idnits 2.17.1 draft-ietf-tsvwg-sctp-failover-05.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 17 pages Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The abstract seems to contain references ([RFC4960]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 20, 2014) is 3565 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 4960 (Obsoleted by RFC 9260) Summary: 2 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group Y. Nishida 3 Internet-Draft GE Global Research 4 Intended status: Experimental P. Natarajan 5 Expires: January 21, 2015 Cisco Systems 6 A. Caro 7 BBN Technologies 8 P. Amer 9 University of Delaware 10 K. Nielsen 11 Ericsson 12 July 20, 2014 14 Quick Failover Algorithm in SCTP 15 draft-ietf-tsvwg-sctp-failover-05.txt 17 Abstract 19 One of the major advantages of SCTP is that it supports 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 [RFC4960] specified 23 failover operation is followed there can be a significant delay in 24 the migration to the active destination addresses, thus severely 25 reducing the effectiveness of SCTP multi-homed operation. 27 The memo complements RFC4960 by the introduction of the Potentially 28 Failed state and associated new Quick Failover operation to apply 29 during network failure and specifies for SCTP senders to support this 30 more performance optimal failover procedure as an add-on to the 31 [RFC4960] failover operation. The memo in addition complements 32 [RFC4960] by introduction of alternative switchover operation modes 33 for the data transfer path management after a failover. These 34 operation modes offer for more performance optimal operation in some 35 network environments. From the perspective of this memo the 36 implementation of the additional switchover operation modes is 37 considered optional. 39 The procedures defined require only minimal modifications to the 40 current specification. The procedures are sender-side only and do 41 not impact the SCTP receiver. 43 Status of This Memo 45 This Internet-Draft is submitted in full conformance with the 46 provisions of BCP 78 and BCP 79. 48 Internet-Drafts are working documents of the Internet Engineering 49 Task Force (IETF). Note that other groups may also distribute 50 working documents as Internet-Drafts. The list of current Internet- 51 Drafts is at http://datatracker.ietf.org/drafts/current/. 53 Internet-Drafts are draft documents valid for a maximum of six months 54 and may be updated, replaced, or obsoleted by other documents at any 55 time. It is inappropriate to use Internet-Drafts as reference 56 material or to cite them other than as "work in progress." 58 This Internet-Draft will expire on January 21, 2015. 60 Copyright Notice 62 Copyright (c) 2014 IETF Trust and the persons identified as the 63 document authors. All rights reserved. 65 This document is subject to BCP 78 and the IETF Trust's Legal 66 Provisions Relating to IETF Documents 67 (http://trustee.ietf.org/license-info) in effect on the date of 68 publication of this document. Please review these documents 69 carefully, as they describe your rights and restrictions with respect 70 to this document. Code Components extracted from this document must 71 include Simplified BSD License text as described in Section 4.e of 72 the Trust Legal Provisions and are provided without warranty as 73 described in the Simplified BSD License. 75 Table of Contents 77 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 78 2. Conventions and Terminology . . . . . . . . . . . . . . . . . 3 79 3. Issues with the SCTP Path Management . . . . . . . . . . . . 4 80 4. SCTP with Potentially-Failed Destination State (SCTP-PF) . . 5 81 4.1. SCTP-PF Description . . . . . . . . . . . . . . . . . . . 5 82 4.2. Permanent Failover . . . . . . . . . . . . . . . . . . . 9 83 5. Socket API Considerations . . . . . . . . . . . . . . . . . . 10 84 5.1. Support for the Potentially Failed Path State . . . . . . 11 85 5.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket 86 Option . . . . . . . . . . . . . . . . . . . . . . . . . 12 87 5.3. Exposing the Potentially Failed Path State 88 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option . . 13 89 6. Security Considerations . . . . . . . . . . . . . . . . . . . 13 90 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 14 91 8. References . . . . . . . . . . . . . . . . . . . . . . . . . 14 92 8.1. Normative References . . . . . . . . . . . . . . . . . . 14 93 8.2. Informative References . . . . . . . . . . . . . . . . . 14 94 Appendix A. Discussions of Alternative Approaches . . . . . . . 15 95 A.1. Reduce Path.Max.Retrans (PMR) . . . . . . . . . . . . . . 15 96 A.2. Adjust RTO related parameters . . . . . . . . . . . . . . 16 97 Appendix B. Discussions for Path Bouncing Effect . . . . . . . . 16 98 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 17 100 1. Introduction 102 The Stream Control Transmission Protocol (SCTP) as specified in 103 [RFC4960] supports multihoming at the transport layer -- an SCTP 104 association can bind to multiple IP addresses at each endpoint. 105 SCTP's multihoming features include failure detection and failover 106 procedures to provide network interface redundancy and improved end- 107 to-end fault tolerance. 109 In SCTP's current failure detection procedure, the sender must 110 experience Path.Max.Retrans (PMR) number of consecutive failed 111 retransmissions on a destination before detecting a path failure. 112 The sender fails over to an alternate active destination only after 113 failure detection. Until detecting the failover, the sender 114 continues to transmit data on the failed path, which degrades the 115 SCTP performance. Concurrent Multipath Transfer (CMT) [IYENGAR06] is 116 an extension to SCTP and allows the sender to transmit data on 117 multiple paths simultaneously. Research [NATARAJAN09] shows that the 118 current failure detection procedure worsens CMT performance during 119 failover and can be significantly improved by employing a better 120 failover algorithm. 122 This document specifies an alternative failure detection procedure 123 for SCTP that improves the SCTP performance during a failover. 125 Also the operation after a failover impacts the performance of the 126 protocol. With [RFC4960] procedures, SCTP will, after a failover 127 from the primary path, switch back to use the primary path for data 128 transfer as soon as this path becomes available. From a performance 129 perspective, as confirmed in research [CARO02], such a switchback of 130 the data transmission path is not optimal in general. As an optional 131 alternative to the switchback operation of [RFC4960], this document 132 specifies for SCTP to support the Permanent Failover switchover 133 procedures proposed by [CARO02]. 135 2. Conventions and Terminology 137 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 138 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 139 document are to be interpreted as described in [RFC2119]. 141 3. Issues with the SCTP Path Management 143 This section describes issues in the current SCTP to be fixed by the 144 approach described in this document. 146 SCTP can utilize multiple IP addresses for a single SCTP association. 147 Each SCTP endpoint exchanges the list of its usable addresses during 148 initial negotiation with its peer. Then the endpoints select one 149 address from the peer's list and define this as the primary 150 destination. During normal transmission, SCTP sends all user data to 151 the primary destination. Also, it sends heartbeat packets to all 152 idle destinations at a certain interval to check the reachability of 153 the path. Idle destinations normally include all non-primary 154 destinations. 156 If a sender has multiple active destination addresses, it can 157 retransmit data to secondary destination address, when the 158 transmission to the primary times out. 160 When a sender receives an acknowledgment for DATA or HEARTBEAT chunks 161 sent to one of the destination addresses, it considers that 162 destination to be active. If it fails to receive acknowledgments, 163 the error count for the address is increased. If the error counter 164 exceeds the protocol parameter 'Path.Max.Retrans', SCTP endpoint 165 considers the address to be inactive. 167 The failover process of SCTP is initiated when the primary path 168 becomes inactive (error counter for the primary path exceeds 169 Path.Max.Retrans). If the primary path is marked inactive, SCTP 170 chooses a new destination address from one of the active destinations 171 and start using this address to send data to. If the primary path 172 becomes active again, SCTP uses the primary destination for 173 subsequent data transmissions and stop using non-primary one. 175 One issue with this failover process is that it usually takes 176 significant amount of time before SCTP switches to the new 177 destination. Let's say the primary path on a multi-homed host 178 becomes unavailable and the RTO value for the primary path at that 179 time is around 1 second, it usually takes over 60 seconds before SCTP 180 starts to use the secondary path. This is because the recommended 181 value for Path.Max.Retrans in the standard is 5, which requires 6 182 consecutive timeouts before failover takes place. Before SCTP 183 switches to the secondary address, SCTP keeps trying to send packets 184 to the primary and only retransmitted packets are sent to the 185 secondary and can thus be reached at the receiver. This slow 186 failover process can cause significant performance degradation and 187 will not be acceptable in some situations. 189 Another issue is that once the primary path is active again, the 190 traffic is switched back. This is not optimal in some situations. 191 This is further discussed in Section 4.2. 193 4. SCTP with Potentially-Failed Destination State (SCTP-PF) 195 In order to address the issues described in Section 3, We propose to 196 update SCTP path management scheme as follows. 198 4.1. SCTP-PF Description 200 SCTP-PF stems from the following two observations about SCTP's 201 failure detection procedure: 203 o In order to minimize performance impact during failover, the 204 sender should avoid transmitting data to the failed destination as 205 early as possible. In the current SCTP path management scheme, 206 the sender stops transmitting data to a destination only after the 207 destination is marked Failed (inactive). Thus, a smaller PMR 208 value is ideal so that the sender transitions a destination to the 209 Failed (inactive) state quicker. 211 o Smaller PMR values increase the chances of spurious failure 212 detection where the sender incorrectly marks a destination as 213 Failed (inactive) during periods of temporary congestion. As 214 [RFC4960] recommends for a coupling of the PMR value and the AMR 215 value such spurious failure detection risks to carry over to 216 spurious association failure detection and closure. Larger PMR 217 values are preferable to avoid spurious failure detection. 219 From the above observations it is clear that tweaking the PMR value 220 involves the following tradeoff -- a lower value improves performance 221 but increases the chances of spurious failure detection, whereas a 222 higher value degrades performance and reduces spurious failure 223 detection in a wide range of path conditions. Thus, tweaking the 224 association's PMR value is an incomplete solution to address 225 performance impact during failure. 227 This proposal introduces a new "Potentially-Failed" (PF) destination 228 state in SCTP's path management procedure. The PF state was 229 originally proposed to improve CMT performance [NATARAJAN09]. The PF 230 state is an intermediate state between Active and Failed states. 231 SCTP's failure detection procedure is modified to include the PF 232 state. The new failure detection algorithm assumes that loss 233 detected by a timeout implies either severe congestion or failure en- 234 route. After a number of consecutive timeouts on a path, the sender 235 is unsure, and marks the corresponding destination as PF. A PF 236 destination is not used for data transmission except in special cases 237 (discussed below). The new failure detection algorithm requires only 238 sender-side changes. The details are: 240 1. The sender maintains a new tunable parameter called Potentially- 241 Failed.Max.Retrans (PFMR). The recommended value of PFMR = 0 242 when quick failover is used. When PFMR is larger or equal to 243 PMR, quick failover is turned off. 245 2. The error counter of an active destination address is 246 incremented as specified in [RFC4960]. This means that the 247 error counter of the destination address will be incremented 248 each time the T3-rtx timer expires, or at times where a 249 HEARTBEAT sent to an idle, active address is not acknowledged 250 within an RTO. When the value in the destination address error 251 counter exceeds PFMR, the endpoint MUST mark the destination 252 transport address as PF. 254 3. The sender SHOULD avoid data transmission to PF destinations. 255 When the destinations are all in PF state or some in PF state 256 and some in inactive state, the sender MUST choose one 257 destination in PF state and transmit data to this destination. 258 The sender SHOULD choose the destination in PF state with least 259 error count (fewest consecutive timeouts) for data transmission 260 and transmit data to this destination. In case of multiple PF 261 destinations with same error count, the sender SHOULD let the 262 choice among the multiple PF destination with equal error count 263 be based on the [RFC4960], section 6.4.1, principles of choosing 264 most divergent source-destination pairs when executing 265 (potentially consecutive) retransmission. This means that the 266 sender SHOULD attempt to pick the most divergent source - 267 destination pair from the last source - destination pair on 268 which data were transmitted or retransmitted. Rules for picking 269 the most divergent source-destination pair are an implementation 270 decision and are not specified within this document. A sender 271 MAY choose to deploy other strategies than the above when 272 choosing among multiple PF destinations with equal error count. 273 In all cases the sender MUST NOT change the state of chosen 274 destination and it MUST NOT clear the destination's error 275 counter as a result of choosing the destination for data 276 transmission. 278 4. Heartbeats SHOULD be sent to PF destination(s) once per RTO. 279 This means the sender MUST ignore HB.interval for PF 280 destinations. If an heartbeat is unanswered, the sender SHOULD 281 increment the error counter and exponentially back off the RTO 282 value. If error counter is less than PMR, the sender SHOULD 283 transmit another heartbeat immediately after T3-timer 284 expiration. When data is transmitted to a PF destination the 285 transmittal of heartbeats may be omitted as SACK or T3-rtx timer 286 expiration can provide equivalent information. This 287 specification recommends that heartbeats be send to PF 288 destinations independently from whether the Path Heartbeat 289 function (Section 8.3 of [RFC4960]) is enabled for the 290 destination address or not. 292 5. When the sender receives an heartbeat ACK from a PF destination, 293 the sender MUST clear the destination's error counter and 294 transition the PF destination back to Active state. When the 295 sender resumes data transmission on the destination it MUST do 296 this following the prescriptions of Section 7.2 of [RFC4960]. 298 6. Additional (PMR - PFMR) consecutive timeouts on a PF destination 299 confirm the path failure, upon which the destination transitions 300 to the Inactive state. As described in [RFC4960], the sender 301 (i) SHOULD notify ULP about this state transition, and (ii) 302 transmit heartbeats to the Inactive destination at a lower 303 frequency as described in Section 8.3 of [RFC4960] (when this 304 function is enabled for the destination address). 306 7. When all destinations are in inactive state (association dormant 307 state) the sender MUST also choose one destination to transmit 308 data to. The sender SHOULD choose the destination in inactive 309 state with least error count (fewest consecutive timeouts) for 310 data transmission and transmit data to this destination. In 311 case of multiple destinations with same error count in inactive 312 state, the sender SHOULD attempt to pick the most divergent 313 source - destination pair from the last source - destination 314 pair on which data were transmitted or retransmitted following 315 [RFC4960]. Rules for picking the most divergent source- 316 destination pair are an implementation decision and are not 317 specified within this document. In order to support this 318 prescription a sender SHOULD allow for increment of the 319 destination error counters up to some reasonable limit above 320 PMR+1, thus changing the prescriptions of [RFC4960], section 321 8.3, in this respect. The exact limit to apply is not specified 322 in this document but it is considered reasonable to require for 323 such to be an order of magnitude higher than the PMR value. A 324 sender MAY choose to deploy other strategies than the above. 325 For example, a sender could choose to prioritize the last active 326 destination during dormant state. The strategy to prioritize 327 the last active destination is optimal when some paths are 328 permanently inactive, but suboptimal when paths' instability is 329 transient. While the increment of the error counters above 330 PMR+1 is a prerequisite for the error counter values to serve to 331 guide the path selection in dormant state, then it is noted that 332 by virtue of the introduction of the Potentially Failed state, 333 one may deploy higher values of PMR without compromising the 334 efficiency of the failover operation, and thus making the 335 increase of path error counters above PMR+1 less critical as the 336 dormant state will be less likely to happen. The downside of 337 increasing the PMR value relative to the AMR value, however, is 338 that the per destination address failure detection and 339 notification of such to ULP thereby is weakened. In all cases 340 the sender MUST NOT change the state of the chosen destination 341 and it MUST NOT clear the destination's error counter as a 342 result of choosing the destination for data transmission. 344 8. ACKs for chunks which have been transmitted to multiple 345 destinations (i.e., a chunk which has been retransmitted to a 346 different destination than the destination to which the chunk 347 was first transmitted) SHOULD NOT clear the error count of an 348 inactive destination and SHOULD NOT transition a PF destination 349 back to Active state, since a sender cannot disambiguate whether 350 the ack was for the original transmission or the 351 retransmission(s). The same ambiguity concerns the related 352 congestion window growth. In this respect then it is specified 353 that bytes of a newly acknowledged chunk which has been 354 transmitted to multiple destinations SHOULD, when the conditions 355 for such contribution is fulfilled following the prescriptions 356 of Section 7.2 of [RFC4960], contribute to the congestion window 357 growth towards the destination to which the chunk was last 358 transmitted. A SCTP sender MAY apply a different approach for 359 both the error count handling as well as the congestion control 360 growth handling does it have unequivocally information as to 361 which destination (including multiple destinations) the chunk 362 reached. This document makes no reference to what such 363 unequivocally information could consist of, neither how such 364 unequivocally information could be obtained. The implementation 365 of such an alternative approach is left to implementations. 367 9. ACKs for chunks which has been transmitted to one destination 368 address only MUST clear the error counter of the destination 369 address and MUST transition a PF destination back to Active 370 state. This situation can happen when new data is sent to a 371 destination address in PF state. It can also happen in 372 situations where the destination address is in PF state due to 373 the occurrence of a spurious T3-rtx timer and ACKs start to 374 arrive for data sent prior to occurrence of the spurious T3-rtx 375 and data has not yet been retransmitted towards other 376 destinations. This document does not specify special handling 377 for detection of or reaction to spurious T3-rtx timeouts, e.g., 378 for special operation vis-a-vis the congestion control handling 379 or data retransmission operation towards a destination address 380 which undergoes a transition from active to PF to active state 381 due to a spurious T3-rtx timeout. But it is noted that this is 382 an area which would benefit from additional attention, 383 experimentation and specification for Single Homed SCTP as well 384 as for Multi Homed SCTP protocol operation. 386 10. SCTP SHOULD provide the means to expose the PF state of its 387 destinations as well as the means to notify the ULP of the state 388 transitions from Active to PF and from PF to Active state . When 389 doing such SCTP MUST provide the means to suppress exposure of 390 PF state and association state transitions and in this case the 391 ULP MAY make SCTP suppress exposure of PF state to ULP. If 392 exposure of PF state is suppressed, the ULP will rely solely on 393 the [RFC4960] state machine even if Quick Failover function is 394 activated in SCTP. 396 4.2. Permanent Failover 398 Post failover then, by [RFC4960] behavior, an SCTP sender migrates 399 the traffic back to the original primary destination once this 400 destination becomes active anew. As the CWND towards the original 401 primary destination has to be rebuilt once data transfer resumes, the 402 switch back to use the original primary path is not always optimal. 403 Indeed [CARO02] shows that the switch back to the original primary 404 may degrade SCTP performance compared to continuing data transmission 405 on the same path, especially, but not only, in scenarios where this 406 path's characteristics are better. In order to mitigate this 407 performance degradation, Permanent Failover operation was proposed in 408 [CARO02]. When SCTP changes the destination due to failover, 409 Permanent Failover operation allows SCTP sender to continue data 410 transmission on the new working path even if the old primary 411 destination becomes active again. This is achieved by having SCTP 412 perform a switch over of the primary path to the alternative working 413 path rather than having SCTP switch back data transfer to the 414 (previous) primary path. 416 The manner of switch over operation that is most optimal in a given 417 scenario depends on the relative quality of a set primary path versus 418 the quality of alternative paths available as well as it depends on 419 the extent to which it is desired for the mode of operation to 420 enforce traffic distribution over a number of network paths. I.e., 421 load distribution of traffic from multiple SCTP associations may be 422 sought to be enforced by distribution of the set primary paths with 423 [RFC4960] switchback operation. However as [RFC4960] switchback 424 behavior is suboptimal in certain situations, especially in scenarios 425 where a number of equally good paths are available, it is recommended 426 for SCTP to support also, as alternative behavior, the Permanent 427 Failover switch over modes of operation. 429 The Permanent Failover operation requires only sender side changes. 430 The details are: 432 1. The sender maintains a new tunable parameter, called 433 Primary.Switchover.Max.Retrans (PSMR). The PSMR SHOULD be set 434 greater or equal to the PFMR value. Any setting of PSMR < PFMR 435 MUST be rejected by the implementation. 437 2. When the path error counter on a set primary path exceeds PSMR, 438 the SCTP implementation autonomously selects and sets a new 439 primary path. 441 3. The primary path selected by the SCTP implementation shall be the 442 path which at the given time would be chosen for data transfer. 443 A previously failed primary path may come in use as data transfer 444 path as per normal path selection when the present data transfer 445 path fails. 447 4. The recommended value of PSMR is PFMR when Permanent Failover is 448 used. This means that no forced switchback to a previously 449 failed primary path is performed. An implementation of Permanent 450 Failover MUST support set of PSMR = PFMR. An implementation of 451 Permanent Failover MAY support setting of PSMR > PFMR. 453 5. It MUST be possible to disable the Permanent Failover and obtain 454 the standard switchback operation of [RFC4960]. 456 We recommend that SCTP-PF sticks to the standard RFC4960 switchback 457 behavior as default, i.e., switch back to the old primary destination 458 once the destination becomes active again. However in order to 459 support optimal operation in a wider range of network scenarios, an 460 implementation MAY implement Permanent Failover operation as detailed 461 above and MAY enable it based on network configurations or users' 462 requests. 464 5. Socket API Considerations 466 This section describes how the socket API defined in [RFC6458] is 467 extended to provide a way for the application to control and observe 468 the quick failover behavior. 470 Please note that this section is informational only. 472 A socket API implementation based on [RFC6458] is, by means of the 473 existing SCTP_PEER_ADDR_CHANGE event, extended to provide the event 474 notification when a peer address enters or leaves the potentially 475 failed state as well as the socket API implementation is extended to 476 expose the potentially failed state of a peer address in the existing 477 SCTP_GET_PEER_ADDR_INFO structure. 479 Furthermore, two new read/write socket options for the level 480 IPPROTO_SCTP and the name SCTP_PEER_ADDR_THLDS and 481 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE are defined as described below. 482 The first socket option is used to control the values of the PFMR and 483 PSMR parameters described in Section 4. The second one controls the 484 exposition of the potentially failed path state. 486 Support for the SCTP_PEER_ADDR_THLDS and 487 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE socket options need also to be 488 added to the function sctp_opt_info(). 490 5.1. Support for the Potentially Failed Path State 492 As defined in [RFC6458], the SCTP_PEER_ADDR_CHANGE event is provided 493 if the status of a peer address changes. In addition to the state 494 changes described in [RFC6458], this event is also provided, if a 495 peer address enters or leaves the potentially failed state. The 496 notification as defined in [RFC6458] uses the following structure: 498 struct sctp_paddr_change { 499 uint16_t spc_type; 500 uint16_t spc_flags; 501 uint32_t spc_length; 502 struct sockaddr_storage spc_aaddr; 503 uint32_t spc_state; 504 uint32_t spc_error; 505 sctp_assoc_t spc_assoc_id; 506 } 508 [RFC6458] defines the constants SCTP_ADDR_AVAILABLE, 509 SCTP_ADDR_UNREACHABLE, SCTP_ADDR_REMOVED, SCTP_ADDR_ADDED, and 510 SCTP_ADDR_MADE_PRIM to be provided in the spc_state field. This 511 document defines in addition to that the new constant 512 SCTP_ADDR_POTENTIALLY_FAILED, which is reported if the affected 513 address becomes potentially failed. 515 The SCTP_GET_PEER_ADDR_INFO socket option defined in [RFC6458] can be 516 used to query the state of a peer address. It uses the following 517 structure: 519 struct sctp_paddrinfo { 520 sctp_assoc_t spinfo_assoc_id; 521 struct sockaddr_storage spinfo_address; 522 int32_t spinfo_state; 523 uint32_t spinfo_cwnd; 524 uint32_t spinfo_srtt; 525 uint32_t spinfo_rto; 526 uint32_t spinfo_mtu; 527 }; 529 [RFC6458] defines the constants SCTP_UNCONFIRMED, SCTP_ACTIVE, and 530 SCTP_INACTIVE to be provided in the spinfo_state field. This 531 document defines in addition to that the new constant 532 SCTP_POTENTIALLY_FAILED, which is reported if the peer address is 533 potentially failed. 535 5.2. Peer Address Thresholds (SCTP_PEER_ADDR_THLDS) Socket Option 537 Applications can control the quick failover behavior by getting or 538 setting the number of consecutive timeouts before a peer address is 539 considered potentially failed or unreachable and before the primary 540 path is changed automatically. This socket option uses the level 541 IPPROTO_SCTP and the name SCTP_PEER_ADDR_THLDS. 543 The following structure is used to access and modify the thresholds: 545 struct sctp_paddrthlds { 546 sctp_assoc_t spt_assoc_id; 547 struct sockaddr_storage spt_address; 548 uint16_t spt_pathmaxrxt; 549 uint16_t spt_pathpfthld; 550 uint16_t spt_pathcpthld; 551 }; 553 spt_assoc_id: This parameter is ignored for one-to-one style 554 sockets. For one-to-many style sockets the application may fill 555 in an association identifier or SCTP_FUTURE_ASSOC. It is an error 556 to use SCTP_{CURRENT|ALL}_ASSOC in spt_assoc_id. 558 spt_address: This specifies which peer address is of interest. If a 559 wildcard address is provided, this socket option applies to all 560 current and future peer addresses. 562 spt_pathmaxrxt: Each peer address of interest is considered 563 unreachable, if its path error counter exceeds spt_pathmaxrxt. 565 spt_pathpfthld: Each peer address of interest is considered 566 potentially failed, if its path error counter exceeds 567 spt_pathpfthld. 569 spt_pathcpthld: Each peer address of interest is not considered the 570 primary remote address anymore, if its path error counter exceeds 571 spt_pathcpthld. Using a value of 0xffff disables the selection of 572 a new primary peer address. If an implementation does not support 573 the automatically selection of a new primary address, it should 574 indicate an error with errno set to EINVAL if a value different 575 from 0xffff is used in spt_pathcpthld. Setting of spt_pathcpthld 576 < spt_pathpfthld should be rejected with errno set to EINVAL. An 577 implementation MAY support only setting of spt_pathcpthld = 578 spt_pathpfthld and spt_pathcpthld = 0xffff. In this case it shall 579 reject setting of other values with errno set to EINVAL. 581 5.3. Exposing the Potentially Failed Path State 582 (SCTP_EXPOSE_POTENTIALLY_FAILED_STATE) Socket Option 584 Applications can control the exposure of the potentially failed path 585 state in the SCTP_PEER_ADDR_CHANGE event and the 586 SCTP_GET_PEER_ADDR_INFO as described in Section 5.1. The default 587 value is implementation specific. 589 This socket option uses the level IPPROTO_SCTP and the name 590 SCTP_EXPOSE_POTENTIALLY_FAILED_STATE. 592 The following structure is used to control the exposition of the 593 potentially failed path state: 595 struct sctp_assoc_value { 596 sctp_assoc_t assoc_id; 597 uint32_t assoc_value; 598 }; 600 assoc_id: This parameter is ignored for one-to-one style sockets. 601 For one-to-many style sockets the application may fill in an 602 association identifier or SCTP_FUTURE_ASSOC. It is an error to 603 use SCTP_{CURRENT|ALL}_ASSOC in assoc_id. 605 assoc_value: The potentially failed path state is exposed if and 606 only if this parameter is non-zero. 608 6. Security Considerations 610 There are no new security considerations introduced in this document. 612 7. IANA Considerations 614 This document does not create any new registries or modify the rules 615 for any existing registries managed by IANA. 617 8. References 619 8.1. Normative References 621 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 622 Requirement Levels", BCP 14, RFC 2119, March 1997. 624 [RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC 625 4960, September 2007. 627 8.2. Informative References 629 [CARO02] Caro Jr., A., Iyengar, J., Amer, P., Heinz, G., and R. 630 Stewart, "A Two-level Threshold Recovery Mechanism for 631 SCTP", Tech report, CIS Dept, University of Delaware , 7 632 2002. 634 [CARO04] Caro Jr., A., Amer, P., and R. Stewart, "End-to-End 635 Failover Thresholds for Transport Layer Multihoming", 636 MILCOM 2004 , 11 2004. 638 [CARO05] Caro Jr., A., "End-to-End Fault Tolerance using Transport 639 Layer Multihoming", Ph.D Thesis, University of Delaware , 640 1 2005. 642 [FALLON08] 643 Fallon, S., Jacob, P., Qiao, Y., Murphy, L., Fallon, E., 644 and A. Hanley, "SCTP Switchover Performance Issues in WLAN 645 Environments", IEEE CCNC 2008, 1 2008. 647 [GRINNEMO04] 648 Grinnemo, K-J. and A. Brunstrom, "Performance of SCTP- 649 controlled failovers in M3UA-based SIGTRAN networks", 650 Advanced Simulation Technologies Conference , 4 2004. 652 [IYENGAR06] 653 Iyengar, J., Amer, P., and R. Stewart, "Concurrent 654 Multipath Transfer using SCTP Multihoming over Independent 655 End-to-end Paths.", IEEE/ACM Trans on Networking 14(5), 10 656 2006. 658 [JUNGMAIER02] 659 Jungmaier, A., Rathgeb, E., and M. Tuexen, "On the use of 660 SCTP in failover scenarios", World Multiconference on 661 Systemics, Cybernetics and Informatics , 7 2002. 663 [NATARAJAN09] 664 Natarajan, P., Ekiz, N., Amer, P., and R. Stewart, 665 "Concurrent Multipath Transfer during Path Failure", 666 Computer Communications , 5 2009. 668 [RFC6458] Stewart, R., Tuexen, M., Poon, K., Lei, P., and V. 669 Yasevich, "Sockets API Extensions for the Stream Control 670 Transmission Protocol (SCTP)", RFC 6458, December 2011. 672 Appendix A. Discussions of Alternative Approaches 674 This section lists alternative approaches for the issues desribed in 675 this document. Although these approaches do not require to update 676 RFC4960, we do not recommend them from the reasons described below. 678 A.1. Reduce Path.Max.Retrans (PMR) 680 Smaller values for Path.Max.Retrans shorten the failover duration. 681 In fact, this is recommended in some research results [JUNGMAIER02] 682 [GRINNEMO04] [FALLON08]. For example, if when Path.Max.Retrans=0, 683 SCTP switches to another destination on a single timeout. This 684 smaller value for Path.Max.Retrans can results in spurious failover, 685 which might be a problem. 687 Unlike SCTP-PF, the interval for heartbeat packets is governed by 688 'HB.interval' even during failover process. 'HB.interval' is usually 689 set in the order of seconds (recommended value is 30 seconds). When 690 the primary path becomes inactive, the next HB can be transmitted 691 only seconds later. Meanwhile, the primary path may have recovered. 692 In such situations, post failover, an endpoint is forced to wait on 693 the order of seconds before the endpoint can resume transmission on 694 the primary path. However, using smaller value for 'HB.interval' 695 might help this situation, but it will be the waste of bandwidth in 696 most cases. 698 In addition, smaller Path.Max.Retrans values also affect 699 'Association.Max.Retrans' values. When the SCTP association's error 700 count (sum of error counts on all ACTIVE paths) exceeds 701 Association.Max.Retrans threshold, the SCTP sender considers the peer 702 endpoint unreachable and terminates the association. Therefore, 703 Section 8.2 in [RFC4960] recommends that Association.Max.Retrans 704 value should not be larger than the summation of the Path.Max.Retrans 705 of each of the destination addresses, else the SCTP sender considers 706 its peer reachable even when all destinations are INACTIVE. To avoid 707 such inconsistent behavior an SCTP implementation SHOULD reduce 708 Association.Max.Retrans accordingly whenever it reduces 709 Path.Max.Retrans. However, smaller Association.Max.Retrans value 710 increases chances of association termination during minor congestion 711 events. 713 A.2. Adjust RTO related parameters 715 As several research results indicate, we can also shorten the 716 duration of failover process by adjusting RTO related parameters 717 [JUNGMAIER02] [FALLON08]. During failover process, RTO keeps being 718 doubled. However, if we can choose smaller value for RTO.max, we can 719 stop the exponential growth of RTO at some point. Also, choosing 720 smaller values for RTO.initial or RTO.min can contribute to keep RTO 721 value small. 723 Similar to reducing Path.Max.Retrans, the advantage of this approach 724 is that it requires no modification to the current specification, 725 although it needs to ignore several recommendations described in the 726 Section 15 of [RFC4960]. However, this approach requires to have 727 enough knowledge about the network characteristics between end 728 points. Otherwise, it can introduce adverse side-effects such as 729 spurious timeouts. 731 Appendix B. Discussions for Path Bouncing Effect 733 The methods described in the document can accelerate the failover 734 process. Hence, they might introduce the path bouncing effect where 735 the sender keeps changing the data transmission path frequently. 736 This sounds harmful to the data transfer, however several research 737 results indicate that there is no serious problem with SCTP in terms 738 of path bouncing effect [CARO04] [CARO05]. 740 There are two main reasons for this. First, SCTP is basically 741 designed for multipath communication, which means SCTP maintains all 742 path related parameters (CWND, ssthresh, RTT, error count, etc) per 743 each destination address. These parameters cannot be affected by 744 path bouncing. In addition, when SCTP migrates the data transfer to 745 another path, it starts with the minimal or the initial CWND. Hence, 746 there is little chance for packet reordering or duplicating. 748 Second, even if all communication paths between the end-nodes share 749 the same bottleneck, the quick failover results in a behavior already 750 allowed by [RFC4960]. 752 Authors' Addresses 754 Yoshifumi Nishida 755 GE Global Research 756 2623 Camino Ramon 757 San Ramon, CA 94583 758 USA 760 Email: nishida@wide.ad.jp 762 Preethi Natarajan 763 Cisco Systems 764 510 McCarthy Blvd 765 Milpitas, CA 95035 766 USA 768 Email: prenatar@cisco.com 770 Armando Caro 771 BBN Technologies 772 10 Moulton St. 773 Cambridge, MA 02138 774 USA 776 Email: acaro@bbn.com 778 Paul D. Amer 779 University of Delaware 780 Computer Science Department - 434 Smith Hall 781 Newark, DE 19716-2586 782 USA 784 Email: amer@udel.edu 786 Karen E. E. Nielsen 787 Ericsson 788 Kistavaegen 25 789 Stockholm 164 80 790 Sweden 792 Email: karen.nielsen@tieto.com