idnits 2.17.1 draft-stewart-tsvwg-sctpecn-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: ---------------------------------------------------------------------------- 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 (January 15, 2014) is 3755 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) -- Obsolete informational reference (is this intentional?): RFC 2481 (Obsoleted by RFC 3168) -- Obsolete informational reference (is this intentional?): RFC 2960 (Obsoleted by RFC 4960) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group R. Stewart 3 Internet-Draft Adara Networks 4 Intended status: Standards Track M. Tuexen 5 Expires: July 19, 2014 Muenster Univ. of Appl. Sciences 6 X. Dong 7 Huawei 8 January 15, 2014 10 ECN for Stream Control Transmission Protocol (SCTP) 11 draft-stewart-tsvwg-sctpecn-05.txt 13 Abstract 15 This document describes the addition of the ECN to the Stream Control 16 Transmission Protocol (SCTP). 18 Status of This Memo 20 This Internet-Draft is submitted in full conformance with the 21 provisions of BCP 78 and BCP 79. 23 Internet-Drafts are working documents of the Internet Engineering 24 Task Force (IETF). Note that other groups may also distribute 25 working documents as Internet-Drafts. The list of current Internet- 26 Drafts is at http://datatracker.ietf.org/drafts/current/. 28 Internet-Drafts are draft documents valid for a maximum of six months 29 and may be updated, replaced, or obsoleted by other documents at any 30 time. It is inappropriate to use Internet-Drafts as reference 31 material or to cite them other than as "work in progress." 33 This Internet-Draft will expire on July 19, 2014. 35 Copyright Notice 37 Copyright (c) 2014 IETF Trust and the persons identified as the 38 document authors. All rights reserved. 40 This document is subject to BCP 78 and the IETF Trust's Legal 41 Provisions Relating to IETF Documents 42 (http://trustee.ietf.org/license-info) in effect on the date of 43 publication of this document. Please review these documents 44 carefully, as they describe your rights and restrictions with respect 45 to this document. Code Components extracted from this document must 46 include Simplified BSD License text as described in Section 4.e of 47 the Trust Legal Provisions and are provided without warranty as 48 described in the Simplified BSD License. 50 Table of Contents 52 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 53 2. Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 3 54 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 55 4. Chunk and Parameter Formats . . . . . . . . . . . . . . . . . 3 56 4.1. ECN Support Parameter (32768) . . . . . . . . . . . . . . 3 57 4.2. ECN Echo (12) . . . . . . . . . . . . . . . . . . . . . . 3 58 4.3. CWR Chunk(13) . . . . . . . . . . . . . . . . . . . . . . 4 59 5. Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 5 60 5.1. SCTP Initialization . . . . . . . . . . . . . . . . . . . 5 61 5.2. The SCTP Sender . . . . . . . . . . . . . . . . . . . . . 6 62 5.3. The SCTP Receiver . . . . . . . . . . . . . . . . . . . . 8 63 5.4. Congestion on the SACK path . . . . . . . . . . . . . . . 9 64 5.5. Retransmitted SCTP Packets . . . . . . . . . . . . . . . 9 65 5.6. SCTP Window Probes . . . . . . . . . . . . . . . . . . . 10 66 6. Security Considerations . . . . . . . . . . . . . . . . . . . 10 67 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10 68 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 10 69 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 10 70 9.1. Normative references . . . . . . . . . . . . . . . . . . 10 71 9.2. Informational References . . . . . . . . . . . . . . . . 10 72 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 11 74 1. Introduction 76 At the time SCTP was initially defined in [RFC2960] ECN - [RFC2481] 77 was still an experimental document. This left the authors of SCTP in 78 a position where they could not directly refer to ECN without 79 creating a normative reference in a standards track document to an 80 experimental RFC. To work around this problem the authors of SCTP 81 decided to add two reserved chunk types for ECN (CWR and ECNE) but 82 did not fully specify how they were to be used except in a vague way 83 within an appendix of the document. This worked around the document 84 reference problem, but left ECN and its implementation for SCTP 85 unspecified. This document is intended to fill in the details of ECN 86 processing in SCTP in a standards track document. 88 This document assumes that the reader is familiar with ECN [RFC3168]. 89 Readers unfamiliar with ECN are strongly encouraged to first read 90 [RFC3168] since this document will not repeat any of the details on 91 how the various IP level bits are set. This document will use the 92 same terminology has [RFC3168]. For example the term ECT is used to 93 indicate that the IP level packet is marked indicating the transport 94 (SCTP) supports ECN. 96 2. Conventions 98 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 99 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 100 document are to be interpreted as described in [RFC2119]. 102 3. Terminology 104 All integer fields defined in this document included in an SCTP 105 packet MUST be transmitted in network byte order, unless otherwise 106 stated. 108 ECT - The term used to indicate that the IP level packet is 109 marked indicating the transport is willing to support ECN for this 110 packet. 112 not-ECT - The term used to indicate that the IP level packet is 113 marked indicating the transport is NOT willing to support ECN for 114 this packet. 116 CE - The term used to indicate that the IP level packet is 117 marked indicating that a router in the network has marked the 118 packet as having experienced congestion 120 4. Chunk and Parameter Formats 122 4.1. ECN Support Parameter (32768) 124 0 1 2 3 125 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 126 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 127 | Parameter Type = 32768 | Parameter Length = 4 | 128 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 130 This parameter is used to indicate the support for ECN. If this 131 parameter is present, the sender of the chunk is indicating that it 132 supports ECN and wishes to use ECN for the newly forming association. 134 Valid Chunk Appearance 136 The ECN Supported Parameter may appear in the INIT, or the INIT-ACK 137 chunk type. 139 4.2. ECN Echo (12) 140 0 1 2 3 141 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 142 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 143 | Chunk Type=12 | Flags=00000000| Chunk Length = 12 | 144 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 145 | Lowest TSN Number | 146 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 147 | Number CE Marked Packets Seen since CWR | 148 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 150 Chunk Flags: 8 bits 152 Set to all zeros on transmit and ignored on receipt. 154 Lowest TSN Number: 32 bits (unsigned integer) 156 This parameter contains the lowest TSN number contained in the 157 last packet received that was marked by the network with a CE 158 indication. 160 Number CE Marked Packets: 32 bits (unsigned integer) 162 This parameter contains the total number of CE marked packets that 163 has been seen since the first CE mark received while waiting for a 164 CWR chunk. Note that the CE counter will overflow from 0xffffffff 165 to 0 if a CWR chunk is not recieved. 167 Note that the appendix of [RFC4960] did not have the field Number CE 168 Marked Packets. Implementations SHOULD accept an 8 byte form of this 169 chunk that does not include this field. In such a case the 170 implementation SHOULD treat the missing field as indicating one CE 171 marked packet for any purpose for which the implementation is using 172 this field. 174 4.3. CWR Chunk(13) 176 0 1 2 3 177 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 178 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 179 | Chunk Type=13 | Flags=0000000R| Chunk Length = 8 | 180 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 181 | TSN Number | 182 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 184 Chunk Flags: 8 bits 186 The R Bit indicates if the CWR is a retransmission of an earlier CWR 187 that may have been lost. If this bit is set, then the TSN number 188 included is the latest TSN that a CWR has been responded to. If the 189 o bit is clear, than the TSN indicated is the latest TSN for that 190 destination. 192 Set to all zeros on transmit and ignored on receipt. 194 TSN Number: 32 bits (unsigned integer) 196 This parameter contains the TSN number to which the sender has 197 reduced his congestion window to. 199 5. Procedures 201 5.1. SCTP Initialization 203 In the SCTP association setup phase, the source and destination SCTP 204 endpoints exchange information about their willingness to use ECN. 205 After the completion of this negotiation, an SCTP sender sets an ECT 206 codepoint in the IP header of data packets to indicate to the network 207 that the transport is capable and willing to participate in ECN for 208 this packet. This indicates to the routers that they may mark this 209 packet with the CE codepoint. 211 If the SCTP association does not wish to use ECN notification for a 212 particular packet, the sending SCTP sets the ECN codepoint to not- 213 ECT, and the SCTP receiver ignores the CE codepoint in the received 214 packet. 216 For this discussion we will call the endpoint initiating the SCTP 217 association as EP-A and the listening SCTP endpoint as EP-Z. 219 Before an SCTP association can use ECN, EP-A sends an INIT chunk 220 which includes the ECN Support parameter. By including the ECN 221 Support parameter the sending endpoint (EP-A) will participate in ECN 222 as both a sender and a receiver. Specifically, as a receiver, it 223 will respond to incoming data packets that have the CE codepoint set 224 in the IP header by sending an ECN Echo chunk bundled with the next 225 outgoing SACK Chunk. As a sender, it will respond to incoming 226 packets that include an ECN Echo chunk by reducing the congestion 227 window and sending a CWR chunk when appropriate. 229 Including an ECN Support parameter in an INIT or INIT-ACK does not 230 commit the SCTP sender to setting the ECT codepoint in any or all of 231 the packets it may transmit. However, the commitment to respond 232 appropriately to incoming packets with the CE codepoint set remains. 234 When EP-Z sends INIT-ACK chunk, it also includes an ECN Support 235 parameter. Including the ECN Support parameter indicates that the 236 SCTP transmitting the INIT-ACK chunk is ECN-Capable. 238 The following rules apply to the use of ECN for an SCTP association. 240 * If the SCTP Endpoint supports ECN a sender of either an INIT or 241 INIT-ACK chunk MUST ALWAYS include the ECN Supported Parameter. 243 * After the exchange of the INIT and INIT-ACK if both endpoints have 244 NOT indicated support of ECN by including an ECN Supported 245 Parameter, then ECT MUST NOT be set on any IP packets sent by any 246 endpoint which is ECN capable. Furthermore upon receiving IP 247 packets with a CE codepoint set, the ECN capable endpoint SHOULD 248 ignore the CE codepoint. 250 * If both endpoints have included an ECN Supported Parameter in the 251 INIT and INIT-ACK exchange, then both endpoints MUST follow the 252 ECN procedures defined in the rest of this document. 254 * A sending endpoint SHOULD set the ECT code points on IP packets 255 that carry Data chunk. This includes IP packets that have other 256 control chunks bundled with the Data. 258 5.2. The SCTP Sender 260 For an SCTP association using ECN, new data packets are transmitted 261 with an ECT codepoint set in the IP header. When only one ECT 262 codepoint is needed by a sender for all packets sent on an SCTP 263 association ECT(0) SHOULD be used. If the sender receives an ECN- 264 Echo chunk packet, then the sender knows that congestion was 265 encountered in the network on the path from the sender to the 266 receiver. The indication of congestion should be treated just as a 267 congestion loss in non-ECN-Capable SCTP. That is, the SCTP source 268 halves the congestion window "cwnd" for the destination address that 269 the sender transmitted the data to and reduces the slow start 270 threshold "ssthresh". A packet containing an ECN-Echo chunk 271 shouldn't trigger new data to be sent. SCTP follows the normal 272 procedures for increasing the congestion window when it receives a 273 packet with a SACK chunk without the ECN Echo chunk. 275 SCTP should not react to congestion indications more than once every 276 round-trip time. That is, the SCTP sender's congestion window should 277 be reduced only once in response to a series of dropped and/or CE 278 packets from a single window of data. In addition, the SCTP source 279 should not decrease the slow-start threshold, ssthresh, if it has 280 been decreased within the last round trip time. 282 One method to accomplish this is as following: 284 1) During association setup, create a new state variable ECN_ECHO_TSN 285 and ECN_ECHO_LAST for each destination. The initial value of 286 these variables are set to the initial TSN that will be assigned 287 minus 1. 289 2) When an ECN Echo chunk arrives, use the TSN in the ECN Echo to 290 establish which destination the packet was sent to. We will call 291 this destination the selected destination.If the chunk cannot be 292 found note that an override is occuring From the selected 293 destination (if found) select its ECN Echo TSN. 295 3) Compare the ECN Echo TSN with the ECN_ECHO_TSN for the selected 296 destination. If an override is not noted and the value of the 297 ECN_ECHO_TSN is greater than the ECN Echo TSN proceed to step 4; 298 else proceed to step 6b. 300 4) Reduce the cwnd and ssthresh for the selected destination the same 301 as if a loss was detected during a fast retransmit. For details, 302 see [RFC4960] Section 7.2.3 and Section 7.2.4. 304 5) Record in the ECN_ECHO_TSN value, the last TSN that was sent and 305 recorded in ECN_ECHO_LAST the TSN number from the ECN Echo Chunk. 307 6a) If the implementation is tracking the number of marked packets, 308 record the value found in the 'Number CE Marked Packets Seen since 309 CWR' field and also add this number to the running loss count. If 310 such a count is not being maintained, then proceed to step 7. 312 6b) If the implementation is tracking the number of marked packets, 313 compare the number in the ECN Echo Chunk TSN to the ECN_ECHO_LAST. 314 If it is greater than ECN_ECHO_LAST, update ECN_ECHO_LAST with 315 this value. Take the difference between the stored 'Number CE 316 Marked Packets' field and the value from the newly arriving 317 'Number CE Marked Packets' and add this difference to the total 318 loss count. Then update the stored 'Number CE Marked Packets' 319 with the ECN Echo Chunk TSN. 321 7) Create a CWR chunk with the value found in the ECN_ECHO_LAST for 322 the selected destination.If an override was noted, set the 'O' bit 323 within the CWR flags. Queue this chunk for transmission to the 324 peer destination. Note if there is already such a chunk in queue 325 to be sent, remove that chunk and replace it with the new chunk. 327 After the sending SCTP reduces its congestion window in response to a 328 ECN Echo, incoming SACKs that continue to arrive can "clock out" 329 outgoing packets as allowed by the reduced congestion window. Note 330 that continued arrival of ECN Echo chunks should still be processed 331 as described above, possibly reducing the cwnd, but always sending a 332 CWR to the receiving SCTP. This assures that the ECN Echo and CWR 333 are robust with regard to loss in either direction and that the 334 implementation, if it desires, can maintain an accurate loss count 335 per destination. 337 Note, originally in the appendix of [RFC4960] a definition was 338 supplied for the ECN Echo chunk. This definition did NOT include the 339 'Number CE Marked Packets' field. An implementation SHOULD accept 340 such a chunk, delineating it from the standards track version by the 341 fact that the length field will be 8 bytes instead of 12. When 342 processing this older style chunk, the 'Number CE Marked Packets' 343 should be treated as if it contains the number 1. This may cause 344 incorrect loss counts but will NOT cause any issues with SCTP's ECN 345 handling. 347 5.3. The SCTP Receiver 349 When an SCTP endpoint first receives a CE data packet at the 350 destination end-system, the SCTP data receiver creates an ECN Echo 351 chunk and records the lowest TSN number found in the data packet. It 352 also sets the 'Number CE Marked Packets' to 1 and queues this chunk 353 for transmission at the next opportunity. If there is any ACK 354 withholding implemented, as in current "delayed-SACK" SCTP 355 implementations where the SCTP receiver can send an SACK for two 356 arriving data packets, then the ECN Echo chunk will not be sent until 357 the SACK is sent. If the next arriving data packet also has the CE 358 codepoint set, then the receiver updates the queued ECN Echo chunk to 359 have a higher TSN value (the lowest one in the newly arriving data 360 packet) and increments the 'Number CE Marked Packets' field in the 361 queued chunk. 363 Multi-homing requires one added restriction upon the ECN Echo chunk, 364 such a chunk MUST be bundled with a SACK, and the SACK MUST follow 365 the ECN Echo Chunk. This ordering is necessary so that the receiver 366 of the ECN Echo chunk will at least one time find the proper 367 destination to which the chunk was originally sent. Without this 368 restriction it is possible a SACK could arrive ahead of the ECN Echo 369 Chunk, no matter what the sending order, causing the sender to free 370 the DATA chunk and thus loose the association with what destination 371 it was sent to. For the same reason we also require the ECN Echo 372 Chunk be earlier in the packet ahead of the SACK so that the SACK is 373 not processed before the ECN Echo Chunk. 375 After transmission of the ECN Echo chunk, usually bundled with the 376 SACK, the receiver does NOT discard the ECN Echo chunk. Instead it 377 keeps the chunk in its queue and continues to send this chunk bundled 378 with at least a SACK chunk on each outgoing packet, updating it as 379 described above if other CE codepoint data packets arrive. The ECN 380 Echo chunk should only be discarded when a CWR Chunk arrives holding 381 a TSN value that is greater than or equal to the value inside the ECN 382 Echo Chunk. 384 This provides robustness against the possibility of a dropped SACK 385 packet carrying an ECN Echo chunk. The SCTP receiver continues to 386 transmit the ECN Echo chunk in subsequent SACK packets until the 387 correct CWR is received. 389 After the receipt of the CWR chunk, acknowledgments for subsequent 390 non-CE data packets will not have an ECN Echo chunk bundled with 391 them. If another CE packet is received by the data receiver, the 392 receiver would once again send SACK packets bundled with a newly 393 created ECN Echo chunk. The receipt of a CWR packet guarantees that 394 the data sender has received the ECN Echo chunk for the TSN 395 specified, and reduced its congestion window at some point *after* it 396 sent the data packet for which the CE codepoint was set. 398 When processing a CWR, it is important that the receiver of the CWR 399 validate the source address from which the CWR came from. It SHOULD 400 match the destination the ECN Echo was sent to unless the override 401 bit is set in the CWR Chunk. 403 5.4. Congestion on the SACK path 405 For the current generation of SCTP congestion control algorithms, 406 pure acknowledgement packets (e.g., packets that do not contain any 407 accompanying data) MUST be sent with the not-ECT codepoint. Current 408 SCTP receivers have no mechanisms for reducing traffic on the SACK- 409 path in response to congestion notification. Mechanisms for 410 responding to congestion on the SACK-path are areas for current and 411 future research. For current SCTP implementations, a single dropped 412 SACK generally has only a very small effect on SCTP's sending rate. 414 5.5. Retransmitted SCTP Packets 416 This document specifies ECN-capable SCTP implementations MUST NOT set 417 either ECT codepoint (ECT(0) or ECT(1)) in the IP header for 418 retransmitted data packets, and that the SCTP data receiver SHOULD 419 ignore the ECN field on arriving data packets that are outside of the 420 receiver's current window. The reasons for this can be found in 421 [RFC3168] Section 6.1.5. 423 5.6. SCTP Window Probes 425 When the SCTP data receiver advertises a zero window, the SCTP data 426 sender sends window probes to determine if the receiver's window has 427 increased. Window probe packets for SCTP do contain user data (one 428 chunk). If a window probe packet is dropped in the network, this 429 loss can be detected by the receiver. Therefore, the SCTP data 430 sender MAY set an ECT codepoint on the initial send of the window 431 probe, but the SCTP sender MUST NOT set the ECT codepoint on 432 retransmissions of that TSN. 434 6. Security Considerations 436 [RFC3168] defines the security considerations for ECN. These same 437 consideration that are described for TCP are applicable to SCTP. 439 7. IANA Considerations 441 TBD 443 8. Acknowledgements 445 Thanks to Richard Scheffenegger for his helpful comments and review. 447 9. References 449 9.1. Normative references 451 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 452 Requirement Levels", BCP 14, RFC 2119, March 1997. 454 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 455 of Explicit Congestion Notification (ECN) to IP", RFC 456 3168, September 2001. 458 [RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC 459 4960, September 2007. 461 9.2. Informational References 463 [RFC2481] Ramakrishnan, K. and S. Floyd, "A Proposal to add Explicit 464 Congestion Notification (ECN) to IP", RFC 2481, January 465 1999. 467 [RFC2960] Stewart, R., Xie, Q., Morneault, K., Sharp, C., 468 Schwarzbauer, H., Taylor, T., Rytina, I., Kalla, M., 469 Zhang, L., and V. Paxson, "Stream Control Transmission 470 Protocol", RFC 2960, October 2000. 472 Authors' Addresses 474 Randall R. Stewart 475 Adara Networks 476 Chapin, SC 29036 477 USA 479 Email: randall@lakerest.net 481 Michael Tuexen 482 Muenster University of Applied Sciences 483 Stegerwaldstr. 39 484 48565 Steinfurt 485 Germany 487 Email: tuexen@fh-muenster.de 489 Xuesong Dong 490 Huawei 491 Pleasanton, CA 94566 492 USA 494 Email: stevedong@huawei.com