idnits 2.17.1 draft-ietf-tsvwg-sctp-failover-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 date (July 17, 2015) is 3206 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) ** Obsolete normative reference: RFC 4960 (Obsoleted by RFC 9260) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group Y. Nishida 3 Internet-Draft GE Global Research 4 Intended status: Standards Track P. Natarajan 5 Expires: January 18, 2016 Cisco Systems 6 A. Caro 7 BBN Technologies 8 P. Amer 9 University of Delaware 10 K. Nielsen 11 Ericsson 12 July 17, 2015 14 SCTP-PF: Quick Failover Algorithm in SCTP 15 draft-ietf-tsvwg-sctp-failover-11.txt 17 Abstract 19 SCTP supports multi-homing. However, when the failover operation 20 specified in RFC4960 is followed, there can be significant delay and 21 performance degradation in the data transfer path failover. To 22 overcome this problem this document specifies a quick failover 23 algorithm (SCTP-PF) based on the introduction of a Potentially Failed 24 (PF) state in SCTP Path Management. 26 The document also specifies a dormant state operation of SCTP. This 27 dormant state operation is required to be followed by an SCTP-PF 28 implementation, but it may equally well be applied by a standard 29 RFC4960 SCTP implementation. 31 Additionally, the document introduces an alternative switchback mode 32 called Permanent Failover that will be beneficial in some situations. 33 This mode of operation applies to both a standard RFC4960 SCTP 34 implementation as well as to a SCTP-PF implementation. 36 The procedures defined in the document require only minimal 37 modifications to the RFC4960 specification. The procedures are 38 sender-side only and do not impact the SCTP receiver. 40 Status of This Memo 42 This Internet-Draft is submitted in full conformance with the 43 provisions of BCP 78 and BCP 79. 45 Internet-Drafts are working documents of the Internet Engineering 46 Task Force (IETF). Note that other groups may also distribute 47 working documents as Internet-Drafts. The list of current Internet- 48 Drafts is at http://datatracker.ietf.org/drafts/current/. 50 Internet-Drafts are draft documents valid for a maximum of six months 51 and may be updated, replaced, or obsoleted by other documents at any 52 time. It is inappropriate to use Internet-Drafts as reference 53 material or to cite them other than as "work in progress." 55 This Internet-Draft will expire on January 18, 2016. 57 Copyright Notice 59 Copyright (c) 2015 IETF Trust and the persons identified as the 60 document authors. All rights reserved. 62 This document is subject to BCP 78 and the IETF Trust's Legal 63 Provisions Relating to IETF Documents 64 (http://trustee.ietf.org/license-info) in effect on the date of 65 publication of this document. Please review these documents 66 carefully, as they describe your rights and restrictions with respect 67 to this document. Code Components extracted from this document must 68 include Simplified BSD License text as described in Section 4.e of 69 the Trust Legal Provisions and are provided without warranty as 70 described in the Simplified BSD License. 72 Table of Contents 74 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 75 2. Conventions and Terminology . . . . . . . . . . . . . . . . . 4 76 3. SCTP with Potentially-Failed Destination State (SCTP-PF) . . 4 77 3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 4 78 3.2. Specification of the SCTP-PF Procedures . . . . . . . . . 5 79 4. Dormant State Operation . . . . . . . . . . . . . . . . . . . 9 80 4.1. SCTP Dormant State Procedure . . . . . . . . . . . . . . 10 81 5. Permanent Failover . . . . . . . . . . . . . . . . . . . . . 11 82 6. Suggested SCTP Protocol Parameter Values . . . . . . . . . . 12 83 7. Socket API Considerations . . . . . . . . . . . . . . . . . . 12 84 7.1. Support for the Potentially Failed Path State . . . . . . 13 85 7.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket 86 Option . . . . . . . . . . . . . . . . . . . . . . . . . 14 87 7.3. Exposing the Potentially Failed Path State 88 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option . . 15 89 8. Security Considerations . . . . . . . . . . . . . . . . . . . 15 90 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 91 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 16 92 11. Proposed Change of Status (to be Deleted before Publication) 16 93 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 17 94 12.1. Normative References . . . . . . . . . . . . . . . . . . 17 95 12.2. Informative References . . . . . . . . . . . . . . . . . 17 96 Appendix A. Discussions of Alternative Approaches . . . . . . . 18 97 A.1. Reduce Path.Max.Retrans (PMR) . . . . . . . . . . . . . . 18 98 A.2. Adjust RTO related parameters . . . . . . . . . . . . . . 19 99 Appendix B. Discussions for Path Bouncing Effect . . . . . . . . 19 100 Appendix C. SCTP-PF for SCTP Single-homed Operation . . . . . . 20 101 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 20 103 1. Introduction 105 The Stream Control Transmission Protocol (SCTP) specified in 106 [RFC4960] supports multi homing at the transport layer. SCTP's multi 107 homing features include failure detection and failover procedures to 108 provide network interface redundancy and improved end-to-end fault 109 tolerance. In SCTP's current failure detection procedure, the sender 110 must experience Path.Max.Retrans (PMR) number of consecutive failed 111 timer-based retransmissions on a destination address before detecting 112 a path failure. Until detecting the path failure, the sender 113 continues to transmit data on the failed path. The prolonged time in 114 which [RFC4960] SCTP continues to use a failed path severely degrades 115 the performance of the protocol. To address this problem, this 116 document specifies a quick failover algorithm (SCTP-PF) based on the 117 introduction of a new Potentially Failed path state in SCTP path 118 management. The performance deficiencies of the [RFC4960] failover 119 operation, and the improvements obtainable from the introduction of a 120 Potentially Failed state in SCTP, were proposed and documented in 121 [NATARAJAN09] for Concurrent Multipath Transfer SCTP [IYENGAR06]. 123 While SCTP-PF can accelerate failover process and improve 124 performance, the risks that an SCTP endpoint enters in dormant state 125 where all destination addresses are inactive can be increased. 126 [RFC4960] leaves the protocol operation during dormant state to 127 implementations and encourages to avoid entering the state as much as 128 possible by careful tuning of the Path.Max.Retrans (PMR) and 129 Association.Max.Retrans (AMR) parameters. We specify a dormant state 130 operation for SCTP-PF which makes SCTP-PF provide the same disruption 131 tolerance as [RFC4960] despite that the dormant state may be entered 132 more quickly. The dormant state operation may equally well be 133 applied by an [RFC4960] implementation and will here serve to provide 134 added fault tolerance for situations where the tuning of the 135 Path.Max.Retrans (PMR) and Association.Max.Retrans (AMR) parameters 136 fail to provide adequate prevention of the entering of the dormant 137 state. 139 The operation after the recovery of a failed path equally well 140 impacts the performance of the protocol. With the procedures 141 specified in [RFC4960] SCTP will, after a failover from the primary 142 path, switch back to use the primary path for data transfer as soon 143 as this path becomes available again. From a performance perspective 144 such a forced switchback of the data transmission path can be 145 suboptimal as the CWND towards the original primary destination 146 address has to be rebuilt once data transfer resumes, [CARO02]. As 147 an optional alternative to the switchback operation of [RFC4960], 148 this document specifies an alternative Permanent Failover procedure 149 which avoid such forced switchbacks of the data transfer path. The 150 Permanent Failover operation was originally proposed in [CARO02]. 152 While SCTP-PF primarily is motivated by a desire to improve the 153 multi-homed operation, the feature applies also to SCTP single-homed 154 operation. Here the algorithm serves to provide increased failure 155 detection on idle associations, whereas the failover or switchback 156 aspects of the algorithm will not be activated. This is discussed in 157 more detail in Appendix C. 159 A brief description of the motivation for the introduction of the 160 Potentially Failed state including a discussion of alternative 161 approaches to mitigate the deficiencies of the [RFC4960] failover 162 operation are given in the Appendices. Discussion of path bouncing 163 effects that might be caused by frequent switchover, are also 164 provided there. 166 2. Conventions and Terminology 168 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 169 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 170 document are to be interpreted as described in [RFC2119]. 172 3. SCTP with Potentially-Failed Destination State (SCTP-PF) 174 3.1. Overview 176 To minimize the performance impact during failover, the sender should 177 avoid transmitting data to a failed destination address as early as 178 possible. In the [RFC4960] SCTP path management scheme, the sender 179 stops transmitting data to a destination address only after the 180 destination address is marked inactive. This process takes a 181 significant amount of time as it requires the error counter of the 182 destination address to exceed the Path.Max.Retrans (PMR) threshold. 183 The issue cannot simply be mitigated by lowering of the PMR threshold 184 because this may result in spurious failure detection and unnecessary 185 prevention of the usage of a preferred primary path as well as it, 186 due to the coupled tuning of the Path.Max.Retrans (PMR) and the 187 Association.Max.Retrans (AMR) parameter values in [RFC4960], may 188 result in compromisation of the fault tolerance of SCTP. 190 The solution provided in this document is to extend the SCTP path 191 management scheme of [RFC4960] by the addition of the Potentially 192 Failed (PF) state as an intermediate state in between the active and 193 inactive state of a destination address in [RFC4960] path management 194 scheme, and let the failover of data transfer away from a destination 195 address be driven by the entering of the PF state instead of by the 196 entering of the inactive state. Thereby SCTP may perform quick 197 failover without compromising the overall fault tolerance of 198 [RFC4960] SCTP. At the same time, RTO-based HEARTBEAT probing is 199 initiated towards a destination address once it enters PF state. 200 Thereby SCTP may quickly ascertain whether network connectivity 201 towards the destination address is broken or whether the failover was 202 spurious. In the case where the failover was spurious data transfer 203 may quickly resume towards the original destination address. 205 The new failure detection algorithm assumes that loss detected by a 206 timeout implies either severe congestion or network connectivity 207 failure and it assumes that by default a destination address is 208 classified as PF already at the occurrence of one first timeout. 210 3.2. Specification of the SCTP-PF Procedures 212 The SCTP-PF operation is specified as follows: 214 1. The sender maintains a new tunable SCTP Protocol Parameter 215 called PotentiallyFailed.Max.Retrans (PFMR). The PFMR defines 216 the new intermediate PF threshold on the destination address 217 error counter at exceed of which the destination address is 218 classified as PF. The RECOMMENDED value of PFMR is 0, but other 219 values MAY be used. Setting PFMR larger to or equal to 220 Path.Max.Retrans (PMR) does not result in definition of a PF 221 threshold for the destination address. I.e., the destination 222 address will not be classified as PF prior to reaching inactive 223 state. 225 2. The error counter of an active destination address is 226 incremented as specified in [RFC4960]. This means that the 227 error counter of the destination address will be incremented 228 each time the T3-rtx timer expires, or each time a HEARTBEAT 229 chunk is sent when idle and not acknowledged within an RTO. 230 When the value in the destination address error counter exceeds 231 PFMR, the endpoint MUST mark the destination address as in the 232 PF state. 234 3. The PFMR threshold defines the point the destination address no 235 longer is considered a good candidate for data transmission and 236 a SCTP-PF sender SHOULD NOT send data to destination addresses 237 in PF state when alternative destination addresses in active 238 state are available. Specifically this means that: 240 i When there is outbound data to send and the destination 241 address presently used for data transmission is in PF state, 242 the sender SHOULD choose a destination address in active 243 state, if one exists, and failover to deploy this destination 244 address for data transmission. 246 ii When retransmitting data that has timed out and the sender 247 thus by [RFC4960], section 6.4.1, should attempt to pick a 248 new destination address for data retransmission, the sender 249 SHOULD choose an alternate destination transport address in 250 active state if one exists. 252 iii When there is outbound data to send and the SCTP user 253 explicitly requests to send data to a destination address in 254 PF state, the sender SHOULD send the data to an alternate 255 destination address in active state if one exists. 257 When choosing among multiple destination address in active state 258 the following considerations are given: 260 A. An SCTP sender should comply with [RFC4960], section 6.4.1, 261 principles of choosing most divergent source-destination 262 pairs compared with, for i.: the destination address in PF 263 state that it performs a failover from, and for ii.: the 264 destination address towards which the data timed out. Rules 265 for picking the most divergent source-destination pair are 266 an implementation decision and are not specified within this 267 document. 269 B. A SCTP-PF sender MAY choose to send data to a destination 270 address in PF state, even if destination addresses in active 271 state exist, have the SCTP-PF sender other means of 272 information available that disqualifies the destination 273 address in active state from being preferred. However, the 274 discussion of such mechanisms is outside of the scope of the 275 SCTP-PF operation specified in this document. 277 In all cases, the sender MUST NOT change the state of chosen 278 destination address, whether this state be active or PF, and it 279 MUST NOT clear the error counter of the destination address as a 280 result of choosing the destination address for data 281 transmission. 283 4. When the destination addresses are all in PF state or some in PF 284 state and some in inactive state, the sender MUST choose one 285 destination address in PF state and transmit or retransmit data 286 to this destination address using the following rules: 288 A. The sender SHOULD choose the destination in PF state with 289 the lowest error count (fewest consecutive timeouts) for 290 data transmission and transmit or retransmit data to this 291 destination. 293 B. When there are multiple PF destinations with same error 294 count, the sender should let the choice among the multiple 295 PF destination with equal error count be based on the 296 [RFC4960], section 6.4.1, principles of choosing most 297 divergent source-destination pairs when executing 298 (potentially consecutive) retransmission. Rules for picking 299 the most divergent source-destination pair are an 300 implementation decision and are not specified within this 301 document. 303 C. A sender MAY choose to deploy other strategies than the 304 above when choosing among multiple PF destinations have the 305 SCTP-PF sender other means of information available that 306 qualifies a particular destination address for being used. 307 The SCTP-PF protocol operation specified in this document 308 makes no assumption of the existence of such other means of 309 information and specifies for the above as the default 310 operation of an SCTP-PF sender. 312 The sender MUST NOT change the state and the error counter of 313 any destination address regardless of whether it has been chosen 314 for transmission or not. 316 5. The HB.interval of the Path Heartbeat function of [RFC4960] 317 MUST be ignored for destination addresses in PF state. Instead 318 HEARTBEAT chunks are sent to destination addresses in PF state 319 once per RTO. HEARTBEAT chunks SHOULD be sent to destination 320 addresses in PF state, but the sending of HEARTBEATS MUST honor 321 whether the Path Heartbeat function (Section 8.3 of [RFC4960]) 322 is enabled for the destination address or not. I.e., if the 323 Path Heartbeat function is disabled for the destination address 324 in question, HEARTBEATS MUST NOT be sent. Note that when 325 Heartbeat function is disabled, it may take longer to transition 326 PF destination to ACTIVE. 328 6. HEARTBEATs are sent when a destination address reaches the PF 329 state. When a HEARTBEAT chunk is not acknowledged within the 330 RTO, the sender increments the error counter and exponentially 331 backs off the RTO value. If the error counter is less than PMR, 332 the sender transmits another packet containing the HEARTBEAT 333 chunk immediately after timeout expiration on the previous 334 HEARTBEAT. When data is being transmitted to a destination 335 address in the PF state, the transmission of a HEARTBEAT chunk 336 MAY be omitted in case receipt of a SACK of or a T3-rtx timer 337 expiration on the outstanding data can provide equivalent 338 information, such as a case where the data chunk has transmitted 339 to a single destination. Likewise, the timeout of a HEARTBEAT 340 chunk MAY be ignored if data is outstanding towards the 341 destination address. 343 7. When the sender receives a HEARTBEAT ACK from a HEARTBEAT sent 344 to a destination address in PF state, the sender MUST clear the 345 error counter of the destination address and transition the 346 destination address back to active state. When the sender 347 resumes data transmission on the destination address, it MUST do 348 this following the prescriptions of Section 7.2 of [RFC4960]. 350 8. Additional (PMR - PFMR) consecutive timeouts on a destination 351 address in PF state confirm the path failure, upon which the 352 destination address transitions to the inactive state. As 353 described in [RFC4960], the sender (i) SHOULD notify the ULP 354 about this state transition, and (ii) transmit HEARTBEAT chunks 355 to the inactive destination address at a lower HB.interval 356 frequency as described in Section 8.3 of [RFC4960] (when the 357 Path Heartbeat function is enabled for the destination address). 359 9. Acknowledgments for chunks that have been transmitted to 360 multiple destinations (i.e., a chunk which has been 361 retransmitted to a different destination address than the 362 destination address to which the chunk was first transmitted) 363 MUST NOT clear the error count for an inactive destination 364 address and MUST NOT transition a destination address in PF 365 state back to active state, since a sender cannot disambiguate 366 whether the ACK was for the original transmission or the 367 retransmission(s). A SCTP sender MAY apply a different approach 368 for the error count handling based on unequivocally information 369 on which destination (including multiple destination addresses) 370 the chunk reached. This document makes no reference to what 371 such unequivocally information could consist of, neither how 372 such unequivocally information could be obtained. The design of 373 such an alternative approach is left to implementations. 375 10. Acknowledgments for chunks that has been transmitted to one 376 destination address only MUST clear the error counter for the 377 destination address and MUST transition a destination address in 378 PF state back to Active state. This situation can happen when 379 new data is sent to a destination address in the PF state. It 380 can also happen in situations where the destination address is 381 in the PF state due to the occurrence of a spurious T3-rtx timer 382 and Acknowledgments start to arrive for data sent prior to 383 occurrence of the spurious T3-rtx and data has not yet been 384 retransmitted towards other destinations. This document does 385 not specify special handling for detection of or reaction to 386 spurious T3-rtx timeouts, e.g., for special operation vis-a-vis 387 the congestion control handling or data retransmission operation 388 towards a destination address which undergoes a transition from 389 active to PF to active state due to a spurious T3-rtx timeout. 390 But it is noted that this is an area which would benefit from 391 additional attention, experimentation and specification for 392 Single Homed SCTP as well as for Multi Homed SCTP protocol 393 operation. 395 11. When all destination addresses are in inactive state, and SCTP 396 protocol operation thus is said to be in dormant state, the 397 prescriptions given in Section 4 shall be followed. 399 12. The SCTP stack should provide the ULP with the means to expose 400 the PF state of its destinations as well as the means to notify 401 of state transitions from Active to PF, and vice-versa. However 402 it is recommended that an SCTP stack implementing SCTP-PF also 403 allows for that the ULP is kept ignorant of the PF state of its 404 destinations and the associated state transition. For this 405 reason is it recommended that an SCTP stack implementing SCTP-PF 406 also should provide the ULP with the means to suppress exposure 407 of PF state and the associated state transitions. 409 4. Dormant State Operation 411 In a situation with complete disruption of the communication in 412 between the SCTP Endpoints, the aggressive HEARTBEAT transmissions of 413 SCTP-PF on destination addresses in PF state may make the association 414 enter dormant state faster than a standard [RFC4960] SCTP 415 implementation given the same setting of Path.Max.Retrans (PMR) and 416 Association.Max.Retrans (AMR). For example, an SCTP association with 417 two destination addresses typically would reach dormant state in half 418 the time of an [RFC4960] SCTP implementation in such situations. 419 This is because a SCTP PF sender will send HEARTBEATS and data 420 retransmissions in parallel with RTO intervals when there are 421 multiple destinations addresses in PF state. This argument presumes 422 that RTO << HB.interval of [RFC4960]. With the design goal that 423 SCTP-PF shall provide the same level of disruption tolerance as an 424 [RFC4960] SCTP implementation with the same Path.Max.Retrans (PMR) 425 and Association.Max.Retrans (AMR) setting, we prescribe for that an 426 SCTP-PF implementation SHOULD operate as described below in 427 Section 4.1 during dormant state. 429 An SCTP-PF implementation MAY choose a different dormant state 430 operation than the one described below in Section 4.1 provided that 431 the solution chosen does not compromise the fault tolerance of the 432 SCTP-PF operation. 434 The below prescription for SCTP-PF dormant state handling SHOULD NOT 435 be coupled to the value of the PFMR, but solely to the activation of 436 SCTP-PF logic in an SCTP implementation. 438 It is noted that the below dormant state operation is considered to 439 provide added disruption tolerance also for an [RFC4960] SCTP 440 implementation, and that it can be sensible for an [RFC4960] SCTP 441 implementation to follow this mode of operation. For an [RFC4960] 442 SCTP implementation the continuation of data transmission during 443 dormant state makes the fault tolerance of SCTP be more robust 444 towards situations where some, or all, alternative paths of an SCTP 445 association approach, or reach, inactive state prior to that the 446 primary path used for data transmission observes trouble. 448 4.1. SCTP Dormant State Procedure 450 a. When the destination addresses are all in inactive state and data 451 is available for transfer, the sender MUST choose one destination 452 and transmit data to this destination address. 454 b. The sender MUST NOT change the state of the chosen destination 455 address (it remains in inactive state) and it MUST NOT clear the 456 error counter of the destination address as a result of choosing 457 the destination address for data transmission. 459 c. The sender SHOULD choose the destination in inactive state with 460 the lowest error count (fewest consecutive timeouts) for data 461 transmission. When there are multiple destinations with same 462 error count in inactive state, the sender SHOULD attempt to pick 463 the most divergent source - destination pair from the last source 464 - destination pair where failure was observed. Rules for picking 465 the most divergent source-destination pair are an implementation 466 decision and are not specified within this document. To support 467 differentiation of inactive destination addresses based on their 468 error count SCTP will need to allow for increment of the 469 destination address error counters up to some reasonable limit 470 above PMR+1, thus changing the prescriptions of [RFC4960], 471 section 8.3, in this respect. The exact limit to apply is not 472 specified in this document but it is considered reasonable to 473 require for such to be an order of magnitude higher than the PMR 474 value. A sender MAY choose to deploy other strategies that the 475 strategy defined by here. The strategy to prioritize the last 476 active destination address, i.e., the destination address with 477 the fewest error counts is optimal when some paths are 478 permanently inactive, but suboptimal when a path instability is 479 transient. 481 5. Permanent Failover 483 The objective of the Permanent Failover operation is to allow the 484 SCTP sender to continue data transmission on a new working path even 485 when the old primary destination address becomes active again. This 486 is achieved by having SCTP perform a switch over of the primary path 487 to the new working path if the error counter of the primary path 488 exceeds a certain threshold. This mode of operation can be applied 489 not only to SCTP-PF implementations, but also to [RFC4960] 490 implementations. 492 The Permanent Failover operation requires only sender side changes. 493 The details are: 495 1. The sender maintains a new tunable parameter, called 496 Primary.Switchover.Max.Retrans (PSMR). For SCTP-PF 497 implementations, the PSMR MUST be set greater or equal to the 498 PFMR value. For [RFC4960] implementations the PSMR MUST be set 499 greater or equal to the PMR value. Implementations MUST reject 500 any other values of PSMR. 502 2. When the path error counter on a set primary path exceeds PSMR, 503 the SCTP implementation MUST autonomously select and set a new 504 primary path. 506 3. The primary path selected by the SCTP implementation MUST be the 507 path which at the given time would be chosen for data transfer. 508 A previously failed primary path can be used as data transfer 509 path as per normal path selection when the present data transfer 510 path fails. 512 4. For SCTP-PF, the recommended value of PSMR is PFMR when Permanent 513 Failover is used. This means that no forced switchback to a 514 previously failed primary path is performed. An SCTP-PF 515 implementation of Permanent Failover MUST support the setting of 516 PSMR = PFMR. A SCTP-PF implementation of Permanent Failover MAY 517 support setting of PSMR > PFMR. 519 5. For [RFC4960] SCTP, the recommended value of PSMR is PMR when 520 Permanent Failover is used. This means that no forced switchback 521 to a previously failed primary path is performed. A [RFC4960] 522 SCTP implementation of Permanent Failover MUST support the 523 setting of PSMR = PMR An [RFC4960] SCTP implementation of 524 Permanent Failover MAY support larger settings of PSMR > PMR. 526 6. It MUST be possible to disable the Permanent Failover and obtain 527 the standard switchback operation of [RFC4960]. 529 The manner of switch over operation that is most optimal in a given 530 scenario depends on the relative quality of a set primary path versus 531 the quality of alternative paths available as well as it depends on 532 the extent to which it is desired for the mode of operation to 533 enforce traffic distribution over a number of network paths. I.e., 534 load distribution of traffic from multiple SCTP associations may be 535 sought to be enforced by distribution of the set primary paths with 536 [RFC4960] switchback operation. However as [RFC4960] switchback 537 behavior is suboptimal in certain situations, especially in scenarios 538 where a number of equally good paths are available, an SCTP 539 implementation MAY support also, as alternative behavior, the 540 Permanent Failover mode of operation and MAY enable it based on 541 users' requests. 543 For an SCTP implementation that implements Permanent Failover, this 544 specification RECOMMENDS that the standard RFC4960 switchback 545 operation is retained as the default operation. 547 6. Suggested SCTP Protocol Parameter Values 549 This document does not alter the [RFC4960] value RECOMMENDATIONS for 550 the SCTP Protocol Parameters defined in [RFC4960]. 552 The following protocol parameter is RECOMMENDED: 554 PotentiallyFailed.Max.Retrans (PFMR) - 0 556 7. Socket API Considerations 558 This section describes how the socket API defined in [RFC6458] is 559 extended to provide a way for the application to control and observe 560 the SCTP-PF behavior as well as the Permanent Failover function. 562 Please note that this section is informational only. 564 A socket API implementation based on [RFC6458] is, by means of the 565 existing SCTP_PEER_ADDR_CHANGE event, extended to provide the event 566 notification when a peer address enters or leaves the potentially 567 failed state as well as the socket API implementation is extended to 568 expose the potentially failed state of a peer address in the existing 569 SCTP_GET_PEER_ADDR_INFO structure. 571 Furthermore, two new read/write socket options for the level 572 IPPROTO_SCTP and the name SCTP_PEER_ADDR_THLDS and 573 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE are defined as described below. 574 The first socket option is used to control the values of the PFMR and 575 PSMR parameters described in Section 3 and in Section 5. The second 576 one controls the exposition of the potentially failed path state. 578 Support for the SCTP_PEER_ADDR_THLDS and 579 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE socket options need also to be 580 added to the function sctp_opt_info(). 582 7.1. Support for the Potentially Failed Path State 584 As defined in [RFC6458], the SCTP_PEER_ADDR_CHANGE event is provided 585 if the status of a peer address changes. In addition to the state 586 changes described in [RFC6458], this event is also provided, if a 587 peer address enters or leaves the potentially failed state. The 588 notification as defined in [RFC6458] uses the following structure: 590 struct sctp_paddr_change { 591 uint16_t spc_type; 592 uint16_t spc_flags; 593 uint32_t spc_length; 594 struct sockaddr_storage spc_aaddr; 595 uint32_t spc_state; 596 uint32_t spc_error; 597 sctp_assoc_t spc_assoc_id; 598 } 600 [RFC6458] defines the constants SCTP_ADDR_AVAILABLE, 601 SCTP_ADDR_UNREACHABLE, SCTP_ADDR_REMOVED, SCTP_ADDR_ADDED, and 602 SCTP_ADDR_MADE_PRIM to be provided in the spc_state field. This 603 document defines in addition to that the new constant 604 SCTP_ADDR_POTENTIALLY_FAILED, which is reported if the affected 605 address becomes potentially failed. 607 The SCTP_GET_PEER_ADDR_INFO socket option defined in [RFC6458] can be 608 used to query the state of a peer address. It uses the following 609 structure: 611 struct sctp_paddrinfo { 612 sctp_assoc_t spinfo_assoc_id; 613 struct sockaddr_storage spinfo_address; 614 int32_t spinfo_state; 615 uint32_t spinfo_cwnd; 616 uint32_t spinfo_srtt; 617 uint32_t spinfo_rto; 618 uint32_t spinfo_mtu; 619 }; 621 [RFC6458] defines the constants SCTP_UNCONFIRMED, SCTP_ACTIVE, and 622 SCTP_INACTIVE to be provided in the spinfo_state field. This 623 document defines in addition to that the new constant 624 SCTP_POTENTIALLY_FAILED, which is reported if the peer address is 625 potentially failed. 627 7.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket Option 629 Applications can control the SCTP-PF behavior by getting or setting 630 the number of consecutive timeouts before a peer address is 631 considered potentially failed or unreachable. The same socket option 632 is used by applications to set and get the number of timeouts before 633 the primary path is changed automatically by the Permanent Failover 634 function. This socket option uses the level IPPROTO_SCTP and the 635 name SCTP_PEER_ADDR_THLDS. 637 The following structure is used to access and modify the thresholds: 639 struct sctp_paddrthlds { 640 sctp_assoc_t spt_assoc_id; 641 struct sockaddr_storage spt_address; 642 uint16_t spt_pathmaxrxt; 643 uint16_t spt_pathpfthld; 644 uint16_t spt_pathcpthld; 645 }; 647 spt_assoc_id: This parameter is ignored for one-to-one style 648 sockets. For one-to-many style sockets the application may fill 649 in an association identifier or SCTP_FUTURE_ASSOC. It is an error 650 to use SCTP_{CURRENT|ALL}_ASSOC in spt_assoc_id. 652 spt_address: This specifies which peer address is of interest. If a 653 wild card address is provided, this socket option applies to all 654 current and future peer addresses. 656 spt_pathmaxrxt: Each peer address of interest is considered 657 unreachable, if its path error counter exceeds spt_pathmaxrxt. 659 spt_pathpfthld: Each peer address of interest is considered 660 Potentially Failed, if its path error counter exceeds 661 spt_pathpfthld. 663 spt_pathcpthld: Each peer address of interest is not considered the 664 primary remote address anymore, if its path error counter exceeds 665 spt_pathcpthld. Using a value of 0xffff disables the selection of 666 a new primary peer address. If an implementation does not support 667 the automatically selection of a new primary address, it should 668 indicate an error with errno set to EINVAL if a value different 669 from 0xffff is used in spt_pathcpthld. For SCTP-PF, the setting 670 of spt_pathcpthld < spt_pathpfthld should be rejected with errno 671 set to EINVAL. For [RFC4960] SCTP, the setting of spt_pathcpthld 672 < spt_pathmaxrxt should be rejected with errno set to EINVAL. A 673 SCTP-PF implementation MAY support only setting of spt_pathcpthld 674 = spt_pathpfthld and spt_pathcpthld = 0xffff and a [RFC4960] SCTP 675 implementation MAY support only setting of spt_pathcpthld = 676 spt_pathmaxrxt and spt_pathcpthld = 0xffff. In these cases SCTP 677 shall reject setting of other values with errno set to EINVAL. 679 7.3. Exposing the Potentially Failed Path State 680 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option 682 Applications can control the exposure of the potentially failed path 683 state in the SCTP_PEER_ADDR_CHANGE event and the 684 SCTP_GET_PEER_ADDR_INFO as described in Section 7.1. The default 685 value is implementation specific. 687 This socket option uses the level IPPROTO_SCTP and the name 688 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE. 690 The following structure is used to control the exposition of the 691 potentially failed path state: 693 struct sctp_assoc_value { 694 sctp_assoc_t assoc_id; 695 uint32_t assoc_value; 696 }; 698 assoc_id: This parameter is ignored for one-to-one style sockets. 699 For one-to-many style sockets the application may fill in an 700 association identifier or SCTP_FUTURE_ASSOC. It is an error to 701 use SCTP_{CURRENT|ALL}_ASSOC in assoc_id. 703 assoc_value: The potentially failed path state is exposed if and 704 only if this parameter is non-zero. 706 8. Security Considerations 708 Security considerations for the use of SCTP and its APIs are 709 discussed in [RFC4960] and [RFC6458]. 711 The logic introduced by this document does not impact existing on- 712 the-wire SCTP messages. Also, this document does not introduce any 713 new on-the-wire SCTP messages that require new security 714 considerations. 716 SCTP-PF makes SCTP not only more robust during primary path failure/ 717 congestion but also more vulnerable to network connectivity/ 718 congestion attacks on the primary path. SCTP-PF makes it easier for 719 an attacker to trick SCTP to change data transfer path, since the 720 duration of time that an attacker needs to compromise the network 721 connectivity is much shorter than [RFC4960]. However, SCTP-PF does 722 not constitute a significant change in the duration of time and 723 effort an attacker needs to keep SCTP away from the primary path. 724 With the standard switchback operation [RFC4960] SCTP resumes data 725 transfer on its primary path as soon as the next HEARTBEAT succeeds. 727 On the other hand, usage of the Permanent Failover mechanism, does 728 change the treat analysis. This is because attackers can force a 729 permanent change of the data transfer path by blocking the primary 730 path until the switchover of the primary path is triggered by the 731 Permanent Failover algorithm. This especially will be the case when 732 Permanent Failover is used together with SCTP-PF with the particular 733 setting of PSMR = PFMR = 0, as Permanent Failover here happens 734 already at the first RTO timeout experienced. Users of the Permanent 735 Failover mechanism should be aware of this fact. 737 The event notification of path state transfer from active to 738 potentially failed state and vice versa gives attackers an increased 739 possibility to generate more local events. However, it is assumed 740 that event notifications are rate-limited in the implementation to 741 address this threat. 743 9. IANA Considerations 745 This document does not create any new registries or modify the rules 746 for any existing registries managed by IANA. 748 10. Acknowledgements 750 The authors wish to thank Michael Tuexen for his many invaluable 751 comments and for his very substantial support with the making of this 752 document. 754 11. Proposed Change of Status (to be Deleted before Publication) 756 Initially this work looked to entail some changes of the Congestion 757 Control (CC) operation of SCTP and for this reason the work was 758 proposed as Experimental. These intended changes of the CC operation 759 have since been judged to be irrelevant and are no longer part of the 760 specification. As the specification entails no other potential 761 harmful features, consensus exists in the WG to bring the work 762 forward as PS. 764 Initially concerns have been expressed about the possibility for the 765 mechanism to introduce path bouncing with potential harmful network 766 impacts. These concerns are believed to be unfounded. This issue is 767 addressed in Appendix B. 769 It is noted that the feature specified by this document is 770 implemented by multiple SCTP SW implementations and furthermore that 771 various variants of the solution have been deployed in Telco 772 signaling environments for several years with good results. 774 12. References 776 12.1. Normative References 778 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 779 Requirement Levels", BCP 14, RFC 2119, March 1997. 781 [RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC 782 4960, September 2007. 784 12.2. Informative References 786 [CARO02] Caro Jr., A., Iyengar, J., Amer, P., Heinz, G., and R. 787 Stewart, "A Two-level Threshold Recovery Mechanism for 788 SCTP", Tech report, CIS Dept, University of Delaware , 7 789 2002. 791 [CARO04] Caro Jr., A., Amer, P., and R. Stewart, "End-to-End 792 Failover Thresholds for Transport Layer Multihoming", 793 MILCOM 2004 , 11 2004. 795 [CARO05] Caro Jr., A., "End-to-End Fault Tolerance using Transport 796 Layer Multihoming", Ph.D Thesis, University of Delaware , 797 1 2005. 799 [FALLON08] 800 Fallon, S., Jacob, P., Qiao, Y., Murphy, L., Fallon, E., 801 and A. Hanley, "SCTP Switchover Performance Issues in WLAN 802 Environments", IEEE CCNC 2008, 1 2008. 804 [GRINNEMO04] 805 Grinnemo, K-J. and A. Brunstrom, "Performance of SCTP- 806 controlled failovers in M3UA-based SIGTRAN networks", 807 Advanced Simulation Technologies Conference , 4 2004. 809 [IYENGAR06] 810 Iyengar, J., Amer, P., and R. Stewart, "Concurrent 811 Multipath Transfer using SCTP Multihoming over Independent 812 End-to-end Paths.", IEEE/ACM Trans on Networking 14(5), 10 813 2006. 815 [JUNGMAIER02] 816 Jungmaier, A., Rathgeb, E., and M. Tuexen, "On the use of 817 SCTP in failover scenarios", World Multiconference on 818 Systemics, Cybernetics and Informatics , 7 2002. 820 [NATARAJAN09] 821 Natarajan, P., Ekiz, N., Amer, P., and R. Stewart, 822 "Concurrent Multipath Transfer during Path Failure", 823 Computer Communications , 5 2009. 825 [RFC6458] Stewart, R., Tuexen, M., Poon, K., Lei, P., and V. 826 Yasevich, "Sockets API Extensions for the Stream Control 827 Transmission Protocol (SCTP)", RFC 6458, December 2011. 829 Appendix A. Discussions of Alternative Approaches 831 This section lists alternative approaches for the issues described in 832 this document. Although these approaches do not require to update 833 RFC4960, we do not recommend them from the reasons described below. 835 A.1. Reduce Path.Max.Retrans (PMR) 837 Smaller values for Path.Max.Retrans shorten the failover duration and 838 in fact this is recommended in some research results [JUNGMAIER02] 839 [GRINNEMO04] [FALLON08]. However to significantly reduce the 840 failover time it is required to go down (as with PFMR) to 841 Path.Max.Retrans=0 and with this setting SCTP switches to another 842 destination address already on a single timeout which may result in 843 spurious failover. Spurious failover is a problem in [RFC4960] SCTP 844 as the transmission of HEARTBEATS on the left primary path, unlike in 845 SCTP-PF, is governed by 'HB.interval' also during the failover 846 process. 'HB.interval' is usually set in the order of seconds 847 (recommended value is 30 seconds) and when the primary path becomes 848 inactive, the next HEARTBEAT may be transmitted only many seconds 849 later. Indeed as recommended, only 30 secs later. Meanwhile, the 850 primary path may since long have recovered, if it needed recovery at 851 all (indeed the failover could be truely spurious). In such 852 situations, post failover, an endpoint is forced to wait in the order 853 of many seconds before the endpoint can resume transmission on the 854 primary path and furthermore once it returns on the primary path the 855 CWND needs to be rebuild anew - a process which the throughput 856 already have had to suffer from on the alternate path. Using a 857 smaller value for 'HB.interval' might help this situation, but it 858 would result in a general waste of bandwidth as such more frequent 859 HEARBEATING would take place also when there are no observed 860 troubles. The bandwidth overhead may be diminished by having the ULP 861 use a smaller 'HB.interval' only on the path which at any given time 862 is set to be the primary path, but this adds complication in the ULP. 864 In addition, smaller Path.Max.Retrans values also affect the 865 'Association.Max.Retrans' value. When the SCTP association's error 866 count exceeds Association.Max.Retrans threshold, the SCTP sender 867 considers the peer endpoint unreachable and terminates the 868 association. Section 8.2 in [RFC4960] recommends that 869 Association.Max.Retrans value should not be larger than the summation 870 of the Path.Max.Retrans of each of the destination addresses. Else 871 the SCTP sender considers its peer reachable even when all 872 destinations are INACTIVE and to avoid this dormant state operation, 873 [RFC4960] SCTP implementation SHOULD reduce Association.Max.Retrans 874 accordingly whenever it reduces Path.Max.Retrans. However, smaller 875 Association.Max.Retrans value compromizes the fault tolerance of SCTP 876 as it increases the chances of association termination during minor 877 congestion events. 879 A.2. Adjust RTO related parameters 881 As several research results indicate, we can also shorten the 882 duration of failover process by adjusting RTO related parameters 883 [JUNGMAIER02] [FALLON08]. During failover process, RTO keeps being 884 doubled. However, if we can choose smaller value for RTO.max, we can 885 stop the exponential growth of RTO at some point. Also, choosing 886 smaller values for RTO.initial or RTO.min can contribute to keep the 887 RTO value small. 889 Similar to reducing Path.Max.Retrans, the advantage of this approach 890 is that it requires no modification to the current specification, 891 although it needs to ignore several recommendations described in the 892 Section 15 of [RFC4960]. However, this approach requires to have 893 enough knowledge about the network characteristics between end 894 points. Otherwise, it can introduce adverse side-effects such as 895 spurious timeouts. 897 The significant issue with this approach, however, is that even if 898 the RTO.max is lowered to an optimal low value, then as long as the 899 Path.Max.Retrans is kept at the [RFC4960] recommended value, the 900 reduction of the RTO.max doesn't reduce the failover time 901 sufficiently enough to prevent severe performance degradation during 902 failover. 904 Appendix B. Discussions for Path Bouncing Effect 906 The methods described in the document can accelerate the failover 907 process. Hence, they might introduce the path bouncing effect where 908 the sender keeps changing the data transmission path frequently. 909 This sounds harmful to the data transfer, however several research 910 results indicate that there is no serious problem with SCTP in terms 911 of path bouncing effect [CARO04] [CARO05]. 913 There are two main reasons for this. First, SCTP is basically 914 designed for multipath communication, which means SCTP maintains all 915 path related parameters (CWND, ssthresh, RTT, error count, etc) per 916 each destination address. These parameters cannot be affected by 917 path bouncing. In addition, when SCTP migrates the data transfer to 918 another path, it starts with the minimal or the initial CWND. Hence, 919 there is little chance for packet reordering or duplicating. 921 Second, even if all communication paths between the end-nodes share 922 the same bottleneck, the SCTP-PF results in a behavior already 923 allowed by [RFC4960]. 925 Appendix C. SCTP-PF for SCTP Single-homed Operation 927 For a single-homed SCTP association the only tangible effect of the 928 activation of SCTP-PF operation is enhanced failure detection in 929 terms of potential notification of the PF state of the sole 930 destination address as well as, for idle associations, more rapid 931 entering, and notification, of inactive state of the destination 932 address and more rapid end-point failure detection. It is believed 933 that neither of these effects are harmful, provided adequate dormant 934 state operation is implemented, and furthermore that they may be 935 particularly useful for applications that deploys multiple SCTP 936 associations for load balancing purposes. The early notification of 937 the PF state may be used for preventive measures as the entering of 938 the PF state can be used as a warning of potential congestion. 939 Depending on the PMR value, the aggressive HEARTBEAT transmission in 940 PF state may speed up the end-point failure detection (exceed of AMR 941 threshold on the sole path error counter) on idle associations in 942 case where relatively large HB.interval value compared to RTO (e.g. 943 30secs) is used. 945 Authors' Addresses 947 Yoshifumi Nishida 948 GE Global Research 949 2623 Camino Ramon 950 San Ramon, CA 94583 951 USA 953 Email: nishida@wide.ad.jp 955 Preethi Natarajan 956 Cisco Systems 957 510 McCarthy Blvd 958 Milpitas, CA 95035 959 USA 961 Email: prenatar@cisco.com 962 Armando Caro 963 BBN Technologies 964 10 Moulton St. 965 Cambridge, MA 02138 966 USA 968 Email: acaro@bbn.com 970 Paul D. Amer 971 University of Delaware 972 Computer Science Department - 434 Smith Hall 973 Newark, DE 19716-2586 974 USA 976 Email: amer@udel.edu 978 Karen E. E. Nielsen 979 Ericsson 980 Kistavaegen 25 981 Stockholm 164 80 982 Sweden 984 Email: karen.nielsen@tieto.com