multicast Listener Discovery (MLD) Snooping 859 Switches", RFC 4541, DOI 10.17487/RFC4541, May 2006, 860 . 862 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 863 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 864 DOI 10.17487/RFC4861, September 2007, 865 . 867 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 868 Address Autoconfiguration", RFC 4862, 869 DOI 10.17487/RFC4862, September 2007, 870 . 872 [RFC6282] Hui, J., Ed. and P. Thubert, "Compression Format for IPv6 873 Datagrams over IEEE 802.15.4-Based Networks", RFC 6282, 874 DOI 10.17487/RFC6282, September 2011, 875 . 877 [RFC6775] Shelby, Z., Ed., Chakrabarti, S., Nordmark, E., and C. 878 Bormann, "Neighbor Discovery Optimization for IPv6 over 879 Low-Power Wireless Personal Area Networks (6LoWPANs)", 880 RFC 6775, DOI 10.17487/RFC6775, November 2012, 881 . 883 [Tramarin2017] 884 Tramarin, F., Vitturi, S., and M. Luvisotto, "IEEE 802.11n 885 for Distributed Measurement Systems", 2017 IEEE 886 International Instrumentation and Measurement Technology 887 Conference (I2MTC) pp. 1-6, May 2017. 889 [uli] Pat Kinney, "LLC Proposal for 802.15.4", Nov 2015. 891 Authors' Addresses 893 Charles E. Perkins 894 Futurewei Inc. 895 2330 Central Expressway 896 Santa Clara, CA 95050 897 USA 899 Phone: +1-408-330-4586 900 Email: charliep@computer.org 901 Mike McBride 902 Futurewei Inc. 903 2330 Central Expressway 904 Santa Clara, CA 95055 905 USA 907 Email: michael.mcbride@huawei.com 909 Dorothy Stanley 910 Hewlett Packard Enterprise 911 2000 North Naperville Rd. 912 Naperville, IL 60566 913 USA 915 Phone: +1 630 979 1572 916 Email: dstanley@arubanetworks.com 918 Warren Kumari 919 Google 920 1600 Amphitheatre Parkway 921 Mountain View, CA 94043 922 USA 924 Email: warren@kumari.net 926 Juan Carlos Zuniga 927 SIGFOX 928 425 rue Jean Rostand 929 Labege 31670 930 France 932 Email: j.c.zuniga@ieee.org