idnits 2.17.1 draft-ietf-tsvwg-sctp-failover-10.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: ---------------------------------------------------------------------------- == It seems as if not all pages are separated by form feeds - found 0 form feeds but 21 pages 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 (March 9, 2015) is 3307 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 (~~), 2 warnings (==), 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: September 10, 2015 Cisco Systems 6 A. Caro 7 BBN Technologies 8 P. Amer 9 University of Delaware 10 K. Nielsen 11 Ericsson 12 March 9, 2015 14 SCTP-PF: Quick Failover Algorithm in SCTP 15 draft-ietf-tsvwg-sctp-failover-10.txt 17 Abstract 19 One of the major advantages of SCTP is the support of multi-homed 20 communication. A multi-homed SCTP end-point has the ability to 21 withstand network failures by migrating the traffic from an inactive 22 network to an active one. However, if the failover operation as 23 specified in RFC4960 is followed, there can be a significant delay in 24 the migration to the active destination addresses, thus severely 25 reducing the effectiveness of the SCTP failover operation. 27 This document complements RFC4960 by the introduction of a new path 28 state, the Potentially Failed (PF) path state, and an associated new 29 failover operation to apply during a network failure. The algorithm 30 defined is called SCTP Potentially Failed Algorithm, SCTP-PF for 31 short. In addition, the document complements RFC4960 by introducing 32 alternative switchover operation modes for the data transfer path 33 management after the recovery of a failed primary path. These modes 34 can allow improvements in the performance of the operation in some 35 network environments. The implementation of the additional 36 switchover operation modes is an optional part of SCTP-PF. 38 The procedures defined in the document require only minimal 39 modifications to the current specification. The procedures are 40 sender-side only and do not impact the SCTP receiver. 42 Status of This Memo 44 This Internet-Draft is submitted in full conformance with the 45 provisions of BCP 78 and BCP 79. 47 Internet-Drafts are working documents of the Internet Engineering 48 Task Force (IETF). Note that other groups may also distribute 49 working documents as Internet-Drafts. The list of current Internet- 50 Drafts is at http://datatracker.ietf.org/drafts/current/. 52 Internet-Drafts are draft documents valid for a maximum of six months 53 and may be updated, replaced, or obsoleted by other documents at any 54 time. It is inappropriate to use Internet-Drafts as reference 55 material or to cite them other than as "work in progress." 57 This Internet-Draft will expire on September 10, 2015. 59 Copyright Notice 61 Copyright (c) 2015 IETF Trust and the persons identified as the 62 document authors. All rights reserved. 64 This document is subject to BCP 78 and the IETF Trust's Legal 65 Provisions Relating to IETF Documents 66 (http://trustee.ietf.org/license-info) in effect on the date of 67 publication of this document. Please review these documents 68 carefully, as they describe your rights and restrictions with respect 69 to this document. Code Components extracted from this document must 70 include Simplified BSD License text as described in Section 4.e of 71 the Trust Legal Provisions and are provided without warranty as 72 described in the Simplified BSD License. 74 Table of Contents 76 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 77 2. Conventions and Terminology . . . . . . . . . . . . . . . . . 4 78 3. Issues with the SCTP Path Management . . . . . . . . . . . . 4 79 4. SCTP with Potentially-Failed Destination State (SCTP-PF) . . 5 80 4.1. SCTP-PF Concept . . . . . . . . . . . . . . . . . . . . . 5 81 4.2. Specification of the SCTP-PF Algorithm . . . . . . . . . 6 82 4.2.1. Dormant State Operation . . . . . . . . . . . . . . . 10 83 4.3. Permanent Failover . . . . . . . . . . . . . . . . . . . 12 84 4.3.1. Background . . . . . . . . . . . . . . . . . . . . . 12 85 4.3.2. Permanent Failover Algorithm . . . . . . . . . . . . 12 86 5. Socket API Considerations . . . . . . . . . . . . . . . . . . 13 87 5.1. Support for the Potentially Failed Path State . . . . . . 14 88 5.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket 89 Option . . . . . . . . . . . . . . . . . . . . . . . . . 15 90 5.3. Exposing the Potentially Failed Path State 91 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option . . 16 92 6. Security Considerations . . . . . . . . . . . . . . . . . . . 16 93 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 17 94 8. Proposed Change of Status (to be Deleted before Publication) 17 95 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 17 96 9.1. Normative References . . . . . . . . . . . . . . . . . . 17 97 9.2. Informative References . . . . . . . . . . . . . . . . . 17 98 Appendix A. Discussions of Alternative Approaches . . . . . . . 18 99 A.1. Reduce Path.Max.Retrans (PMR) . . . . . . . . . . . . . . 18 100 A.2. Adjust RTO related parameters . . . . . . . . . . . . . . 19 101 Appendix B. Discussions for Path Bouncing Effect . . . . . . . . 20 102 Appendix C. SCTP-PF for SCTP Single-homed Operation . . . . . . 20 103 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 20 105 1. Introduction 107 The Stream Control Transmission Protocol (SCTP) as specified in 108 [RFC4960] supports multihoming at the transport layer -- an SCTP 109 endpoint can bind to multiple IP addresses. SCTP's multihoming 110 features include failure detection and failover procedures to provide 111 network interface redundancy and improved end-to-end fault tolerance. 113 In SCTP's current failure detection procedure, the sender must 114 experience Path.Max.Retrans (PMR) number of consecutive failed timer- 115 based retransmissions on a destination address before detecting a 116 path failure. The sender fails over to an alternate active 117 destination address only after failure detection. Until detecting 118 the failover, the sender continues to transmit data on the failed 119 path, which degrades the SCTP performance. Concurrent Multipath 120 Transfer (CMT) [IYENGAR06] is an proposed extension to SCTP that 121 allows the sender to transmit data on multiple paths simultaneously. 122 Research [NATARAJAN09] shows that the current failure detection 123 procedure worsens CMT performance during failover and can be 124 significantly improved by employing a better failover algorithm. 126 This document specifies an alternative failure detection and failover 127 procedure, the SCTP Potentially Failed algorithm, that improves the 128 performance of SCTP multi-homed operation during a failover. 130 For multi-homed SCTP the operation after the recovery of a failed 131 path equally well impacts the performance of the protocol. With the 132 procedures specified in [RFC4960], SCTP will, after a failover from 133 the primary path, switch back to the primary path for data transfer 134 as soon as this path becomes available again. From a performance 135 perspective, as confirmed in research [CARO02], such a switchback of 136 the data transmission path is not optimal in general. As an optional 137 alternative to the switchback operation of [RFC4960], this document 138 specifies the Permanent Failover procedures proposed by [CARO02]. 140 Additional discussion for alternative approaches that do not require 141 modifications to [RFC4960], as well as discussion of path bouncing 142 effects that might be caused by frequent switchover, are provided in 143 the Appendices. 145 While the Potentially Failed algorithm primarily is motivated for 146 improvement of the SCTP multi-homed operation, the feature applies 147 also to SCTP single-homed operation. Here the algorithm serves to 148 provide increased failure detection on idle associations, whereas the 149 failover or switchback aspects of the algorithm will not be 150 activated. This is discussed in more detail in Appendix C. 152 2. Conventions and Terminology 154 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 155 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 156 document are to be interpreted as described in [RFC2119]. 158 3. Issues with the SCTP Path Management 160 This section describes issues in the SCTP as specified in [RFC4960] 161 to be fixed by the approach described in this document. 163 An SCTP endpoint can support multiple IP addresses. Each SCTP 164 endpoint exchanges the list of its usable addresses during the 165 initial negotiation with its peer. Then the endpoints select one 166 address from the peer's list and use this as the primary destination 167 address. During normal transmission, an SCTP endpoint sends all user 168 data to the primary destination address. Also, it sends packets 169 containing a HEARTBEAT chunk to all idle destination addresses at a 170 certain interval to check the reachability of these destination 171 addresses. Idle destination addresses normally include all non- 172 primary destination addresses. 174 If a sender has multiple active destination addresses, it can 175 retransmit data to an non-primary destination address, if the 176 transmission to the primary times out. 178 When a sender receives an acknowledgment for DATA or HEARTBEAT chunks 179 sent to one of the destination addresses, it considers that 180 destination address to be active and clears the error counter for the 181 destination address. If it fails to receive acknowledgments, the 182 error count for the destination address is increased. If the error 183 counter exceeds the tunable protocol parameter Path.Max.Retrans 184 (PMR), the SCTP endpoint considers the destination address to be 185 inactive. 187 The failover process of SCTP is initiated when the primary path 188 becomes inactive (the error counter for the primary path exceeds 189 Path.Max.Retrans). If the primary path is marked inactive, SCTP 190 chooses a new destination address from one of the active destinations 191 and starts using this as the destination address for sending data. 192 If the primary path becomes active again, SCTP reverts to using the 193 primary destination address for subsequent data transmissions and 194 stop using the non-primary one. 196 One issue with this failover process defined in [RFC4960] is that it 197 usually takes a significant amount of time before SCTP switches to 198 the new destination address. Let's say the primary path on a multi- 199 homed host becomes unavailable and the RTO value for the primary path 200 at that time is around 1 second, it usually takes over 60 seconds 201 before SCTP starts to use the non-primary path for initial data 202 transmission. This is because the recommended value for 203 Path.Max.Retrans in the [RFC4960] is 5, which requires 6 consecutive 204 timeouts before the failover takes place. Before SCTP switches to 205 the non-primary address, SCTP keeps trying to send packets to the 206 primary address and only retransmitted packets are sent to the non- 207 primary address and thus can be received by the receiver. This slow 208 failover process can cause significant performance degradation and is 209 not acceptable in some situations. 211 Another issue with RFC4960 failover and switchback operation is that 212 once the primary path becomes active again, the traffic is 213 unconditionally switched back to use this path. This is not optimal 214 in some situations. This is further discussed in Section 4.3. 216 4. SCTP with Potentially-Failed Destination State (SCTP-PF) 218 To address the issues described in Section 3, this document extends 219 SCTP path management scheme by adding the Potentially Failed state 220 and associated protocol operation. The algorithm is called SCTP 221 Potentially Failed algorithm. SCTP-PF for short. The resulting SCTP 222 path management operation is called SCTP Potentially Failed 223 operation. 225 4.1. SCTP-PF Concept 227 The introduction of the Potentially Failed state stems from the 228 following two observations about SCTP's failure detection procedure: 230 o To minimize the performance impact during failover, the sender 231 should avoid transmitting data to the failed destination address 232 as early as possible. In the current SCTP path management scheme, 233 the sender stops transmitting data to a destination address only 234 after the destination address is marked Failed (inactive). Thus, 235 a smaller PMR value is better because the sender can transition a 236 destination address to the Failed (inactive) state quicker. 238 o Smaller PMR values increase the chances of spurious failure 239 detection where the sender incorrectly marks a destination address 240 as Failed (inactive) during periods of temporary congestion. As 242 [RFC4960] recommends for a coupling of the PMR value and the 243 protocol parameter Association.Max.Retrans (AMR) value such 244 spurious failure detection risks to carry over to spurious 245 association failure detection and closure. Larger PMR values are 246 preferable to avoid spurious failure detection. 248 From the above observations it is clear that tuning the PMR value 249 involves the following trade off -- a lower value improves 250 performance but increases the chances of spurious failure detection, 251 whereas a higher value degrades performance and reduces spurious 252 failure detection in a wide range of path conditions. Thus, tuning 253 the association's PMR value is an incomplete solution to address the 254 performance impact during failure. 256 SCTP-PF defined in this document introduces the new Potentially 257 Failed (PF) destination address state in SCTP's path management 258 procedure. The new Potentially Failed (PF) destination address state 259 applies to SCTP single-homed operation as well as to SCTP multi-homed 260 operation. The PF state was originally proposed to improve CMT 261 performance [NATARAJAN09]. The PF state is an intermediate state 262 between the Active and Failed states. SCTP's failure detection 263 procedure is modified to include the PF state. The new failure 264 detection algorithm assumes that loss detected by a timeout implies 265 either severe congestion or failure en-route. After a number of 266 consecutive timeouts on a path, the sender is unsure, and marks the 267 corresponding destination address as in the PF state. A PF 268 destination address is not used for data transmission except when it 269 is the only destination address available (e.g., for single-homed 270 SCTP) or in other special cases (discussed below). The new failure 271 detection algorithm requires only sender-side changes. 273 4.2. Specification of the SCTP-PF Algorithm 275 The SCTP-PF operation is specified as follows: 277 1. The sender maintains a new tunable parameter called 278 PotentiallyFailed.Max.Retrans (PFMR). The RECOMMENDED value of 279 PFMR is 0 when SCTP-PF is used. The PFMR defines a new 280 intermediate PF threshold on the destination address error 281 counter at exceed of which the destination address is classified 282 as PF and related PF state actions are to be taken. By standard 283 RFC4960 semantics a destination address is classified as 284 Inactive once the error counter exceeds PMR. Setting PFMR 285 larger to or equal to PMR does not result in definition of a PF 286 threshold for the destination address. I.e., PFMR set larger to 287 or equal to PMR means that the destination address never will be 288 classified as PF. 290 2. The error counter of an active destination address is 291 incremented as specified in [RFC4960]. This means that the 292 error counter of the destination address will be incremented 293 each time the T3-rtx timer expires, or each time a HEARTBEAT 294 chunk is sent when idle and not acknowledged within an RTO. 295 When the value in the destination address error counter exceeds 296 PFMR, the endpoint MUST mark the destination address as in the 297 PF state. 299 3. The PFMR threshold defines the point the destination address no 300 longer is considered a good candidate for data transmission and 301 a SCTP-PF sender SHOULD NOT send data to destination addresses 302 in PF state when alternative destination addresses in active 303 state are available. Specifically this means that: 305 i When there is outbound data to send and the destination 306 address presently used for data transmission is in PF state, 307 the sender SHOULD choose a destination address in active 308 state, if one exists, and failover to deploy this destination 309 address for data transmission. 311 ii When retransmitting data that has timed out and the sender 312 thus by [RFC4960], section 6.4.1, should attempt to pick a 313 new destination address for data retransmission, the sender 314 SHOULD choose an alternate destination transport address in 315 active state if one exists. 317 iii When there is outbound data to send and the SCTP user 318 explicitly requests to send data to a destination address in 319 PF state, the sender SHOULD send the data to an alternate 320 destination address in active state if one exists. 322 When choosing among multiple destination address in active state 323 the following considerations are given: 325 A. An SCTP sender should comply with [RFC4960], section 6.4.1, 326 principles of choosing most divergent source-destination 327 pairs compared with, for i.: the destination address in PF 328 state that it performs a failover from, and for ii.: the 329 destination address towards which the data timed out. Rules 330 for picking the most divergent source-destination pair are 331 an implementation decision and are not specified within this 332 document. 334 B. A SCTP-PF sender MAY choose to send data to a destination 335 address in PF state, even if destination addresses in active 336 state exist, have the SCTP-PF sender other means of 337 information available that disqualifies the destination 338 address in active state from being preferred. However, the 339 discussion of such mechanisms is outside of the scope of the 340 SCTP_PF operation specified in this document. 342 In all cases, the sender MUST NOT change the state of chosen 343 destination address, whether this state be active or PF, and it 344 MUST NOT clear the error counter of the destination address as a 345 result of choosing the destination address for data 346 transmission. 348 4. When the destination addresses are all in PF state or some in PF 349 state and some in inactive state, the sender MUST choose one 350 destination address in PF state and transmit or retransmit data 351 to this destination address using the following rules: 353 A. The sender SHOULD choose the destination in PF state with 354 the lowest error count (fewest consecutive timeouts) for 355 data transmission and transmit or retransmit data to this 356 destination. 358 B. When there are multiple PF destinations with same error 359 count, the sender should let the choice among the multiple 360 PF destination with equal error count be based on the 361 [RFC4960], section 6.4.1, principles of choosing most 362 divergent source-destination pairs when executing 363 (potentially consecutive) retransmission. Rules for picking 364 the most divergent source-destination pair are an 365 implementation decision and are not specified within this 366 document. 368 C. A sender MAY choose to deploy other strategies than the 369 above when choosing among multiple PF destinations have the 370 SCTP-PF sender other means of information available that 371 qualifies a particular destination address for being used. 372 The SCTP-PF protocol operation specified in this document 373 makes no assumption of the existence of such other means of 374 information and specifies for the above as the default 375 operation of an SCTP-PF sender. 377 The sender MUST NOT change the state and the error counter of 378 any destination address regardless of whether it has been chosen 379 for transmission or not. 381 5. HEARTBEAT chunks MUST be send to PF destination addresses 382 regardless of whether the Path Heartbeat function (Section 8.3 383 of [RFC4960]) is enabled for the destination address or not. 384 The HB.interval of the Path Heartbeat function of [RFC4960] MUST 385 be ignored for destination addresses in PF state, instead 386 HEARTBEAT chunks are sent to destination addresses in PF state 387 once per RTO. The HEARTBEAT sending begins upon that a 388 destination address reaches the PF state. When a HEARTBEAT 389 chunk is not acknowledged within the RTO, the sender increments 390 the error counter and exponentially back off the RTO value. If 391 the error counter is less than PMR, the sender transmits another 392 packet containing the HEARTBEAT chunk immediately after timeout 393 expiration on the previous HEARTBEAT. When data is being 394 transmitted to a destination address in the PF state, the 395 transmission of a HEARTBEAT chunk MAY be omitted in case receipt 396 of a SACK of or a T3-rtx timer expiration on the outstanding 397 data can provide equivalent information. Likewise the timeout 398 of a HEARTBEAT chunk MAY be ignored if data is outstanding 399 towards the destination address. 401 6. When the sender receives a HEARTBEAT ACK from a destination 402 address in PF state, the sender MUST clear the error counter of 403 the destination address and transition the destination address 404 back to active state. When the sender resumes data transmission 405 on the destination address, it MUST do this following the 406 prescriptions of Section 7.2 of [RFC4960]. 408 7. Additional (PMR - PFMR) consecutive timeouts on a destination 409 address in PF state confirm the path failure, upon which the 410 destination address transitions to the inactive state. As 411 described in [RFC4960], the sender (i) SHOULD notify the ULP 412 about this state transition, and (ii) transmit HEARTBEAT chunks 413 to the inactive destination address at a lower frequency as 414 described in Section 8.3 of [RFC4960] (when this function is 415 enabled for the destination address). 417 8. Acknowledgments for chunks that have been transmitted to 418 multiple destinations (i.e., a chunk which has been 419 retransmitted to a different destination address than the 420 destination address to which the chunk was first transmitted) 421 MUST NOT clear the error count for an inactive destination 422 address and MUST NOT transition a PF destination address back to 423 active state, since a sender cannot disambiguate whether the ACK 424 was for the original transmission or the retransmission(s). The 425 same ambiguity concerns the related congestion window growth. 426 The bytes of a newly acknowledged chunk which has been 427 transmitted to multiple destination addresses SHOULD be 428 considered for contribution to the congestion window growth 429 towards the destination address where the chunk was last sent. 430 The contribution of the ACKed bytes to the window growth is 431 subject to the prescriptions described in Section 7.2 of 432 [RFC4960] is fulfilled. A SCTP sender MAY apply a different 433 approach for both the error count handling and the congestion 434 control growth handling based on unequivocally information on 435 which destination (including multiple destination addresses) the 436 chunk reached. This document makes no reference to what such 437 unequivocally information could consist of, neither how such 438 unequivocally information could be obtained. The design of such 439 an alternative approach is left to implementations. 441 9. Acknowledgments for chunks that has been transmitted to one 442 destination address only MUST clear the error counter for the 443 destination address and MUST transition a PF destination address 444 back to Active state. This situation can happen when new data 445 is sent to a destination address in the PF state. It can also 446 happen in situations where the destination address is in the PF 447 state due to the occurrence of a spurious T3-rtx timer and 448 Acknowledgments start to arrive for data sent prior to 449 occurrence of the spurious T3-rtx and data has not yet been 450 retransmitted towards other destinations. This document does 451 not specify special handling for detection of or reaction to 452 spurious T3-rtx timeouts, e.g., for special operation vis-a-vis 453 the congestion control handling or data retransmission operation 454 towards a destination address which undergoes a transition from 455 active to PF to active state due to a spurious T3-rtx timeout. 456 But it is noted that this is an area which would benefit from 457 additional attention, experimentation and specification for 458 Single Homed SCTP as well as for Multi Homed SCTP protocol 459 operation. 461 10. The SCTP stack SHOULD provide the ULP with the means to expose 462 the PF state of its destinations as well as the means to notify 463 the state transitions from Active to PF, and vice-versa. When 464 doing this, such an SCTP stack MUST provide the ULP with the 465 means to suppress exposure of PF state and associated state 466 transitions as well. 468 4.2.1. Dormant State Operation 470 In a situation with complete disruption of the communication in 471 between the SCTP Endpoints, the aggressive HEARTBEAT transmissions of 472 SCTP-PF on destination addresses in PF state may make the association 473 enter dormant state faster than a standard [RFC4960] SCTP 474 implementation given the same setting of Path.Max.Retrans (PMR) and 475 Association.Max.Retrans (AMR). For example, an SCTP association with 476 two destination addresses typically would reach dormant state in half 477 the time of an [RFC4960] SCTP implementation in such situations. 478 This is because a SCTP PF sender will send HEARTBEATS and data 479 retransmissions in parallel with RTO intervals when there are 480 multiple destinations addresses in PF state. This argument pressumes 481 that RTO << HB.interval of [RFC4960]. One could use higher values of 482 PMR, which makes the dormant state situations less likely to happen. 483 The downside of increasing the PMR value is that destination address 484 failure detections and notifications of such events to ULP is 485 weakened. 487 A design goal of SCTP-PF is that it should provide the same level of 488 disruption tolerance as an [RFC4960] SCTP implementation with the 489 same Path.Max.Retrans (PMR) and Association.Max.Retrans (AMR) 490 setting. For this reason, SCTP-PF SHOULD perform the following 491 operations during dormant state, while this is an implementation 492 decision in [RFC4960]. 494 a. When the destination addresses are all in inactive state, the 495 sender MUST choose one destination when data is transmitted. The 496 sender MUST NOT change the state and the error counter of any 497 destination address regardless of whether it has been chosen for 498 transmission or not. 500 b. The sender SHOULD choose the destination in inactive state with 501 the lowest error count (fewest consecutive timeouts) for data 502 transmission. When there are multiple destinations with same 503 error count in inactive state, the sender SHOULD attempt to pick 504 the most divergent source - destination pair from the last source 505 - destination pair where failure was observed. Rules for picking 506 the most divergent source-destination pair are an implementation 507 decision and are not specified within this document. To support 508 differentiation of inactive destination addresses based on their 509 error count SCTP will need to allow for increment of the 510 destination address error counters up to some reasonable limit 511 above PMR+1, thus changing the prescriptions of [RFC4960], 512 section 8.3, in this respect. The exact limit to apply is not 513 specified in this document but it is considered reasonable to 514 require for such to be an order of magnitude higher than the PMR 515 value. A sender MAY choose to deploy other strategies that the 516 strategy defined by here. The strategy to prioritize the last 517 active destination address,i.e., the destination address with the 518 fewest error counts is optimal when some paths are permanently 519 inactive, but suboptimal when a path instability is transient. 521 An SCTP-PF implementation MAY keep the operation during dormant state 522 an implementation decision, but it should be careful not to 523 compromise the fault tolerance of the SCTP operation. 525 The above prescriptions for SCTP-PF dormant state handling SHOULD NOT 526 be coupled to the value of the PFMR, but solely to the activation of 527 SCTP-PF logic in an SCTP implementation. It is further noted that 528 also a standard [RFC4960] SCTP implementation can use this mode of 529 operation to improve the fault tolerance (which some implementations 530 already do). 532 4.3. Permanent Failover 534 This section describes an OPTIONAL switchback feature called 535 Permanent Failover which is beneficiary to deploy in certain 536 situations. 538 4.3.1. Background 540 In [RFC4960], an SCTP sender migrates the traffic back to the 541 original primary destination address once this address becomes active 542 again. As the CWND towards the original primary destination address 543 has to be rebuilt once data transfer resumes, the switch back to use 544 the original primary address is not always optimal. Indeed [CARO02] 545 shows that the switch back to the original primary may degrade SCTP 546 performance compared to continuing data transmission on the same 547 path, especially, but not only, in scenarios where this path's 548 characteristics are better. In order to mitigate this performance 549 degradation, the Permanent Failover operation was proposed in 550 [CARO02]. When SCTP changes the destination address due to failover, 551 Permanent Failover operation allows SCTP sender to continue data 552 transmission on the new working path even when the old primary 553 destination address becomes active again. This is achieved by having 554 SCTP perform a switch over of the primary path to the alternative 555 working path rather than having SCTP switch back data transfer to the 556 (previous) primary path. 558 The manner of switch over operation that is most optimal in a given 559 scenario depends on the relative quality of a set primary path versus 560 the quality of alternative paths available as well as it depends on 561 the extent to which it is desired for the mode of operation to 562 enforce traffic distribution over a number of network paths. I.e., 563 load distribution of traffic from multiple SCTP associations may be 564 sought to be enforced by distribution of the set primary paths with 565 [RFC4960] switchback operation. However as [RFC4960] switchback 566 behavior is suboptimal in certain situations, especially in scenarios 567 where a number of equally good paths are available, it is recommended 568 for SCTP to support also, as alternative behavior, the Permanent 569 Failover switch over modes of operation. 571 4.3.2. Permanent Failover Algorithm 573 The Permanent Failover operation requires only sender side changes. 574 The details are: 576 1. The sender maintains a new tunable parameter, called 577 Primary.Switchover.Max.Retrans (PSMR). The PSMR MUST be set 578 greater or equal to the PFMR value. Implementations MUST reject 579 any other values of PSMR. 581 2. When the path error counter on a set primary path exceeds PSMR, 582 the SCTP implementation MUST autonomously select and set a new 583 primary path. 585 3. The primary path selected by the SCTP implementation MUST be the 586 path which at the given time would be chosen for data transfer. 587 A previously failed primary path can be used as data transfer 588 path as per normal path selection when the present data transfer 589 path fails. 591 4. The recommended value of PSMR is PFMR when Permanent Failover is 592 used. This means that no forced switchback to a previously 593 failed primary path is performed. An implementation of Permanent 594 Failover MUST support the setting of PSMR = PFMR. An 595 implementation of Permanent Failover MAY support setting of PSMR 596 > PFMR. 598 5. It MUST be possible to disable the Permanent Failover and obtain 599 the standard switchback operation of [RFC4960]. 601 To support optimal operation in a wider range of network scenarios, 602 it it proposed for an SCTP-PF implementation to implement Permanent 603 Failover operation as an optional feature. The implementation of the 604 Permanent Failover feature is optional for an SCTP-PF implementation. 605 For an SCTP implementation that implements Permanent Failover, this 606 specification RECOMMENDS that the standard RFC4960 switchback 607 operation is retained as the default operation. 609 5. Socket API Considerations 611 This section describes how the socket API defined in [RFC6458] is 612 extended to provide a way for the application to control and observe 613 the SCTP-PF behavior. 615 Please note that this section is informational only. 617 A socket API implementation based on [RFC6458] is, by means of the 618 existing SCTP_PEER_ADDR_CHANGE event, extended to provide the event 619 notification when a peer address enters or leaves the potentially 620 failed state as well as the socket API implementation is extended to 621 expose the potentially failed state of a peer address in the existing 622 SCTP_GET_PEER_ADDR_INFO structure. 624 Furthermore, two new read/write socket options for the level 625 IPPROTO_SCTP and the name SCTP_PEER_ADDR_THLDS and 626 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE are defined as described below. 627 The first socket option is used to control the values of the PFMR and 628 PSMR parameters described in Section 4. The second one controls the 629 exposition of the potentially failed path state. 631 Support for the SCTP_PEER_ADDR_THLDS and 632 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE socket options need also to be 633 added to the function sctp_opt_info(). 635 5.1. Support for the Potentially Failed Path State 637 As defined in [RFC6458], the SCTP_PEER_ADDR_CHANGE event is provided 638 if the status of a peer address changes. In addition to the state 639 changes described in [RFC6458], this event is also provided, if a 640 peer address enters or leaves the potentially failed state. The 641 notification as defined in [RFC6458] uses the following structure: 643 struct sctp_paddr_change { 644 uint16_t spc_type; 645 uint16_t spc_flags; 646 uint32_t spc_length; 647 struct sockaddr_storage spc_aaddr; 648 uint32_t spc_state; 649 uint32_t spc_error; 650 sctp_assoc_t spc_assoc_id; 651 } 653 [RFC6458] defines the constants SCTP_ADDR_AVAILABLE, 654 SCTP_ADDR_UNREACHABLE, SCTP_ADDR_REMOVED, SCTP_ADDR_ADDED, and 655 SCTP_ADDR_MADE_PRIM to be provided in the spc_state field. This 656 document defines in addition to that the new constant 657 SCTP_ADDR_POTENTIALLY_FAILED, which is reported if the affected 658 address becomes potentially failed. 660 The SCTP_GET_PEER_ADDR_INFO socket option defined in [RFC6458] can be 661 used to query the state of a peer address. It uses the following 662 structure: 664 struct sctp_paddrinfo { 665 sctp_assoc_t spinfo_assoc_id; 666 struct sockaddr_storage spinfo_address; 667 int32_t spinfo_state; 668 uint32_t spinfo_cwnd; 669 uint32_t spinfo_srtt; 670 uint32_t spinfo_rto; 671 uint32_t spinfo_mtu; 672 }; 674 [RFC6458] defines the constants SCTP_UNCONFIRMED, SCTP_ACTIVE, and 675 SCTP_INACTIVE to be provided in the spinfo_state field. This 676 document defines in addition to that the new constant 677 SCTP_POTENTIALLY_FAILED, which is reported if the peer address is 678 potentially failed. 680 5.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket Option 682 Applications can control the SCTP-PF behavior by getting or setting 683 the number of consecutive timeouts before a peer address is 684 considered potentially failed or unreachable and before the primary 685 path is changed automatically. This socket option uses the level 686 IPPROTO_SCTP and the name SCTP_PEER_ADDR_THLDS. 688 The following structure is used to access and modify the thresholds: 690 struct sctp_paddrthlds { 691 sctp_assoc_t spt_assoc_id; 692 struct sockaddr_storage spt_address; 693 uint16_t spt_pathmaxrxt; 694 uint16_t spt_pathpfthld; 695 uint16_t spt_pathcpthld; 696 }; 698 spt_assoc_id: This parameter is ignored for one-to-one style 699 sockets. For one-to-many style sockets the application may fill 700 in an association identifier or SCTP_FUTURE_ASSOC. It is an error 701 to use SCTP_{CURRENT|ALL}_ASSOC in spt_assoc_id. 703 spt_address: This specifies which peer address is of interest. If a 704 wild card address is provided, this socket option applies to all 705 current and future peer addresses. 707 spt_pathmaxrxt: Each peer address of interest is considered 708 unreachable, if its path error counter exceeds spt_pathmaxrxt. 710 spt_pathpfthld: Each peer address of interest is considered 711 potentially failed, if its path error counter exceeds 712 spt_pathpfthld. 714 spt_pathcpthld: Each peer address of interest is not considered the 715 primary remote address anymore, if its path error counter exceeds 716 spt_pathcpthld. Using a value of 0xffff disables the selection of 717 a new primary peer address. If an implementation does not support 718 the automatically selection of a new primary address, it should 719 indicate an error with errno set to EINVAL if a value different 720 from 0xffff is used in spt_pathcpthld. Setting of spt_pathcpthld 721 < spt_pathpfthld should be rejected with errno set to EINVAL. An 722 implementation MAY support only setting of spt_pathcpthld = 723 spt_pathpfthld and spt_pathcpthld = 0xffff. In this case it shall 724 reject setting of other values with errno set to EINVAL. 726 5.3. Exposing the Potentially Failed Path State 727 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option 729 Applications can control the exposure of the potentially failed path 730 state in the SCTP_PEER_ADDR_CHANGE event and the 731 SCTP_GET_PEER_ADDR_INFO as described in Section 5.1. The default 732 value is implementation specific. 734 This socket option uses the level IPPROTO_SCTP and the name 735 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE. 737 The following structure is used to control the exposition of the 738 potentially failed path state: 740 struct sctp_assoc_value { 741 sctp_assoc_t assoc_id; 742 uint32_t assoc_value; 743 }; 745 assoc_id: This parameter is ignored for one-to-one style sockets. 746 For one-to-many style sockets the application may fill in an 747 association identifier or SCTP_FUTURE_ASSOC. It is an error to 748 use SCTP_{CURRENT|ALL}_ASSOC in assoc_id. 750 assoc_value: The potentially failed path state is exposed if and 751 only if this parameter is non-zero. 753 6. Security Considerations 755 Security considerations for the use of SCTP and its APIs are 756 discussed in [RFC4960] and [RFC6458]. The logic described here is 757 for sender-side only enabled by configuration and does not have any 758 impacts on protocol messages on the wire. No new chunk type or new 759 field parameter is not required in this document. 761 7. IANA Considerations 763 This document does not create any new registries or modify the rules 764 for any existing registries managed by IANA. 766 8. Proposed Change of Status (to be Deleted before Publication) 768 Initially this work looked to entail some changes of the Congestion 769 Control (CC) operation of SCTP and for this reason the work was 770 proposed as Experimental. These intended changes of the CC operation 771 have since been judged to be irrelevant and are no longer part of the 772 specification. As the specification entails no other potential 773 harmful features, consensus exists in the WG to bring the work 774 forward as PS. 776 Initially concerns have been expressed about the possibility for the 777 mechanism to introduce path bouncing with potential harmful network 778 impacts. These concerns are believed to be unfounded. This issue is 779 addressed in Appendix B. 781 It is noted that the feature specified by this document is 782 implemented by multiple SCTP SW implementations and furthermore that 783 various variants of the solution have been deployed in Telco 784 signaling environments for several years with good results. 786 9. References 788 9.1. Normative References 790 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 791 Requirement Levels", BCP 14, RFC 2119, March 1997. 793 [RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC 794 4960, September 2007. 796 9.2. Informative References 798 [CARO02] Caro Jr., A., Iyengar, J., Amer, P., Heinz, G., and R. 799 Stewart, "A Two-level Threshold Recovery Mechanism for 800 SCTP", Tech report, CIS Dept, University of Delaware , 7 801 2002. 803 [CARO04] Caro Jr., A., Amer, P., and R. Stewart, "End-to-End 804 Failover Thresholds for Transport Layer Multihoming", 805 MILCOM 2004 , 11 2004. 807 [CARO05] Caro Jr., A., "End-to-End Fault Tolerance using Transport 808 Layer Multihoming", Ph.D Thesis, University of Delaware , 809 1 2005. 811 [FALLON08] 812 Fallon, S., Jacob, P., Qiao, Y., Murphy, L., Fallon, E., 813 and A. Hanley, "SCTP Switchover Performance Issues in WLAN 814 Environments", IEEE CCNC 2008, 1 2008. 816 [GRINNEMO04] 817 Grinnemo, K-J. and A. Brunstrom, "Performance of SCTP- 818 controlled failovers in M3UA-based SIGTRAN networks", 819 Advanced Simulation Technologies Conference , 4 2004. 821 [IYENGAR06] 822 Iyengar, J., Amer, P., and R. Stewart, "Concurrent 823 Multipath Transfer using SCTP Multihoming over Independent 824 End-to-end Paths.", IEEE/ACM Trans on Networking 14(5), 10 825 2006. 827 [JUNGMAIER02] 828 Jungmaier, A., Rathgeb, E., and M. Tuexen, "On the use of 829 SCTP in failover scenarios", World Multiconference on 830 Systemics, Cybernetics and Informatics , 7 2002. 832 [NATARAJAN09] 833 Natarajan, P., Ekiz, N., Amer, P., and R. Stewart, 834 "Concurrent Multipath Transfer during Path Failure", 835 Computer Communications , 5 2009. 837 [RFC6458] Stewart, R., Tuexen, M., Poon, K., Lei, P., and V. 838 Yasevich, "Sockets API Extensions for the Stream Control 839 Transmission Protocol (SCTP)", RFC 6458, December 2011. 841 Appendix A. Discussions of Alternative Approaches 843 This section lists alternative approaches for the issues described in 844 this document. Although these approaches do not require to update 845 RFC4960, we do not recommend them from the reasons described below. 847 A.1. Reduce Path.Max.Retrans (PMR) 849 Smaller values for Path.Max.Retrans shorten the failover duration. 850 In fact, this is recommended in some research results [JUNGMAIER02] 851 [GRINNEMO04] [FALLON08]. For example, if when Path.Max.Retrans=0, 852 SCTP switches to another destination address on a single timeout. 853 This smaller value for Path.Max.Retrans can results in spurious 854 failover, which might be a problem. 856 Unlike SCTP-PF, the interval for heartbeat packets is governed by 857 'HB.interval' even during failover process. 'HB.interval' is usually 858 set in the order of seconds (recommended value is 30 seconds). When 859 the primary path becomes inactive, the next HEARTBEAT can be 860 transmitted only seconds later. Meanwhile, the primary path may have 861 recovered. In such situations, post failover, an endpoint is forced 862 to wait on the order of seconds before the endpoint can resume 863 transmission on the primary path. However, using smaller value for 864 'HB.interval' might help this situation, but it will be the waste of 865 bandwidth in most cases. 867 In addition, smaller Path.Max.Retrans values also affect 868 'Association.Max.Retrans' values. When the SCTP association's error 869 count (sum of error counts on all ACTIVE paths) exceeds 870 Association.Max.Retrans threshold, the SCTP sender considers the peer 871 endpoint unreachable and terminates the association. Therefore, 872 Section 8.2 in [RFC4960] recommends that Association.Max.Retrans 873 value should not be larger than the summation of the Path.Max.Retrans 874 of each of the destination addresses, else the SCTP sender considers 875 its peer reachable even when all destinations are INACTIVE. To avoid 876 such inconsistent behavior an SCTP implementation SHOULD reduce 877 Association.Max.Retrans accordingly whenever it reduces 878 Path.Max.Retrans. However, smaller Association.Max.Retrans value 879 increases chances of association termination during minor congestion 880 events. 882 A.2. Adjust RTO related parameters 884 As several research results indicate, we can also shorten the 885 duration of failover process by adjusting RTO related parameters 886 [JUNGMAIER02] [FALLON08]. During failover process, RTO keeps being 887 doubled. However, if we can choose smaller value for RTO.max, we can 888 stop the exponential growth of RTO at some point. Also, choosing 889 smaller values for RTO.initial or RTO.min can contribute to keep RTO 890 value small. 892 Similar to reducing Path.Max.Retrans, the advantage of this approach 893 is that it requires no modification to the current specification, 894 although it needs to ignore several recommendations described in the 895 Section 15 of [RFC4960]. However, this approach requires to have 896 enough knowledge about the network characteristics between end 897 points. Otherwise, it can introduce adverse side-effects such as 898 spurious timeouts. 900 Appendix B. Discussions for Path Bouncing Effect 902 The methods described in the document can accelerate the failover 903 process. Hence, they might introduce the path bouncing effect where 904 the sender keeps changing the data transmission path frequently. 905 This sounds harmful to the data transfer, however several research 906 results indicate that there is no serious problem with SCTP in terms 907 of path bouncing effect [CARO04] [CARO05]. 909 There are two main reasons for this. First, SCTP is basically 910 designed for multipath communication, which means SCTP maintains all 911 path related parameters (CWND, ssthresh, RTT, error count, etc) per 912 each destination address. These parameters cannot be affected by 913 path bouncing. In addition, when SCTP migrates the data transfer to 914 another path, it starts with the minimal or the initial CWND. Hence, 915 there is little chance for packet reordering or duplicating. 917 Second, even if all communication paths between the end-nodes share 918 the same bottleneck, the SCTP-PF results in a behavior already 919 allowed by [RFC4960]. 921 Appendix C. SCTP-PF for SCTP Single-homed Operation 923 For a single-homed SCTP association the only tangible effect of the 924 activation of SCTP-PF operation is enhanced failure detection in 925 terms of potential notification of the PF state of the sole 926 destination address as well as, for idle associations, more rapid 927 entering, and notification, of inactive state of the destination 928 address and more rapid end-point failure detection. It is believed 929 that neither of these effects are harmful, provided adequate dormant 930 state operation is implemented, and furthermore that they may be 931 particularly useful for applications that deploys multiple SCTP 932 associations for load balancing purposes. The early notification of 933 the PF state may be used for preventive measures as the entering of 934 the PF state can be used as a warning of potential congestion. 935 Depending on the PMR value, the aggressive HEARTBEAT transmission in 936 PF state may speed up the end-point failure detection (exceed of AMR 937 threshold on the sole path error counter) on idle associations in 938 case where relatively large HB.interval value compared to RTO (e.g. 939 30secs) is used. 941 Authors' Addresses 942 Yoshifumi Nishida 943 GE Global Research 944 2623 Camino Ramon 945 San Ramon, CA 94583 946 USA 948 Email: nishida@wide.ad.jp 950 Preethi Natarajan 951 Cisco Systems 952 510 McCarthy Blvd 953 Milpitas, CA 95035 954 USA 956 Email: prenatar@cisco.com 958 Armando Caro 959 BBN Technologies 960 10 Moulton St. 961 Cambridge, MA 02138 962 USA 964 Email: acaro@bbn.com 966 Paul D. Amer 967 University of Delaware 968 Computer Science Department - 434 Smith Hall 969 Newark, DE 19716-2586 970 USA 972 Email: amer@udel.edu 974 Karen E. E. Nielsen 975 Ericsson 976 Kistavaegen 25 977 Stockholm 164 80 978 Sweden 980 Email: karen.nielsen@tieto.com