idnits 2.17.1 draft-ietf-dime-overload-reqs-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 : ---------------------------------------------------------------------------- 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 seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords -- however, there's a paragraph with a matching beginning. Boilerplate error? (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). -- The document date (August 26, 2013) is 3889 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- No issues found here. Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group E. McMurry 3 Internet-Draft B. Campbell 4 Intended status: Informational Tekelec 5 Expires: February 27, 2014 August 26, 2013 7 Diameter Overload Control Requirements 8 draft-ietf-dime-overload-reqs-11 10 Abstract 12 When a Diameter server or agent becomes overloaded, it needs to be 13 able to gracefully reduce its load, typically by informing clients to 14 reduce sending traffic for some period of time. Otherwise, it must 15 continue to expend resources parsing and responding to Diameter 16 messages, possibly resulting in congestion collapse. The existing 17 Diameter mechanisms are not sufficient for this purpose. This 18 document describes the limitations of the existing mechanisms. 19 Requirements for new overload management mechanisms are also 20 provided. 22 Status of this Memo 24 This Internet-Draft is submitted in full conformance with the 25 provisions of BCP 78 and BCP 79. 27 Internet-Drafts are working documents of the Internet Engineering 28 Task Force (IETF). Note that other groups may also distribute 29 working documents as Internet-Drafts. The list of current Internet- 30 Drafts is at http://datatracker.ietf.org/drafts/current/. 32 Internet-Drafts are draft documents valid for a maximum of six months 33 and may be updated, replaced, or obsoleted by other documents at any 34 time. It is inappropriate to use Internet-Drafts as reference 35 material or to cite them other than as "work in progress." 37 This Internet-Draft will expire on February 27, 2014. 39 Copyright Notice 41 Copyright (c) 2013 IETF Trust and the persons identified as the 42 document authors. All rights reserved. 44 This document is subject to BCP 78 and the IETF Trust's Legal 45 Provisions Relating to IETF Documents 46 (http://trustee.ietf.org/license-info) in effect on the date of 47 publication of this document. Please review these documents 48 carefully, as they describe your rights and restrictions with respect 49 to this document. Code Components extracted from this document must 50 include Simplified BSD License text as described in Section 4.e of 51 the Trust Legal Provisions and are provided without warranty as 52 described in the Simplified BSD License. 54 Table of Contents 56 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 57 1.1. Documentation Conventions . . . . . . . . . . . . . . . . 3 58 1.2. Causes of Overload . . . . . . . . . . . . . . . . . . . . 4 59 1.3. Effects of Overload . . . . . . . . . . . . . . . . . . . 5 60 1.4. Overload vs. Network Congestion . . . . . . . . . . . . . 5 61 1.5. Diameter Applications in a Broader Network . . . . . . . . 6 62 2. Overload Control Scenarios . . . . . . . . . . . . . . . . . . 6 63 2.1. Peer to Peer Scenarios . . . . . . . . . . . . . . . . . . 7 64 2.2. Agent Scenarios . . . . . . . . . . . . . . . . . . . . . 9 65 2.3. Interconnect Scenario . . . . . . . . . . . . . . . . . . 12 66 3. Diameter Overload Case Studies . . . . . . . . . . . . . . . . 13 67 3.1. Overload in Mobile Data Networks . . . . . . . . . . . . . 13 68 3.2. 3GPP Study on Core Network Overload . . . . . . . . . . . 15 69 4. Existing Mechanisms . . . . . . . . . . . . . . . . . . . . . 15 70 5. Issues with the Current Mechanisms . . . . . . . . . . . . . . 16 71 5.1. Problems with Implicit Mechanism . . . . . . . . . . . . . 17 72 5.2. Problems with Explicit Mechanisms . . . . . . . . . . . . 17 73 6. Extensibility and Application Independence . . . . . . . . . . 18 74 7. Solution Requirements . . . . . . . . . . . . . . . . . . . . 19 75 7.1. General . . . . . . . . . . . . . . . . . . . . . . . . . 19 76 7.2. Performance . . . . . . . . . . . . . . . . . . . . . . . 20 77 7.3. Heterogeneous Support for Solution . . . . . . . . . . . . 21 78 7.4. Granular Control . . . . . . . . . . . . . . . . . . . . . 21 79 7.5. Priority and Policy . . . . . . . . . . . . . . . . . . . 22 80 7.6. Security . . . . . . . . . . . . . . . . . . . . . . . . . 22 81 7.7. Flexibility and Extensibility . . . . . . . . . . . . . . 23 82 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 24 83 9. Security Considerations . . . . . . . . . . . . . . . . . . . 24 84 9.1. Access Control . . . . . . . . . . . . . . . . . . . . . . 24 85 9.2. Denial-of-Service Attacks . . . . . . . . . . . . . . . . 25 86 9.3. Replay Attacks . . . . . . . . . . . . . . . . . . . . . . 25 87 9.4. Man-in-the-Middle Attacks . . . . . . . . . . . . . . . . 25 88 9.5. Compromised Hosts . . . . . . . . . . . . . . . . . . . . 26 89 10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 26 90 10.1. Normative References . . . . . . . . . . . . . . . . . . . 26 91 10.2. Informative References . . . . . . . . . . . . . . . . . . 26 92 Appendix A. Contributors . . . . . . . . . . . . . . . . . . . . 27 93 Appendix B. Acknowledgements . . . . . . . . . . . . . . . . . . 27 94 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 28 96 1. Introduction 98 A Diameter [RFC6733] node is said to be overloaded when it has 99 insufficient resources to successfully process all of the Diameter 100 requests that it receives. When a node becomes overloaded, it needs 101 to be able to gracefully reduce its load, typically by informing 102 clients to reduce sending traffic for some period of time. 103 Otherwise, it must continue to expend resources parsing and 104 responding to Diameter messages, possibly resulting in congestion 105 collapse. The existing mechanisms provided by Diameter are not 106 sufficient for this purpose. This document describes the limitations 107 of the existing mechanisms, and provides requirements for new 108 overload management mechanisms. 110 This document draws on the work done on SIP overload control 111 ([RFC5390], [RFC6357]) as well as on experience gained via overload 112 handling in Signaling System No. 7 (SS7) networks and studies done by 113 the Third Generation Partnership Project (3GPP) (Section 3). 115 Diameter is not typically an end-user protocol; rather it is 116 generally used as one component in support of some end-user activity. 118 For example, a SIP server might use Diameter to authenticate and 119 authorize user access. Overload in the Diameter backend 120 infrastructure will likely impact the experience observed by the end 121 user in the SIP application. 123 The impact of Diameter overload on the client application (a client 124 application may use the Diameter protocol and other protocols to do 125 its job) is beyond the scope of this document. 127 This document presents non-normative descriptions of causes of 128 overload along with related scenarios and studies. Finally, it 129 offers a set of normative requirements for an improved overload 130 indication mechanism. 132 1.1. Documentation Conventions 134 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 135 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 136 document are to be interpreted as defined in [RFC2119], with the 137 exception that they are not intended for interoperability of 138 implementations. Rather, they are used to describe requirements 139 towards future specifications where the interoperability requirements 140 will be defined. 142 The terms "client", "server", "agent", "node", "peer", "upstream", 143 and "downstream" are used as defined in [RFC6733]. 145 1.2. Causes of Overload 147 Overload occurs when an element, such as a Diameter server or agent, 148 has insufficient resources to successfully process all of the traffic 149 it is receiving. Resources include all of the capabilities of the 150 element used to process a request, including CPU processing, memory, 151 I/O, and disk resources. It can also include external resources such 152 as a database or DNS server, in which case the CPU, processing, 153 memory, I/O, and disk resources of those elements are effectively 154 part of the logical element processing the request. 156 External resources can include upstream Diameter nodes; for example, 157 a Diameter agent can become effectively overloaded if one or more 158 upstream nodes are overloaded. While overload is not the same thing 159 as network congestion, network congestion can reduce a Diameter nodes 160 ability to process and respond to requests, thus contributing to 161 overload. 163 A Diameter node can become overloaded due to request levels that 164 exceed its capacity, a reduction of available resources ( for 165 example, a local or upstream hardware failure) or a combination of 166 the two. 168 Overload can occur for many reasons, including: 170 Inadequate capacity: When designing Diameter networks, that is, 171 application layer multi-node Diameter deployments, it can be very 172 difficult to predict all scenarios that may cause elevated 173 traffic. It may also be more costly to implement support for some 174 scenarios than a network operator may deem worthwhile. This 175 results in the likelihood that a Diameter network will not have 176 adequate capacity to handle all situations. 178 Dependency failures: A Diameter node can become overloaded because a 179 resource on which it is dependent has failed or become overloaded, 180 greatly reducing the logical capacity of the node. In these 181 cases, even minimal traffic might cause the node to go into 182 overload. Examples of such dependency overloads include DNS 183 servers, databases, disks, and network interfaces. 185 Component failures: A Diameter node can become overloaded when it is 186 a member of a cluster of servers that each share the load of 187 traffic, and one or more of the other members in the cluster fail. 188 In this case, the remaining nodes take over the work of the failed 189 nodes. Normally, capacity planning takes such failures into 190 account, and servers are typically run with enough spare capacity 191 to handle failure of another node. However, unusual failure 192 conditions can cause many nodes to fail at once. This is often 193 the case with software failures, where a bad packet or bad 194 database entry hits the same bug in a set of nodes in a cluster. 196 Network Initiated Traffic Flood: Issues with the radio access 197 network in a mobile network such as radio overlays with frequent 198 handovers, and operational changes are examples of network events 199 that can precipitate a flood of Diameter signaling traffic, such 200 as an avalanche restart. Failure of a Diameter proxy may also 201 result in a large amount of signaling as connections and sessions 202 are reestablished. 204 Subscriber Initiated Traffic Flood: Large gatherings of subscribers 205 or events that result in many subscribers interacting with the 206 network in close time proximity can result in Diameter signaling 207 traffic floods. For example, the finale of a large fireworks show 208 could be immediately followed by many subscribers posting 209 messages, pictures, and videos concentrated on one portion of a 210 network. Subscriber devices, such as smartphones, may use 211 aggressive registration strategies that generate unusually high 212 Diameter traffic loads. 214 DoS attacks: An attacker, wishing to disrupt service in the network, 215 can cause a large amount of traffic to be launched at a target 216 element. This can be done from a central source of traffic or 217 through a distributed DoS attack. In all cases, the volume of 218 traffic well exceeds the capacity of the element, sending the 219 system into overload. 221 1.3. Effects of Overload 223 Modern Diameter networks, comprised of application layer multi-node 224 deployments of Diameter elements, may operate at very large 225 transaction volumes. If a Diameter node becomes overloaded, or even 226 worse, fails completely, a large number of messages may be lost very 227 quickly. Even with redundant servers, many messages can be lost in 228 the time it takes for failover to complete. While a Diameter client 229 or agent should be able to retry such requests, an overloaded peer 230 may cause a sudden large increase in the number of transaction 231 transactions needing to be retried, rapidly filling local queues or 232 otherwise contributing to local overload. Therefore Diameter devices 233 need to be able to shed load before critical failures can occur. 235 1.4. Overload vs. Network Congestion 237 This document uses the term "overload" to refer to application-layer 238 overload at Diameter nodes. This is distinct from "network 239 congestion", that is, congestion that occurs at the lower networking 240 layers that may impact the delivery of Diameter messages between 241 nodes. The authors recognize that element overload and network 242 congestion are interrelated, and that overload can contribute to 243 network congestion and vice versa. 245 Network congestion issues are better handled by the transport 246 protocols. Diameter uses TCP and SCTP, both of which include 247 congestion management features. Analysis of whether those features 248 are sufficient for transport level congestion between Diameter nodes, 249 and any work to further mitigate network congestion is out of scope 250 both for this document, and for the work proposed by this document. 252 1.5. Diameter Applications in a Broader Network 254 Most elements using Diameter applications do not use Diameter 255 exclusively. It is important to realize that overload of an element 256 can be caused by a number of factors that may be unrelated to the 257 processing of Diameter or Diameter applications. 259 An element that doesn't use Diameter exclusively needs to be able to 260 signal to Diameter peers that it is experiencing overload regardless 261 of the cause of the overload, since the overload will affect that 262 element's ability to process Diameter transactions. If the element 263 communicates with other protocols than Diameter, it may also need to 264 signal the overload situation on these protocols depending on its 265 function and the architecture of the network and application it is 266 providing services for. Whether that is necessary can only be 267 decided within the context of that architecture and use cases. A 268 mechanism for signaling overload with Diameter, which this 269 specification details the requirements for, provides Diameter nodes 270 the ability to signal their Diameter peers of overload, mitigating 271 that part of the issue. Diameter nodes may need to use this, as well 272 as other mechanisms, to solve their broader overload issues. 273 Indicating overload on protocols other than Diameter is out of scope 274 for this document, and for the work proposed by this document. 276 2. Overload Control Scenarios 278 Several Diameter deployment scenarios exist that may impact overload 279 management. The following scenarios help motivate the requirements 280 for an overload management mechanism. 282 These scenarios are by no means exhaustive, and are in general 283 simplified for the sake of clarity. In particular, the authors 284 assume for the sake of clarity that the client sends Diameter 285 requests to the server, and the server sends responses to client, 286 even though Diameter supports bidirectional applications. Each 287 direction in such an application can be modeled separately. 289 In a large scale deployment, many of the nodes represented in these 290 scenarios would be deployed as clusters of servers. The authors 291 assume that such a cluster is responsible for managing its own 292 internal load balancing and overload management so that it appears as 293 a single Diameter node. That is, other Diameter nodes can treat it 294 as single, monolithic node for the purposes of overload management. 296 These scenarios do not illustrate the client application. As 297 mentioned in Section 1, Diameter is not typically an end-user 298 protocol; rather it is generally used in support of some other client 299 application. These scenarios do not consider the impact of Diameter 300 overload on the client application. 302 2.1. Peer to Peer Scenarios 304 This section describes Diameter peer-to-peer scenarios. That is, 305 scenarios where a Diameter client talks directly with a Diameter 306 server, without the use of a Diameter agent. 308 Figure 1 illustrates the simplest possible Diameter relationship. 309 The client and server share a one-to-one peer-to-peer relationship. 310 If the server becomes overloaded, either because the client exceeds 311 the server's capacity, or because the server's capacity is reduced 312 due to some resource dependency, the client needs to reduce the 313 amount of Diameter traffic it sends to the server. Since the client 314 cannot forward requests to another server, it must either queue 315 requests until the server recovers, or itself become overloaded in 316 the context of the client application and other protocols it may also 317 use. 319 +------------------+ 320 | | 321 | | 322 | Server | 323 | | 324 +--------+---------+ 325 | 326 | 327 +--------+---------+ 328 | | 329 | | 330 | Client | 331 | | 332 +------------------+ 334 Figure 1: Basic Peer to Peer Scenario 336 Figure 2 shows a similar scenario, except in this case the client has 337 multiple servers that can handle work for a specific realm and 338 application. If server 1 becomes overloaded, the client can forward 339 traffic to server 2. Assuming server 2 has sufficient reserve 340 capacity to handle the forwarded traffic, the client should be able 341 to continue serving client application protocol users. If server 1 342 is approaching overload, but can still handle some number of new 343 request, it needs to be able to instruct the client to forward a 344 subset of its traffic to server 2. 346 +------------------+ +------------------+ 347 | | | | 348 | | | | 349 | Server 1 | | Server 2 | 350 | | | | 351 +--------+-`.------+ +------.'+---------+ 352 `. .' 353 `. .' 354 `. .' 355 `. .' 356 +-------`.'--------+ 357 | | 358 | | 359 | Client | 360 | | 361 +------------------+ 363 Figure 2: Multiple Server Peer to Peer Scenario 365 Figure 3 illustrates a peer-to-peer scenario with multiple Diameter 366 realm and application combinations. In this example, server 2 can 367 handle work for both applications. Each application might have 368 different resource dependencies. For example, a server might need to 369 access one database for application A, and another for application B. 370 This creates a possibility that Server 2 could become overloaded for 371 application A but not for application B, in which case the client 372 would need to divert some part of its application A requests to 373 server 1, but should not divert any application B requests. This 374 requires server 2 to be able to distinguish between applications when 375 it indicates an overload condition to the client. 377 On the other hand, it's possible that the servers host many 378 applications. If server 2 becomes overloaded for all applications, 379 it would be undesirable for it to have to notify the client 380 separately for each application. Therefore it also needs a way to 381 indicate that it is overloaded for all possible applications. 383 +---------------------------------------------+ 384 | Application A +----------------------+----------------------+ 385 |+------------------+ | +----------------+ | +------------------+| 386 || | | | | | | || 387 || | | | | | | || 388 || Server 1 | | | Server 2 | | | Server 3 || 389 || | | | | | | || 390 |+--------+---------+ | +-------+--------+ | +-+----------------+| 391 | | | | | | | 392 +---------+-----------+----------+-----------+ | | 393 | | | | | 394 | | | | Application B | 395 | +----------+----------------+-----------------+ 396 ``-.._ | | 397 `-..__ | _.-'' 398 `--._ | _.-'' 399 ``-._ | _.-'' 400 +-----`-.-''-----+ 401 | | 402 | | 403 | Client | 404 | | 405 +----------------+ 407 Figure 3: Multiple Application Peer to Peer Scenario 409 2.2. Agent Scenarios 411 This section describes scenarios that include a Diameter agent, 412 either in the form of a Diameter relay or Diameter proxy. These 413 scenarios do not consider Diameter redirect agents, since they are 414 more readily modeled as end-servers. 416 Figure 4 illustrates a simple Diameter agent scenario with a single 417 client, agent, and server. In this case, overload can occur at the 418 server, at the agent, or both. But in most cases, client behavior is 419 the same whether overload occurs at the server or at the agent. From 420 the client's perspective, server overload and agent overload is the 421 same thing. 423 +------------------+ 424 | | 425 | | 426 | Server | 427 | | 428 +--------+---------+ 429 | 430 | 431 +--------+---------+ 432 | | 433 | | 434 | Agent | 435 | | 436 +--------+---------+ 437 | 438 | 439 +--------+---------+ 440 | | 441 | | 442 | Client | 443 | | 444 +------------------+ 446 Figure 4: Basic Agent Scenario 448 Figure 5 shows an agent scenario with multiple servers. If server 1 449 becomes overloaded, but server 2 has sufficient reserve capacity, the 450 agent may be able to transparently divert some or all Diameter 451 requests originally bound for server 1 to server 2. 453 In most cases, the client does not have detailed knowledge of the 454 Diameter topology upstream of the agent. If the agent uses dynamic 455 discovery to find eligible servers, the set of eligible servers may 456 not be enumerable from the perspective of the client. Therefore, in 457 most cases the agent needs to deal with any upstream overload issues 458 in a way that is transparent to the client. If one server notifies 459 the agent that it has become overloaded, the notification should not 460 be passed back to the client in a way that the client could 461 mistakenly perceive the agent itself as being overloaded. If the set 462 of all possible destinations upstream of the agent no longer has 463 sufficient capacity for incoming load, the agent itself becomes 464 effectively overloaded. 466 On the other hand, there are cases where the client needs to be able 467 to select a particular server from behind an agent. For example, if 468 a Diameter request is part of a multiple-round-trip authentication, 469 or is otherwise part of a Diameter "session", it may have a 470 DestinationHost AVP that requires the request to be served by server 471 1. Therefore the agent may need to inform a client that a particular 472 upstream server is overloaded or otherwise unavailable. Note that 473 there can be many ways a server can be specified, which may have 474 different implications (e.g. by IP address, by host name, etc). 476 +------------------+ +------------------+ 477 | | | | 478 | | | | 479 | Server 1 | | Server 2 | 480 | | | | 481 +--------+-`.------+ +------.'+---------+ 482 `. .' 483 `. .' 484 `. .' 485 `. .' 486 +-------`.'--------+ 487 | | 488 | | 489 | Agent | 490 | | 491 +--------+---------+ 492 | 493 | 494 | 495 +--------+---------+ 496 | | 497 | | 498 | Client | 499 | | 500 +------------------+ 502 Figure 5: Multiple Server Agent Scenario 504 Figure 6 shows a scenario where an agent routes requests to a set of 505 servers for more than one Diameter realm and application. In this 506 scenario, if server 1 becomes overloaded or unavailable while server 507 2 still has available capacity, the agent may effectively operate at 508 reduced capacity for application A, but at full capacity for 509 application B. Therefore, the agent needs to be able to report that 510 it is overloaded for one application, but not for another. 512 +--------------------------------------------+ 513 | Application A +----------------------+----------------------+ 514 |+------------------+ | +----------------+ | +------------------+| 515 || | | | | | | || 516 || | | | | | | || 517 || Server 1 | | | Server 2 | | | Server 3 || 518 || | | | | | | || 519 |+---------+--------+ | +-------+--------+ | +--+---------------+| 520 | | | | | | | 521 +----------+----------+----------+-----------+ | | 522 | | | | | 523 | | | | Application B | 524 | +----------+-----------------+----------------+ 525 | | | 526 ``--.__ | _. 527 ``-.__ | __.--'' 528 `--.._ | _..--' 529 +----``-+.''-----+ 530 | | 531 | | 532 | Agent | 533 | | 534 +-------+--------+ 535 | 536 | 537 +-------+--------+ 538 | | 539 | | 540 | Client | 541 | | 542 +----------------+ 544 Figure 6: Multiple Application Agent Scenario 546 2.3. Interconnect Scenario 548 Another scenario to consider when looking at Diameter overload is 549 that of multiple network operators using Diameter components 550 connected through an interconnect service, e.g. using IPX. IPX (IP 551 eXchange) [IR.34] is an Inter-Operator IP Backbone that provides 552 roaming interconnection network between mobile operators and service 553 providers. The IPX is also used to transport Diameter signaling 554 between operators [IR.88]. Figure 7 shows two network operators with 555 an interconnect network in-between. There could be any number of 556 these networks between any two network operator's networks. 558 +-------------------------------------------+ 559 | Interconnect | 560 | | 561 | +--------------+ +--------------+ | 562 | | Server 3 |------| Server 4 | | 563 | +--------------+ +--------------+ | 564 | .' `. | 565 +------.-'--------------------------`.------+ 566 .' `. 567 .-' `. 568 ------------.'-----+ +----`.------------- 569 +----------+ | | +----------+ 570 | Server 1 | | | | Server 2 | 571 +----------+ | | +----------+ 572 | | 573 Network Operator 1 | | Network Operator 2 574 -------------------+ +------------------- 576 Figure 7: Two Network Interconnect Scenario 578 The characteristics of the information that an operator would want to 579 share over such a connection are different from the information 580 shared between components within a network operator's network. 581 Network operators may not want to convey topology or operational 582 information, which limits how much overload and loading information 583 can be sent. For the interconnect scenario shown, Server 2 may want 584 to signal overload to Server 1, to affect traffic coming from Network 585 Operator 1. 587 This case is distinct from those internal to a network operator's 588 network, where there may be many more elements in a more complicated 589 topology. Also, the elements in the interconnect network may not 590 support Diameter overload control, and the network operators may not 591 want the interconnect network to use overload or loading information. 592 They may only want the information to pass through the interconnect 593 network without further processing or action by the interconnect 594 network even if the elements in the interconnect network do support 595 Diameter overload control. 597 3. Diameter Overload Case Studies 599 3.1. Overload in Mobile Data Networks 601 As the number of Third Generation (3G) and Long Term Evolution (LTE) 602 enabled smartphone devices continue to expand in mobile networks, 603 there have been situations where high signaling traffic load led to 604 overload events at the Diameter-based Home Location Registries (HLR) 605 and/or Home Subscriber Servers (HSS) [TR23.843]. The root causes of 606 the HLR congestion events were manifold but included hardware failure 607 and procedural errors. The result was high signaling traffic load on 608 the HLR and HSS. 610 The 3GPP architecture [TS23.002] makes extensive use of Diameter. It 611 is used for mobility management [TS29.272] (and others), (IP 612 Multimedia Subsystem) IMS [TS29.228] (and others), policy and 613 charging control [TS29.212] (and others) as well as other functions. 614 The details of the architecture are out of scope for this document, 615 but it is worth noting that there are quite a few Diameter 616 applications, some with quite large amounts of Diameter signaling in 617 deployed networks. 619 The 3GPP specifications do not currently address overload for 620 Diameter applications or provide an equivalent load control mechanism 621 to those provided in the more traditional SS7 elements in (Global 622 System for Mobile Communications) GSM [TS29.002]. The capabilities 623 specified in the 3GPP standards do not adequately address the 624 abnormal condition where excessively high signaling traffic load 625 situations are experienced. 627 Smartphones, an increasingly large percentage of mobile devices, 628 contribute much more heavily, relative to non-smartphones, to the 629 continuation of a registration surge due to their very aggressive 630 registration algorithms. Smartphone behavior contributes to network 631 loading and can contribute to overload conditions. The aggressive 632 smartphone logic is designed to: 634 a. always have voice and data registration, and 636 b. constantly try to be on 3G or LTE data (and thus on 3G voice or 637 VoLTE [IR.92]) for their added benefits. 639 Non-smartphones typically have logic to wait for a time period after 640 registering successfully on voice and data. 642 The smartphone aggressive registration is problematic in two ways: 644 o first by generating excessive signaling load towards the HSS that 645 is ten times that from a non-smartphone, 647 o and second by causing continual registration attempts when a 648 network failure affects registrations through the 3G data network. 650 3.2. 3GPP Study on Core Network Overload 652 A study in 3GPP SA2 on core network overload has produced the 653 technical report [TR23.843]. This enumerates several causes of 654 overload in mobile core networks including portions that are signaled 655 using Diameter. TR23.843 is a work in progress and is not complete. 656 However, it is useful for pointing out scenarios and the general need 657 for an overload control mechanism for Diameter. 659 It is common for mobile networks to employ more than one radio 660 technology and to do so in an overlay fashion with multiple 661 technologies present in the same location (such as 2nd or 3rd 662 generation mobile technologies along with LTE). This presents 663 opportunities for traffic storms when issues occur on one overlay and 664 not another as all devices that had been on the overlay with issues 665 switch. This causes a large amount of Diameter traffic as locations 666 and policies are updated. 668 Another scenario called out by this study is a flood of registration 669 and mobility management events caused by some element in the core 670 network failing. This flood of traffic from end nodes falls under 671 the network initiated traffic flood category. There is likely to 672 also be traffic resulting directly from the component failure in this 673 case. A similar flood can occur when elements or components recover 674 as well. 676 Subscriber initiated traffic floods are also indicated in this study 677 as an overload mechanism where a large number of mobile devices 678 attempting to access services at the same time, such as in response 679 to an entertainment event or a catastrophic event. 681 While this 3GPP study is concerned with the broader effects of these 682 scenarios on wireless networks and their elements, they have 683 implications specifically for Diameter signaling. One of the goals 684 of this document is to provide guidance for a core mechanism that can 685 be used to mitigate the scenarios called out by this study. 687 4. Existing Mechanisms 689 Diameter offers both implicit and explicit mechanisms for a Diameter 690 node to learn that a peer is overloaded or unreachable. The implicit 691 mechanism is simply the lack of responses to requests. If a client 692 fails to receive a response in a certain time period, it assumes the 693 upstream peer is unavailable, or overloaded to the point of effective 694 unavailability. The watchdog mechanism [RFC3539] ensures that a 695 certain rate of transaction responses occur even when there is 696 otherwise little or no other Diameter traffic. 698 The explicit mechanism can involve specific protocol error responses, 699 where an agent or server tells a downstream peer that it is either 700 too busy to handle a request (DIAMETER_TOO_BUSY) or unable to route a 701 request to an upstream destination (DIAMETER_UNABLE_TO_DELIVER), 702 perhaps because that destination itself is overloaded to the point of 703 unavailability. 705 Another explicit mechanism, a DPR (Disconnect-Peer-Request) message, 706 can be sent with a Disconnect-Cause of BUSY. This signals the 707 sender's intent to close the transport connection, and requests the 708 client not to reconnect. 710 Once a Diameter node learns that an upstream peer has become 711 overloaded via one of these mechanisms, it can then attempt to take 712 action to reduce the load. This usually means forwarding traffic to 713 an alternate destination, if available. If no alternate destination 714 is available, the node must either reduce the number of messages it 715 originates (in the case of a client) or inform the client to reduce 716 traffic (in the case of an agent.) 718 Diameter requires the use of a congestion-managed transport layer, 719 currently TCP or SCTP, to mitigate network congestion. It is 720 expected that these transports manage network congestion and that 721 issues with transport (e.g. congestion propagation and window 722 management) are managed at that level. But even with a congestion- 723 managed transport, a Diameter node can become overloaded at the 724 Diameter protocol or application layers due to the causes described 725 in Section 1.2 and congestion managed transports do not provide 726 facilities (and are at the wrong level) to handle server overload. 727 Transport level congestion management is also not sufficient to 728 address overload in cases of multi-hop and multi-destination 729 signaling. 731 5. Issues with the Current Mechanisms 733 The currently available Diameter mechanisms for indicating an 734 overload condition are not adequate to avoid service outages due to 735 overload. This inadequacy may, in turn, contribute to broader 736 congestion impacts due to unresponsive Diameter nodes causing 737 application or transport layer retransmissions. In particular, they 738 do not allow a Diameter agent or server to shed load as it approaches 739 overload. At best, a node can only indicate that it needs to 740 entirely stop receiving requests, i.e. that it has effectively 741 failed. Even that is problematic due to the inability to indicate 742 durational validity on the transient errors available in the base 743 Diameter protocol. Diameter offers no mechanism to allow a node to 744 indicate different overload states for different categories of 745 messages, for example, if it is overloaded for one Diameter 746 application but not another. 748 5.1. Problems with Implicit Mechanism 750 The implicit mechanism doesn't allow an agent or server to inform the 751 client of a problem until it is effectively too late to do anything 752 about it. The client does not know to take action until the upstream 753 node has effectively failed. A Diameter node has no opportunity to 754 shed load early to avoid collapse in the first place. 756 Additionally, the implicit mechanism cannot distinguish between 757 overload of a Diameter node and network congestion. Diameter treats 758 the failure to receive an answer as a transport failure. 760 5.2. Problems with Explicit Mechanisms 762 The Diameter specification is ambiguous on how a client should handle 763 receipt of a DIAMETER_TOO_BUSY response. The base specification 764 [RFC6733] indicates that the sending client should attempt to send 765 the request to a different peer. It makes no suggestion that the 766 receipt of a DIAMETER_TOO_BUSY response should affect future Diameter 767 messages in any way. 769 The Authentication, Authorization, and Accounting (AAA) Transport 770 Profile [RFC3539] recommends that a AAA node that receives a "Busy" 771 response failover all remaining requests to a different agent or 772 server. But while the Diameter base specification explicitly depends 773 on RFC3539 to define transport behavior, it does not refer to RFC3539 774 in the description of behavior on receipt of DIAMETER_TOO_BUSY. 775 There's a strong likelihood that at least some implementations will 776 continue to send Diameter requests to an upstream peer even after 777 receiving a DIAMETER_TOO_BUSY error. 779 BCP 41 [RFC2914] describes, among other things, how end-to-end 780 application behavior can help avoid congestion collapse. In 781 particular, an application should avoid sending messages that will 782 never be delivered or processed. The DIAMETER_TOO_BUSY behavior as 783 described in the Diameter base specification fails at this, since if 784 an upstream node becomes overloaded, a client attempts each request, 785 and does not discover the need to failover the request until the 786 initial attempt fails. 788 The situation is improved if implementations follow the [RFC3539] 789 recommendation and keep state about upstream peer overload. But even 790 then, the Diameter specification offers no guidance on how long a 791 client should wait before retrying the overloaded destination. If an 792 agent or server supports multiple realms and/or applications, 793 DIAMETER_TOO_BUSY offers no way to indicate that it is overloaded for 794 one application but not another. A DIAMETER_TOO_BUSY error can only 795 indicate overload at a "whole server" scope. 797 Agent processing of a DIAMETER_TOO_BUSY response is also problematic 798 as described in the base specification. DIAMETER_TOO_BUSY is defined 799 as a protocol error. If an agent receives a protocol error, it may 800 either handle it locally or it may forward the response back towards 801 the downstream peer. If a downstream peer receives the 802 DIAMETER_TOO_BUSY response, it may stop sending all requests to the 803 agent for some period of time, even though the agent may still be 804 able to deliver requests to other upstream peers. 806 DIAMETER_UNABLE_TO_DELIVER, or using DPR with cause code BUSY also 807 have no mechanisms for specifying the scope or cause of the failure, 808 or the durational validity. 810 The issues with error responses in [RFC6733] extend beyond the 811 particular issues for overload control and have been addressed in an 812 ad hoc fashion by various implementations. Addressing these in a 813 standard way would be a useful exercise, but it us beyond the scope 814 of this document. 816 6. Extensibility and Application Independence 818 Given the variety of scenarios Diameter elements can be deployed in, 819 and the variety of roles they can fulfill with Diameter and other 820 technologies, a single algorithm for handling overload may not be 821 sufficient. For purposes of this discussion, algorithm is inclusive 822 of behavior for control of overload, but does not encompass the 823 general mechanism or transport of control information. This effort 824 cannot anticipate all possible future scenarios and roles. 825 Extensibility, particularly of algorithms used to deal with overload, 826 will be important to cover these cases. 828 Similarly, the scopes that overload information may apply to may 829 include cases that have not yet been considered. Extensibility in 830 this area will also be important. 832 The basic mechanism is intended to be application-independent, that 833 is, a Diameter node can use it across any existing and future 834 Diameter applications and expect reasonable results. Certain 835 Diameter applications might, however, benefit from application- 836 specific behavior over and above the mechanism's defaults. For 837 example, an application specification might specify relative 838 priorities of messages or selection of a specific overload control 839 algorithm. 841 7. Solution Requirements 843 This section proposes requirements for an improved mechanism to 844 control Diameter overload, with the goals of improving the issues 845 described in Section 5 and supporting the scenarios described in 846 Section 2. These requirements are stated primarily in terms of 847 individual node behavior to inform the design of the improved 848 mechanism; solution designers should keep in mind that the overall 849 goal is improved overall system behavior across all the nodes 850 involved, not just improved behavior from specific individual nodes. 852 7.1. General 854 REQ 1: The solution MUST provide a communication method for Diameter 855 nodes to exchange load and overload information. 857 REQ 2: The solution MUST allow Diameter nodes to support overload 858 control regardless of which Diameter applications they 859 support. Diameter clients and agents must be able to use the 860 received load and overload information to support graceful 861 behavior during an overload condition. Graceful behavior 862 under overload conditions is best described by REQ 3. 864 REQ 3: The solution MUST limit the impact of overload on the overall 865 useful throughput of a Diameter server, even when the 866 incoming load on the network is far in excess of its 867 capacity. The overall useful throughput under load is the 868 ultimate measure of the value of a solution. 870 REQ 4: Diameter allows requests to be sent from either side of a 871 connection and either side of a connection may have need to 872 provide its overload status. The solution MUST allow each 873 side of a connection to independently inform the other of its 874 overload status. 876 REQ 5: Diameter allows nodes to determine their peers via dynamic 877 discovery or manual configuration. The solution MUST work 878 consistently without regard to how peers are determined. 880 REQ 6: The solution designers SHOULD seek to minimize the amount of 881 new configuration required in order to work. For example, it 882 is better to allow peers to advertise or negotiate support 883 for the solution, rather than to require this knowledge to be 884 configured at each node. 886 7.2. Performance 888 REQ 7: The solution and any associated default algorithm(s) MUST 889 ensure that the system remains stable. At some point after 890 an overload condition has ended, the solution MUST enable 891 capacity to stabilize and become equal to what it would be 892 in the absence of an overload condition. Note that this 893 also requires that the solution MUST allow nodes to shed 894 load without introducing non converging oscillations during 895 or after an overload condition. 897 REQ 8: Supporting nodes MUST be able to distinguish current 898 overload information from stale information. 900 REQ 9: The solution MUST function across fully loaded as well as 901 quiescent transport connections. This is partially derived 902 from the requirement for stability in REQ 7. 904 REQ 10: Consumers of overload information MUST be able to determine 905 when the overload condition improves or ends. 907 REQ 11: The solution MUST be able to operate in networks of 908 different sizes. 910 REQ 12: When a single network node fails, goes into overload, or 911 suffers from reduced processing capacity, the solution MUST 912 make it possible to limit the impact of this on other nodes 913 in the network. This helps to prevent a small-scale failure 914 from becoming a widespread outage. 916 REQ 13: The solution MUST NOT introduce substantial additional work 917 for node in an overloaded state. For example, a requirement 918 for an overloaded node to send overload information every 919 time it received a new request would introduce substantial 920 work. Existing messaging is likely to have the 921 characteristic of increasing as an overload condition 922 approaches, allowing for the possibility of increased 923 feedback for information piggybacked on it. 925 REQ 14: Some scenarios that result in overload involve a rapid 926 increase of traffic with little time between normal levels 927 and overload inducing levels. The solution SHOULD provide 928 for rapid feedback when traffic levels increase. 930 REQ 15: The solution MUST NOT interfere with the congestion control 931 mechanisms of underlying transport protocols. For example, 932 a solution that opened additional TCP connections when the 933 network is congested would reduce the effectiveness of the 934 underlying congestion control mechanisms. 936 7.3. Heterogeneous Support for Solution 938 REQ 16: The solution is likely to be deployed incrementally. The 939 solution MUST support a mixed environment where some, but 940 not all, nodes implement it. 942 REQ 17: In a mixed environment with nodes that support the solution 943 and that do not, the solution MUST NOT result in materially 944 less useful throughput as would have resulted if the 945 solution were not present. It SHOULD result in less severe 946 congestion in this environment. 948 REQ 18: In a mixed environment of nodes that support the solution 949 and that do not, the solution MUST NOT preclude elements 950 that support overload control from treating elements that do 951 not support overload control in a equitable fashion relative 952 to those that do. Users and operators of nodes that do not 953 support the solution MUST NOT unfairly benefit from the 954 solution. The solution specification SHOULD provide 955 guidance to implementors for dealing with elements not 956 supporting overload control. 958 REQ 19: It MUST be possible to use the solution between nodes in 959 different realms and in different administrative domains. 961 REQ 20: Any explicit overload indication MUST be clearly 962 distinguishable from other errors reported via Diameter. 964 REQ 21: In cases where a network node fails, is so overloaded that 965 it cannot process messages, or cannot communicate due to a 966 network failure, it may not be able to provide explicit 967 indications of the nature of the failure or its levels of 968 congestion. The solution MUST result in at least as much 969 useful throughput as would have resulted if the solution was 970 not in place. 972 7.4. Granular Control 973 REQ 22: The solution MUST provide a way for a node to throttle the 974 amount of traffic it receives from a peer node. This 975 throttling SHOULD be graded so that it can be applied 976 gradually as offered load increases. Overload is not a 977 binary state; there may be degrees of overload. 979 REQ 23: The solution MUST provide sufficient information to enable a 980 load balancing node to divert messages that are rejected or 981 otherwise throttled by an overloaded upstream node to other 982 upstream nodes that are the most likely to have sufficient 983 capacity to process them. 985 REQ 24: The solution MUST provide a mechanism for indicating load 986 levels even when not in an overloaded condition, to assist 987 nodes making decisions to prevent overload conditions from 988 occurring. 990 7.5. Priority and Policy 992 REQ 25: The base specification for the solution SHOULD offer general 993 guidance on which message types might be desirable to send 994 or process over others during times of overload, based on 995 application-specific considerations. For example, it may be 996 more beneficial to process messages for existing sessions 997 ahead of new sessions. Some networks may have a requirement 998 to give priority to requests associated with emergency 999 sessions. Any normative or otherwise detailed definition of 1000 the relative priorities of message types during an overload 1001 condition will be the responsibility of the application 1002 specification. 1004 REQ 26: The solution MUST NOT prevent a node from prioritizing 1005 requests based on any local policy, so that certain requests 1006 are given preferential treatment, given additional 1007 retransmission, not throttled, or processed ahead of others. 1009 7.6. Security 1011 REQ 27: The solution MUST NOT provide new vulnerabilities to 1012 malicious attack, or increase the severity of any existing 1013 vulnerabilities. This includes vulnerabilities to DoS and 1014 DDoS attacks as well as replay and man-in-the middle 1015 attacks. Note that the Diameter base specification 1016 [RFC6733] lacks end to end security and this must be 1017 considered. Note that this requirement was expressed at a 1018 high level so as to not preclude any particular solution. 1020 Is is expected that the solution will address this in more 1021 detail. 1023 REQ 28: The solution MUST NOT depend on being deployed in 1024 environments where all Diameter nodes are completely 1025 trusted. It SHOULD operate as effectively as possible in 1026 environments where other nodes are malicious; this includes 1027 preventing malicious nodes from obtaining more than a fair 1028 share of service. Note that this does not imply any 1029 responsibility on the solution to detect, or take 1030 countermeasures against, malicious nodes. 1032 REQ 29: It MUST be possible for a supporting node to make 1033 authorization decisions about what information will be sent 1034 to peer nodes based on the identity of those nodes. This 1035 allows a domain administrator who considers the load of 1036 their nodes to be sensitive information to restrict access 1037 to that information. Of course, in such cases, there is no 1038 expectation that the solution itself will help prevent 1039 overload from that peer node. 1041 REQ 30: The solution MUST NOT interfere with any Diameter compliant 1042 method that a node may use to protect itself from overload 1043 from non-supporting nodes, or from denial of service 1044 attacks. 1046 7.7. Flexibility and Extensibility 1048 REQ 31: There are multiple situations where a Diameter node may be 1049 overloaded for some purposes but not others. For example, 1050 this can happen to an agent or server that supports multiple 1051 applications, or when a server depends on multiple external 1052 resources, some of which may become overloaded while others 1053 are fully available. The solution MUST allow Diameter nodes 1054 to indicate overload with sufficient granularity to allow 1055 clients to take action based on the overloaded resources 1056 without unreasonably forcing available capacity to go 1057 unused. The solution MUST support specification of overload 1058 information with granularities of at least "Diameter node", 1059 "realm", and "Diameter application", and MUST allow 1060 extensibility for others to be added in the future. 1062 REQ 32: The solution MUST provide a method for extending the 1063 information communicated and the algorithms used for 1064 overload control. 1066 REQ 33: The solution MUST provide a default algorithm that is 1067 mandatory to implement. 1069 REQ 34: The solution SHOULD provide a method for exchanging overload 1070 and load information between elements that are connected by 1071 intermediaries that do not support the solution. 1073 8. IANA Considerations 1075 This document makes no requests of IANA. 1077 9. Security Considerations 1079 A Diameter overload control mechanism is primarily concerned with the 1080 load and overload related behavior of nodes in a Diameter network, 1081 and the information used to affect that behavior. Load and overload 1082 information is shared between nodes and directly affects the behavior 1083 and thus is potentially vulnerable to a number of methods of attack. 1085 Load and overload information may also be sensitive from both 1086 business and network protection viewpoints. Operators of Diameter 1087 equipment want to control visibility to load and overload information 1088 to keep it from being used for competitive intelligence or for 1089 targeting attacks. It is also important that the Diameter overload 1090 control mechanism not introduce any way in which any other 1091 information carried by Diameter is sent inappropriately. 1093 Note that the Diameter base specification [RFC6733] lacks end to end 1094 security, making verifying the authenticity and ownership of load and 1095 overload information difficult for non-adjacent nodes. 1096 Authentication of load and overload information helps to alleviate 1097 several of the security issues listed in this section. 1099 This document includes requirements intended to mitigate the effects 1100 of attacks and to protect the information used by the mechanism. 1101 This section discusses potential security considerations for overload 1102 control solutions. This discussion provides the motivation for 1103 several normative requirements described in Section 7. The 1104 discussion includes specific references to the normative requirements 1105 that apply for each issue. 1107 9.1. Access Control 1109 To control the visibility of load and overload information, sending 1110 should be subject to some form of authentication and authorization of 1111 the receiver. It is also important to the receivers that they are 1112 confident the load and overload information they receive is from a 1113 legitimate source. REQ 28 requires the solution to work without 1114 assuming that all Diameter nodes in a network are trusted for the 1115 purposes of exchanging overload and load information. REQ 29 1116 requires the solution to let nodes restrict unauthorized parties from 1117 seeing overload information. Note that this implies a certain amount 1118 of configurability on the nodes supporting the Diameter overload 1119 control mechanism. 1121 9.2. Denial-of-Service Attacks 1123 An overload control mechanism provides a very attractive target for 1124 denial-of-service attacks. A small number of messages may affect a 1125 large service disruption by falsely reporting overload conditions. 1126 Alternately, attacking servers nearing, or in, overload may also be 1127 facilitated by disrupting their overload indications, potentially 1128 preventing them from mitigating their overload condition. 1130 A design goal for the Diameter overload control mechanism is to 1131 minimize or eliminate the possibility of using the mechanism for this 1132 type of attack. More strongly, REQ 27 forbids the solution from 1133 introducing new vulnerabilities to malicious attack. Additionally, 1134 REQ 30 stipulates that the solution not interfere with other 1135 mechanisms used for protection against denial of service attacks. 1137 As the intent of some denial-of-service attacks is to induce overload 1138 conditions, an effective overload control mechanism should help to 1139 mitigate the effects of an such an attack. 1141 9.3. Replay Attacks 1143 An attacker that has managed to obtain some messages from the 1144 overload control mechanism may attempt to affect the behavior of 1145 nodes supporting the mechanism by sending those messages at 1146 potentially inopportune times. In addition to time shifting, replay 1147 attacks may send messages to other nodes as well (target shifting). 1149 A design goal for the Diameter overload control solution is to 1150 minimize or eliminate the possibility of causing disruption by using 1151 a replay attack on the Diameter overload control mechanism. 1152 (Allowing a replay attack using the overload control solution would 1153 violate REQ 27.) 1155 9.4. Man-in-the-Middle Attacks 1157 By inserting themselves in between two nodes supporting the Diameter 1158 overload control mechanism, an attacker may potentially both access 1159 and alter the information sent between those nodes. This can be used 1160 for information gathering for business intelligence and attack 1161 targeting, as well as direct attacks. 1163 REQs 27, 28, and 29 imply a need to prevent man-in-the-middle attacks 1164 on the overload control solution. A transport using TLS and/or IPSEC 1165 may be desirable for this purpose. 1167 9.5. Compromised Hosts 1169 A compromised host that supports the Diameter overload control 1170 mechanism could be used for information gathering as well as for 1171 sending malicious information to any Diameter node that would 1172 normally accept information from it. While is is beyond the scope of 1173 the Diameter overload control mechanism to mitigate any operational 1174 interruption to the compromised host, REQs 28 and 29 imply a need to 1175 minimize the impact that a compromised host can have on other nodes 1176 through the use of the Diameter overload control mechanism. Of 1177 course, a compromised host could be used to cause damage in a number 1178 of other ways. This is out of scope for a Diameter overload control 1179 mechanism. 1181 10. References 1183 10.1. Normative References 1185 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1186 Requirement Levels", BCP 14, RFC 2119, March 1997. 1188 [RFC6733] Fajardo, V., Arkko, J., Loughney, J., and G. Zorn, 1189 "Diameter Base Protocol", RFC 6733, October 2012. 1191 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 1192 RFC 2914, September 2000. 1194 [RFC3539] Aboba, B. and J. Wood, "Authentication, Authorization and 1195 Accounting (AAA) Transport Profile", RFC 3539, June 2003. 1197 10.2. Informative References 1199 [RFC5390] Rosenberg, J., "Requirements for Management of Overload in 1200 the Session Initiation Protocol", RFC 5390, December 2008. 1202 [RFC6357] Hilt, V., Noel, E., Shen, C., and A. Abdelal, "Design 1203 Considerations for Session Initiation Protocol (SIP) 1204 Overload Control", RFC 6357, August 2011. 1206 [TR23.843] 1207 3GPP, "Study on Core Network Overload Solutions (Work in 1208 Progress)", TR 23.843 0.6.0, October 2012. 1210 [IR.34] GSMA, "Inter-Service Provider IP Backbone Guidelines", 1211 IR 34 7.0, January 2012. 1213 [IR.88] GSMA, "LTE Roaming Guidelines", IR 88 7.0, January 2012. 1215 [IR.92] GSMA, "IMS Profile for Voice and SMS", IR 92 7.0, 1216 March 2013. 1218 [TS23.002] 1219 3GPP, "Network Architecture", TS 23.002 12.0.0, 1220 September 2012. 1222 [TS29.272] 1223 3GPP, "Evolved Packet System (EPS); Mobility Management 1224 Entity (MME) and Serving GPRS Support Node (SGSN) related 1225 interfaces based on Diameter protocol", TS 29.272 11.4.0, 1226 September 2012. 1228 [TS29.212] 1229 3GPP, "Policy and Charging Control (PCC) over Gx/Sd 1230 reference point", TS 29.212 11.6.0, September 2012. 1232 [TS29.228] 1233 3GPP, "IP Multimedia (IM) Subsystem Cx and Dx interfaces; 1234 Signalling flows and message contents", TS 29.228 11.5.0, 1235 September 2012. 1237 [TS29.002] 1238 3GPP, "Mobile Application Part (MAP) specification", 1239 TS 29.002 11.4.0, September 2012. 1241 Appendix A. Contributors 1243 Significant contributions to this document were made by Adam Roach 1244 and Eric Noel. 1246 Appendix B. Acknowledgements 1248 Review of, and contributions to, this specification by Martin Dolly, 1249 Carolyn Johnson, Jianrong Wang, Imtiaz Shaikh, Jouni Korhonen, Robert 1250 Sparks, Dieter Jacobsohn, Janet Gunn, Jean-Jacques Trottin, Laurent 1251 Thiebaut, Andrew Booth, and Lionel Morand were most appreciated. We 1252 would like to thank them for their time and expertise. 1254 Authors' Addresses 1256 Eric McMurry 1257 Tekelec 1258 17210 Campbell Rd. 1259 Suite 250 1260 Dallas, TX 75252 1261 US 1263 Email: emcmurry@computer.org 1265 Ben Campbell 1266 Tekelec 1267 17210 Campbell Rd. 1268 Suite 250 1269 Dallas, TX 75252 1270 US 1272 Email: ben@nostrum.com