idnits 2.17.1 draft-ietf-pwe3-fat-pw-06.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 (May 6, 2011) is 4732 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 4379 (Obsoleted by RFC 8029) ** Obsolete normative reference: RFC 4447 (Obsoleted by RFC 8077) Summary: 2 errors (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 PWE3 S. Bryant, Ed. 3 Internet-Draft C. Filsfils 4 Intended status: Standards Track Cisco Systems 5 Expires: November 7, 2011 U. Drafz 6 Deutsche Telekom 7 V. Kompella 8 J. Regan 9 Alcatel-Lucent 10 S. Amante 11 Level 3 Communications 12 May 6, 2011 14 Flow Aware Transport of Pseudowires over an MPLS Packet Switched Network 15 draft-ietf-pwe3-fat-pw-06 17 Abstract 19 Where the payload of a pseudowire comprises a number of distinct 20 flows, it can be desirable to carry those flows over the equal cost 21 multiple paths (ECMPs) that exist in the packet switched network. 22 Most forwarding engines are able to hash based on MPLS label stacks 23 and use this mechanism to balance MPLS flows over ECMPs. 25 This document describes a method of identifying the flows, or flow 26 groups, within pseudowires such that Label Switching Routers can 27 balance flows at a finer granularity than individual pseudowires. 28 The mechanism uses an additional label in the MPLS label stack. END 30 Requirements Language 32 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 33 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 34 document are to be interpreted as described in RFC2119 [RFC2119]. 36 Status of this Memo 38 This Internet-Draft is submitted in full conformance with the 39 provisions of BCP 78 and BCP 79. 41 Internet-Drafts are working documents of the Internet Engineering 42 Task Force (IETF). Note that other groups may also distribute 43 working documents as Internet-Drafts. The list of current Internet- 44 Drafts is at http://datatracker.ietf.org/drafts/current/. 46 Internet-Drafts are draft documents valid for a maximum of six months 47 and may be updated, replaced, or obsoleted by other documents at any 48 time. It is inappropriate to use Internet-Drafts as reference 49 material or to cite them other than as "work in progress." 51 This Internet-Draft will expire on November 7, 2011. 53 Copyright Notice 55 Copyright (c) 2011 IETF Trust and the persons identified as the 56 document authors. All rights reserved. 58 This document is subject to BCP 78 and the IETF Trust's Legal 59 Provisions Relating to IETF Documents 60 (http://trustee.ietf.org/license-info) in effect on the date of 61 publication of this document. Please review these documents 62 carefully, as they describe your rights and restrictions with respect 63 to this document. Code Components extracted from this document must 64 include Simplified BSD License text as described in Section 4.e of 65 the Trust Legal Provisions and are provided without warranty as 66 described in the Simplified BSD License. 68 Table of Contents 70 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 71 1.1. ECMP in Label Switching Routers . . . . . . . . . . . . . 5 72 1.2. Flow Label . . . . . . . . . . . . . . . . . . . . . . . . 5 73 2. Native Service Processing Function . . . . . . . . . . . . . . 6 74 3. Pseudowire Forwarder . . . . . . . . . . . . . . . . . . . . . 6 75 3.1. Encapsulation . . . . . . . . . . . . . . . . . . . . . . 7 76 4. Signaling the Presence of the Flow Label . . . . . . . . . . . 8 77 4.1. Structure of Flow Label Sub-TLV . . . . . . . . . . . . . 9 78 5. Static Pseudowires . . . . . . . . . . . . . . . . . . . . . . 9 79 6. Multi-Segment Pseudowires . . . . . . . . . . . . . . . . . . 10 80 7. OAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 81 8. Applicability of PWs using Flow Labels . . . . . . . . . . . . 11 82 8.1. Equal Cost Multiple Paths . . . . . . . . . . . . . . . . 12 83 8.2. Link Aggregation Groups . . . . . . . . . . . . . . . . . 13 84 8.3. Multiple RSVP-TE Paths . . . . . . . . . . . . . . . . . . 13 85 8.4. The Single Large Flow Case . . . . . . . . . . . . . . . . 14 86 8.5. Applicability to MPLS-TP . . . . . . . . . . . . . . . . . 15 87 8.6. Asymmetric Operation . . . . . . . . . . . . . . . . . . . 15 88 9. Applicability to MPLS LSPs . . . . . . . . . . . . . . . . . . 15 89 10. Security Considerations . . . . . . . . . . . . . . . . . . . 16 90 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 91 12. Congestion Considerations . . . . . . . . . . . . . . . . . . 16 92 13. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 17 93 14. References . . . . . . . . . . . . . . . . . . . . . . . . . . 17 94 14.1. Normative References . . . . . . . . . . . . . . . . . . . 17 95 14.2. Informative References . . . . . . . . . . . . . . . . . . 18 96 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 19 98 1. Introduction 100 A pseudowire (PW) [RFC3985] is normally transported over one single 101 network path, even if multiple Equal Cost Multiple Paths (ECMP) exit 102 between the ingress and egress PW provider edge (PE) 103 equipments[RFC4385] [RFC4928]. This is required to preserve the 104 characteristics of the emulated service (e.g. to avoid misordering 105 SAToP PW packets [RFC4553] or subjecting the packets to unusable 106 inter-arrival times ). The use of a single path to preserve order 107 remains the default mode of operation of a PW. The new capability 108 proposed in this document is an OPTIONAL mode which may be used when 109 the use of ECMP is known to be beneficial (and not harmful) to the 110 operation of the PW. 112 Some PWs are used to transport large volumes of IP traffic between 113 routers. One example of this is the use of an Ethernet PW to create 114 a virtual direct link between a pair of routers. Such PWs may carry 115 from hundred's of Mbps to Gbps of traffic. These PWs only require 116 packet ordering to be preserved within the context of each individual 117 transported IP flow. They do not require packet ordering to be 118 preserved between all packets of all IP flows within the pseudowire. 120 The ability to explicitly configure such a PW to leverage the 121 availability of multiple ECMPs allows for better capacity planning as 122 the statistical multiplexing of a larger number of smaller flows is 123 more efficient than with a smaller set of larger flows. 125 Typically, forwarding hardware can deduce that an IP payload is being 126 directly carried by an MPLS label stack, and it is capable of looking 127 at some fields in packets to construct hash buckets for conversations 128 or flows. However, when the MPLS payload is a PW, an intermediate 129 node has no information on the type PW being carried in the packet. 130 This limits the forwarder at the intermediate node to only being able 131 to make an ECMP choice based on a hash of the label stack. In the 132 case of a PW emulating a high bandwidth trunk, the granularity 133 obtained by hashing the label stack is inadequate for satisfactory 134 load-balancing. The ingress node, however, is in the special 135 position of being able to look at the un-encapsulated packet and 136 spread flows amongst any available ECMPs, or even any Loop-Free 137 Alternates [RFC5286] . This document defines a method to introduce 138 granularity on the hashing of traffic running over PWs by introducing 139 an additional label, chosen by the ingress node, and placed at the 140 bottom of the label stack. 142 In addition to providing an indication of the flow structure for use 143 in ECMP forwarding decisions, the mechanism described in the document 144 may also be used to select flows for distribution over an 802.1ad 145 link aggregation group that has been used in an MPLS network. 147 NOTE: Although Ethernet is frequently referenced as a use case in 148 this RFC, the mechanisms described in this document are general 149 mechanisms that may be applied to any PW type in which there are 150 identifiable flows, and in which there is no requirement to preserve 151 the order between those flows. 153 1.1. ECMP in Label Switching Routers 155 Label switching routers (LSRs) commonly generate a hash of the label 156 stack or some elements of the label stack as a method of 157 discriminating between flows, and use this to distribute those flows 158 over the available ECMPs that exist in the network. Since the label 159 at the bottom of stack is usually the label most closely associated 160 with the flow, this normally provides the greatest entropy, and hence 161 is usually included in the hash. This document describes a method of 162 adding an additional label stack entry (LSE) at the bottom of stack 163 in order to facilitate the load balancing of the flows within a PW 164 over the available ECMPs. A similar design for general MPLS use has 165 also been proposed [I-D.kompella-mpls-entropy-label], Section 9. 167 An alternative method of load balancing by creating a number of PWs 168 and distributing the flows amongst them was considered, but was 169 rejected because: 171 o It did not introduce as much entropy as can be introduced by 172 adding an additional LSE. 174 o It required additional PWs to be set up and maintained. 176 1.2. Flow Label 178 An additional LSE [RFC3032] is interposed between the PW LSE and the 179 control word, or if the control word is not present, between the PW 180 LSE and the PW payload. This additional LSE is called the flow LSE 181 and the label carried by the flow LSE is called the flow label. 182 Indivisible flows within the PW MUST be mapped to the same flow label 183 by the ingress PE. The flow label stimulates the correct ECMP load 184 balancing behaviour in the packet switched network (PSN). On receipt 185 of the PW packet at the egress PE (which knows flow LSE is present) 186 the flow LSE is discarded without processing. 188 Note that the flow label MUST NOT be an MPLS reserved label (values 189 in the range 0..15) [RFC3032], but is otherwise unconstrained by the 190 protocol. 192 Considerations of the TTL value are described in the Security section 193 of this document. The flow LSE can never become the top LSE in 194 normal operation, and hence the TTL in the flow LSE is never used to 195 determine whether the packet should be discarded due to TTL expiry. 196 Therefore there are restrictions on the TTL value. 198 This document does not define a use for the TC bits (formerly known 199 as the EXP bits) in the flow label. Future documents may define a 200 use for these bits, therefore implementations conforming to this 201 specification MUST set the TC bits to zero at the ingress and MUST 202 ignore them at the egress. 204 2. Native Service Processing Function 206 The Native Service Processing (NSP) function [RFC3985] is a component 207 of a PE that has knowledge of the structure of the emulated service 208 and is able to take action on the service outside the scope of the 209 PW. In this case it is required that the NSP in the ingress PE 210 identify flows, or groups of flows within the service, and indicate 211 the flow (group) identity of each packet as it is passed to the 212 pseudowire forwarder. As an example, where the PW type is an 213 Ethernet, the NSP might parse the ingress Ethernet traffic and 214 consider all of the IP traffic. This traffic could then be 215 categorised into flows by considering all traffic with the same 216 source and destination address pair to be a single indivisible flow. 217 Since this is an NSP function, by definition, the method used to 218 identify a flow is outside the scope of the PW design. Similarly, 219 since the NSP is internal to the PE, the method of flow indication to 220 the PW forwarder is outside the scope of this document. 222 3. Pseudowire Forwarder 224 The PW forwarder must be provided with a method of mapping flows to 225 load balanced paths. 227 The forwarder must generate a label for the flow or group of flows. 228 How the flow label values are determined is outside the scope of this 229 document, however the flow label allocated to a flow MUST NOT be an 230 MPLS reserved label and SHOULD remain constant for the life of the 231 flow. It is RECOMMENDED that the method chosen to generate the load 232 balancing labels introduces a high degree of entropy in their values, 233 to maximise the entropy presented to the ECMP selection mechanism in 234 the LSRs in the PSN, and hence distribute the flows as evenly as 235 possible over the available PSN ECMP. The forwarder at the ingress 236 PE prepends the PW control word (if applicable), and then pushes the 237 flow label, followed by the PW label. 239 NOTE: Although this document does not attempt to specify any hash 240 algorithms, it is suggested that any such algorithm should be based 241 on the assumption that there will be a high degree of entropy in the 242 values assigned to the load balancing labels. 244 The forwarder at the egress PE uses the pseudowire label to identify 245 the pseudowire. From the context associated with the pseudowire 246 label, the egress PE can determine whether a flow LSE is present. If 247 a flow LSE is present, it MUST be checked to determine whether it 248 carries a reserved label. If it is a reserved label the packet is 249 processed according to the rules associated with that reserved label, 250 otherwise the LSE is discarded. 252 All other PW forwarding operations are unmodified by the inclusion of 253 the flow LSE. 255 3.1. Encapsulation 257 The PWE3 Protocol Stack Reference Model modified to include flow LSE 258 is shown in Figure 1 below 260 +-------------+ +-------------+ 261 | Emulated | | Emulated | 262 | Ethernet | | Ethernet | 263 | (including | Emulated Service | (including | 264 | VLAN) |<==============================>| VLAN) | 265 | Services | | Services | 266 +-------------+ +-------------+ 267 | Flow | | Flow | 268 +-------------+ Pseudowire +-------------+ 269 |Demultiplexer|<==============================>|Demultiplexer| 270 +-------------+ +-------------+ 271 | PSN | PSN Tunnel | PSN | 272 | MPLS |<==============================>| MPLS | 273 +-------------+ +-------------+ 274 | Physical | | Physical | 275 +-----+-------+ +-----+-------+ 277 Figure 1: PWE3 Protocol Stack Reference Model 279 The encapsulation of a PW with a flow LSE is shown in Figure 2 below 280 +---------------------------+ 281 | | 282 | Payload | 283 | | n octets 284 | | 285 +---------------------------+ 286 | Optional Control Word | 4 octets 287 +---------------------------+ 288 | Flow LSE | 4 octets 289 +---------------------------+ 290 | PW LSE | 4 octets 291 +---------------------------+ 292 | MPLS Tunnel LSE (s) | n*4 octets (four octets per LSE) 293 +---------------------------+ 295 Figure 2: Encapsulation of a pseudowire with a pseudowire flow LSE 297 4. Signaling the Presence of the Flow Label 299 When using the signalling procedures in [RFC4447], a new Pseudowire 300 Interface Parameter Sub-TLV, the Flow Label Sub-TLV (FL Sub-TLV), is 301 used to synchronise the flow label states between the ingress and 302 egress PEs. 304 The absence of a FL Sub-TLV indicates that the PE is unable process 305 flow labels. A PE that is using PW signalling and that does not send 306 a FL Sub-TLV MUST NOT include a flow label in the PW packet. A PE 307 that is using PW signalling and which does not receive a FL Sub-TLV 308 from its peer MUST NOT include a flow label in the PW packet. This 309 preserves backwards compatibility with existing PW specifications. 311 A PE that wishes to send a flow label in a PW packet MUST include in 312 its label mapping message a FL Sub-TLV with T = 1 (see Section 4.1). 314 A PE that is willing to receive a flow label MUST include in its 315 label mapping message a FL Sub-TLV with R = 1 (see Section 4.1). 317 A PE that receives a label mapping message a FL Sub-TLV with R = 0 318 MUST NOT include a flow label in the PW packet. 320 Thus a PE sending a FL Sub-TLV with T = 1 and receiving a FL Sub-TLV 321 with R = 1 MUST include a flow label in the PW packet. Under all 322 other combinations of FL Sub-TLV signalling a PE MUST NOT include a 323 flow label in the PW packet. 325 The signalling procedures in [RFC4447] state that "Processing of the 326 interface parameters should continue when unknown interface 327 parameters are encountered, and they MUST be silently ignored." The 328 signalling procedure described here is therefore backwards compatible 329 with existing implementations. 331 Note that what is signalled is the desire to include the flow LSE in 332 the label stack. The value of the flow label is a local matter for 333 the ingress PE, and the label value itself is not signalled. 335 4.1. Structure of Flow Label Sub-TLV 337 The structure of the flow label TLV is shown in Figure 3. 339 0 1 2 3 340 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 1 341 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 342 | FL=0x17 | Length |T|R| Reserved | 343 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 345 Figure 3: Flow Label Sub-TLV 347 Where: 349 o FL (value 0x17) is the flow label sub-TLV identifier assigned by 350 IANA (seeSection 11 ). 352 o Length is the length of the TLV in octets and is 4. 354 o When T=1 the PE is requesting the ability to send a PW packet that 355 includes a flow label. When T= 0, the PE is indicating that it 356 will not send a PW packet containing a flow label. 358 o When R=1 the PE is able to receive a PW packet with a flow label 359 present. When R=0 the PE is unable to receive a PW packet with 360 the flow label present. 362 o Reserved bits MUST be zero on transmit and MUST be ignored on 363 receive. 365 5. Static Pseudowires 367 If PWE3 signalling [RFC4447] is not in use for a PW, then whether the 368 flow label is used MUST be identically provisioned in both PEs at the 369 PW endpoints. If there is no provisioning support for this option, 370 the default behaviour is not to include the flow label. 372 6. Multi-Segment Pseudowires 374 The flow label mechanism described in this document works on multi- 375 segment PWs without requiring modification to the Switching PEs 376 (S-PEs). This is because the flow LSE is transparent to the label 377 swap operation, and because interface parameter Sub-TLV signalling is 378 transitive. 380 7. OAM 382 The following OAM considerations apply to this method of load 383 balancing. 385 Where the OAM is only to be used to perform a basic test that the PWs 386 have been configured at the PEs, VCCV [RFC5085] messages may be sent 387 using any load balance PW path, i.e. using any value for the flow 388 label. 390 Where it is required to verify that a pseudowire is fully functional 391 for all flows, VCCV [RFC5085] connection verification message MUST be 392 sent over each ECMP path to the pseudowire egress PE. This solution 393 may be difficult to achieve and scales poorly. 395 Under these circumstances, it may be sufficient to send VCCV messages 396 using any load balance pseudowire path because if a failure occurs 397 within the PSN the failure will normally be detected and repaired by 398 the PSN. That is, the PSN's Interior Gateway protocol (IGP) link/ 399 node failure detection mechanism (loss of light, bidirectional 400 forwarding detection [RFC5880] or IGP hello detection), and the IGP 401 convergence will naturally modify the ECMP set of network paths 402 between the Ingress and Egress PE's. Hence the PW is only impacted 403 during the normal IGP convergence time. Note that this period may be 404 reduced if a fast re-route or fast convergence technology is deployed 405 in the network [RFC4090], [RFC5286]. 407 If the failure is related to the individual corruption of a Label 408 Forwarding Information database (LFIB) entry in a router, then only 409 the network path using that specific entry is impacted. If the PW is 410 load balanced over multiple network paths, then this failure can only 411 be detected if, by chance, the transported OAM flow is mapped onto 412 the impacted network path, or if all paths are tested. Since testing 413 all paths may present problems as noted above, other mechanisms to 414 detect this type of error may need to be developed, such as an LSP 415 self test technology. 417 To troubleshoot the MPLS PSN, including multiple paths, the 418 techniques described in [RFC4378] and [RFC4379] can be used. 420 Where the PW OAM is carried out of band (VCCV Type 2) [RFC5085] it is 421 necessary to insert an "MPLS Router Alert Label" in the label stack. 422 The resultant label stack is a follows: 424 +-------------------------------+ 425 | | 426 | VCCV Message | n octets 427 | | 428 +-------------------------------+ 429 | Optional Control Word | 4 octets 430 +-------------------------------+ 431 | Flow label | 4 octets 432 +-------------------------------+ 433 | PW label | 4 octets 434 +-------------------------------+ 435 | Router Alert label | 4 octets 436 +-------------------------------+ 437 | MPLS Tunnel label(s) | n*4 octets (four octets per label) 438 +-------------------------------+ 440 Figure 4: Use of Router Alert Label 442 Note that, depending on the number of labels hashed by the LSR, the 443 inclusion of the Router Alert label may cause the OAM packet to be 444 load balanced to a different path from that taken by the data packets 445 with identical Flow and PW labels. 447 8. Applicability of PWs using Flow Labels 449 A node within the PSN is not able to perform deep-packet-inspection 450 (DPI) of the PW as the PW technology is not self-describing: the 451 structure of the PW payload is only known to the ingress and egress 452 PE devices. The method proposed in this document provides a 453 statistical mitigation of the problem of load balance in those cases 454 where a PE is able to discern flows embedded in the traffic received 455 on the attachment circuit. 457 The methods described in this document are transparent to the PSN and 458 as such do not require any new capability from the PSN. 460 The requirement to load-balance over multiple PSN paths occurs when 461 the ratio between the PW access speed and the PSN's core link 462 bandwidth is large (e.g. >= 10%). ATM and FR are unlikely to meet 463 this property. Ethernet may have this property, and for that reason 464 this document focuses on Ethernet. Applications for other high- 465 access-bandwidth PW's (e.g. Fibre Channel) may be defined in the 466 future. 468 This design applies to MPLS PWs where it is meaningful to de- 469 construct the packets presented to the ingress PE into flows. The 470 mechanism described in this document promotes the distribution of 471 flows within the PW over different network paths. This in turn means 472 that whilst packets within a flow are delivered in order (subject to 473 normal IP delivery perturbations due to topology variation), order is 474 no longer maintained for all packets sent over the PW. It is not 475 proposed to associate a different sequence number with each flow. If 476 sequence number support is required the flow label mechanism MUST NOT 477 be used. 479 Where it is known that the traffic carried by the Ethernet PW is IP 480 the flows can be identified and mapped to an ECMP. Such methods 481 typically include hashing on the source and destination addresses, 482 the protocol ID and higher-layer flow-dependent fields such as TCP/ 483 UDP ports, L2TPv3 Session IDs etc. 485 Where it is known that the traffic carried by the Ethernet PW is 486 non-IP, techniques used for link bundling between Ethernet switches 487 may be reused. In this case however the latency distribution would 488 be larger than is found in the link bundle case. The acceptability 489 of the increased latency is for further study. Of particular 490 importance the Ethernet control frames SHOULD always be mapped to the 491 same PSN path to ensure in-order delivery. 493 8.1. Equal Cost Multiple Paths 495 ECMP in packet switched networks is statistical in nature. The 496 mapping of flows to a particular path does not take into account the 497 bandwidth of the flow being mapped or the current bandwidth usage of 498 the members of the ECMP set. This simplification works well when the 499 distribution of flows is evenly spread over the ECMP set and there 500 are a large number of flows that have low bandwidth relative to the 501 paths. The random allocation of a flow to a path provides a good 502 approximation to an even spread of flows, provided that polarisation 503 effects are avoided. The method defined in this document has the 504 same statistical properties as an IP PSN. 506 ECMP is a load-sharing mechanism that is based on sharing the load 507 over a number of layer 3 paths through the PSN. Often however 508 multiple links exist between a pair of LSRs that are considered by 509 the IGP to be a single link. These are known as link bundles. The 510 mechanism described in this document can also be used to distribute 511 the flows within a PW over the members of the link bundle by using 512 the flow label value to identify candidate flows. How that mapping 513 takes place is outside the scope of this specification. Similar 514 considerations apply to link aggregation groups. 516 There is no mechanism currently defined to indicate the bandwidths in 517 use by specific flows using the fields of the MPLS shim header. 518 Furthermore, since the semantics of the MPLS shim header are fully 519 defined in [RFC3032] and [RFC5462], those fields cannot be assigned 520 semantics to carry this information. This document does not define 521 any semantic for use in the TTL or TC fields of the label entry that 522 carries the flow label, but requires that the flow label itself be 523 selected with a high degree of entropy suggesting that the label 524 value should not be overloaded with additional meaning in any 525 subsequent specification. 527 A different type of load balancing is the desire to carry a PW over a 528 set of PSN links in which the bandwidth of members of the link set is 529 less than the bandwidth of the PW. Proposals to address this problem 530 have been made in the past[I-D.stein-pwe3-pwbonding]. Such a 531 mechanism can be considered complementary to this mechanism. 533 8.2. Link Aggregation Groups 535 A Link Aggregation Group (LAG) is used to bond together several 536 physical circuits between two adjacent nodes so they appear to 537 higher-layer protocols as a single, higher bandwidth "virtual" pipe. 538 These may co-exist in various parts of a given network. An advantage 539 of LAGs is that they reduce the number of routing and signalling 540 protocol adjacencies between devices, reducing control plane 541 processing overhead. As with ECMP, the key problem related to LAGs 542 is that due to inefficiencies in LAG load-distribution algorithms, a 543 particular component of a LAG may experience congestion. The 544 mechanism proposed here may be able to assist in producing a more 545 uniform flow distribution. 547 The same considerations requiring a flow to go over a single member 548 of an ECMP set apply to a member of a LAG. 550 8.3. Multiple RSVP-TE Paths 552 In some networks it is desirable for a Label Edge Router (LER) to be 553 able to load balance a PW across multiple RSVP-TE tunnels. The flow 554 label mechanism described in this document may be used to provide the 555 LER with the required flow information, and necessary entropy to 556 provide this type of load balancing. An example of such a case is 557 the of the flow label mechanism in networks using a link bundle with 558 the all ones component [RFC4201]. 560 Methods by which the LER is configured to apply this type of ECMP is 561 outside the scope of this document. 563 8.4. The Single Large Flow Case 565 Clearly the operator should make sure that the service offered using 566 PW technology and the method described in this document does not 567 exceed the maximum planned link capacity, unless it can be guaranteed 568 that it conforms to the Internet traffic profile of a very large 569 number of small flows. 571 If the NSP cannot access sufficient information to distinguish flows, 572 perhaps because the protocol stack required parsing further into the 573 packet than it is able, then the functionality described in this 574 document does not give any benefits. The most common case where a 575 single flow dominates the traffic on a PW is when it is used to 576 transport enterprise traffic. Enterprise traffic may well consist of 577 a single, large TCP flow, or encrypted flows that cannot be handled 578 by the methods described in this document. 580 An operator has four options under these circumstances: 582 1. The operator can choose to do nothing and the system will work as 583 it does without the flow label. 585 2. The operator can make the customer aware that the service 586 offering has a restriction on flow bandwidth and police flows to 587 that restriction. This would allow customers offering multiple 588 flows to use a larger fraction their access bandwidth, whilst 589 preventing a single flow from consuming a fraction of internal 590 link bandwidth that the operator considered excessive. 592 3. The operator could configure the ingress PE to assign a constant 593 flow label to all high bandwidth flows so that only one path was 594 affected by these flows, 596 4. The operator could configure the ingress PE to assign a random 597 flow label to all high bandwidth flows so as to minimise the 598 disruption to the network as a cost of out of order traffic to 599 the user. 601 The issues described above are mitigated by the following two 602 factors: 604 o Firstly, the customer of a high-bandwidth PW service has an 605 incentive to get the best transport service because an inefficient 606 use of the PSN leads to jitter and eventually to loss to the PW's 607 payload. 609 o Secondly, the customer is usually able to tailor their 610 applications to generate many flows in the PSN. A well-known 611 example is massive data transport between servers which use many 612 parallel TCP sessions. This same technique can be used by any 613 transport protocol: multiple UDP ports, multiple L2TPv3 Session 614 ID's, multiple GRE keys may be used to decompose a large flow into 615 smaller components. This approach may be applied to IPsec 616 [RFC4301] where multiple Security Parameters Indexes (SPIs) may be 617 allocated to the same security association. 619 8.5. Applicability to MPLS-TP 621 The MPLS Transport Profile (MPLS-TP) [RFC5654] requirement 44 states 622 that "MPLS-TP MUST support mechanisms that ensure the integrity of 623 the transported customer's service traffic as required by its 624 associated SLA. Loss of integrity may be defined as packet 625 corruption, reordering, or loss during normal network conditions. " 626 The flow aware transport of a PW reorders packets, therefore MUST NOT 627 be deployed in a network conforming to the MPLS-TP unless these 628 integrity requirements specified in the SLA can be satisfied. 630 8.6. Asymmetric Operation 632 The protocol defined in this document supports the asymmetric 633 inclusion of the flow LSE. Asymmetric operation can be expected when 634 there is asymmetry in the bandwidth requirements making it 635 unprofitable for one PE to perform the flow classification, or when 636 that PE is otherwise unable to perform the classification but is able 637 to receive flow labeled packet from its peer. Asymmetric operation 638 of the PW may also be required when one PE has a high transmission 639 bandwidth requirement, but has a need to receive the entire PW on a 640 single interface in order to perform a processing operation that 641 requires the context of the complete PW (for example policing of the 642 egress traffic). 644 9. Applicability to MPLS LSPs 646 A further application of this technique would be to create a basis 647 for hash diversity without having to peek below the label stack for 648 IP traffic carried over LDP LSPs. Work on the generalisation of this 649 to MPLS has been described in [I-D.kompella-mpls-entropy-label]. 650 This is can be regarded as a complementary, but distinct, approach 651 since although similar consideration may apply to the identification 652 of flows and the allocation of flow label values, the flow labels are 653 imposed by different network components, and the associated 654 signalling mechanisms are different. 656 10. Security Considerations 658 The PW generic security considerations described in [RFC3985] and the 659 security considerations applicable to a specific PW type (for 660 example, in the case of an Ethernet PW [RFC4448] apply. The security 661 considerations in [RFC5920] also apply. 663 It is useful to give consideration to the choice of TTL value in the 664 flow LSE [RFC3032]. The flow LSE is at the bottom of label stack, 665 therefore, even when penultimate hop popping is employed, it will 666 always be will preceded by the PW label on arrival at the PE. If the 667 flow label is inadvertently examined as if it were a normal label, 668 the packet might be forwarded. This can be prevented by setting the 669 associated TTL to 1. Note that this may be a departure from 670 considerations that apply to the general MPLS case. 672 11. IANA Considerations 674 IANA is requested to amend the PW Interface Parameters Sub-TLV type 675 Registry value 0x17 (Flow Label indicator) to refer to this RFC. 677 Parameter Length Description 678 ID 680 0x17 4 Flow Label 682 12. Congestion Considerations 684 The congestion considerations applicable to PWs as described in 685 [RFC3985] and any additional congestion considerations developed at 686 the time of publication apply to this design. 688 The ability to explicitly configure a PW to leverage the availability 689 of multiple ECMP is beneficial to capacity planning as, all other 690 parameters being constant, the statistical multiplexing of a larger 691 number of smaller flows is more efficient than with a smaller number 692 of larger flows. 694 Note that if the classification into flows is only performed on IP 695 packets the behaviour of those flows in the face of congestion will 696 be as already defined by the IETF for packets of that type and no 697 additional congestion processing is required. 699 Where flows that are not IP are classified PW congestion avoidance 700 must be applied to each non-IP load balance group. 702 13. Acknowledgements 704 The authors wish to thank Eric Grey, Kireeti Kompella, Joerg 705 Kuechemann, Wilfried Maas, Luca Martini, Mark Townsley, and Lucy Yong 706 for valuable comments on this document. 708 14. References 710 14.1. Normative References 712 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 713 Requirement Levels", BCP 14, RFC 2119, March 1997. 715 [RFC3032] Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y., 716 Farinacci, D., Li, T., and A. Conta, "MPLS Label Stack 717 Encoding", RFC 3032, January 2001. 719 [RFC4379] Kompella, K. and G. Swallow, "Detecting Multi-Protocol 720 Label Switched (MPLS) Data Plane Failures", RFC 4379, 721 February 2006. 723 [RFC4385] Bryant, S., Swallow, G., Martini, L., and D. McPherson, 724 "Pseudowire Emulation Edge-to-Edge (PWE3) Control Word for 725 Use over an MPLS PSN", RFC 4385, February 2006. 727 [RFC4447] Martini, L., Rosen, E., El-Aawar, N., Smith, T., and G. 728 Heron, "Pseudowire Setup and Maintenance Using the Label 729 Distribution Protocol (LDP)", RFC 4447, April 2006. 731 [RFC4448] Martini, L., Rosen, E., El-Aawar, N., and G. Heron, 732 "Encapsulation Methods for Transport of Ethernet over MPLS 733 Networks", RFC 4448, April 2006. 735 [RFC4553] Vainshtein, A. and YJ. Stein, "Structure-Agnostic Time 736 Division Multiplexing (TDM) over Packet (SAToP)", 737 RFC 4553, June 2006. 739 [RFC4928] Swallow, G., Bryant, S., and L. Andersson, "Avoiding Equal 740 Cost Multipath Treatment in MPLS Networks", BCP 128, 741 RFC 4928, June 2007. 743 [RFC5085] Nadeau, T. and C. Pignataro, "Pseudowire Virtual Circuit 744 Connectivity Verification (VCCV): A Control Channel for 745 Pseudowires", RFC 5085, December 2007. 747 14.2. Informative References 749 [I-D.kompella-mpls-entropy-label] 750 Kompella, K., Drake, J., Amante, S., Henderickx, W., and 751 L. Yong, "The Use of Entropy Labels in MPLS Forwarding", 752 draft-kompella-mpls-entropy-label-02 (work in progress), 753 March 2011. 755 [I-D.stein-pwe3-pwbonding] 756 Stein, Y., Mendelsohn, I., and R. Insler, "PW Bonding", 757 draft-stein-pwe3-pwbonding-01 (work in progress), 758 November 2008. 760 [RFC3985] Bryant, S. and P. Pate, "Pseudo Wire Emulation Edge-to- 761 Edge (PWE3) Architecture", RFC 3985, March 2005. 763 [RFC4090] Pan, P., Swallow, G., and A. Atlas, "Fast Reroute 764 Extensions to RSVP-TE for LSP Tunnels", RFC 4090, 765 May 2005. 767 [RFC4201] Kompella, K., Rekhter, Y., and L. Berger, "Link Bundling 768 in MPLS Traffic Engineering (TE)", RFC 4201, October 2005. 770 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 771 Internet Protocol", RFC 4301, December 2005. 773 [RFC4378] Allan, D. and T. Nadeau, "A Framework for Multi-Protocol 774 Label Switching (MPLS) Operations and Management (OAM)", 775 RFC 4378, February 2006. 777 [RFC5286] Atlas, A. and A. Zinin, "Basic Specification for IP Fast 778 Reroute: Loop-Free Alternates", RFC 5286, September 2008. 780 [RFC5462] Andersson, L. and R. Asati, "Multiprotocol Label Switching 781 (MPLS) Label Stack Entry: "EXP" Field Renamed to "Traffic 782 Class" Field", RFC 5462, February 2009. 784 [RFC5654] Niven-Jenkins, B., Brungard, D., Betts, M., Sprecher, N., 785 and S. Ueno, "Requirements of an MPLS Transport Profile", 786 RFC 5654, September 2009. 788 [RFC5880] Katz, D. and D. Ward, "Bidirectional Forwarding Detection 789 (BFD)", RFC 5880, June 2010. 791 [RFC5920] Fang, L., "Security Framework for MPLS and GMPLS 792 Networks", RFC 5920, July 2010. 794 Authors' Addresses 796 Stewart Bryant (editor) 797 Cisco Systems 798 250 Longwater Ave 799 Reading RG2 6GB 800 United Kingdom 802 Phone: +44-208-824-8828 803 Email: stbryant@cisco.com 805 Clarence Filsfils 806 Cisco Systems 807 Brussels 808 Belgium 810 Email: cfilsfil@cisco.com 812 Ulrich Drafz 813 Deutsche Telekom 814 Muenster 815 Germany 817 Email: Ulrich.Drafz@t-com.net 819 Vach Kompella 820 Alcatel-Lucent 822 Email: Alcatel-Lucent vach.kompella@alcatel-lucent.com 824 Joe Regan 825 Alcatel-Lucent 827 Email: joe.regan@alcatel-lucent.comRegan 829 Shane Amante 830 Level 3 Communications 832 Email: shane@castlepoint.net