idnits 2.17.1 draft-ietf-6man-ipv6-alt-mark-12.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 (October 22, 2021) is 915 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-06) exists of draft-chen-pce-pcep-ifit-04 == Outdated reference: A later version (-08) exists of draft-fz-spring-srv6-alt-mark-01 == Outdated reference: A later version (-08) exists of draft-ietf-idr-sr-policy-ifit-02 -- Obsolete informational reference (is this intentional?): RFC 8321 (Obsoleted by RFC 9341) -- Obsolete informational reference (is this intentional?): RFC 8889 (Obsoleted by RFC 9342) Summary: 0 errors (**), 0 flaws (~~), 4 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 6MAN Working Group G. Fioccola 3 Internet-Draft T. Zhou 4 Intended status: Standards Track Huawei 5 Expires: April 25, 2022 M. Cociglio 6 Telecom Italia 7 F. Qin 8 China Mobile 9 R. Pang 10 China Unicom 11 October 22, 2021 13 IPv6 Application of the Alternate Marking Method 14 draft-ietf-6man-ipv6-alt-mark-12 16 Abstract 18 This document describes how the Alternate Marking Method can be used 19 as a passive performance measurement tool in an IPv6 domain. It 20 defines a new Extension Header Option to encode Alternate Marking 21 information in both the Hop-by-Hop Options Header and Destination 22 Options Header. 24 Status of This Memo 26 This Internet-Draft is submitted in full conformance with the 27 provisions of BCP 78 and BCP 79. 29 Internet-Drafts are working documents of the Internet Engineering 30 Task Force (IETF). Note that other groups may also distribute 31 working documents as Internet-Drafts. The list of current Internet- 32 Drafts is at https://datatracker.ietf.org/drafts/current/. 34 Internet-Drafts are draft documents valid for a maximum of six months 35 and may be updated, replaced, or obsoleted by other documents at any 36 time. It is inappropriate to use Internet-Drafts as reference 37 material or to cite them other than as "work in progress." 39 This Internet-Draft will expire on April 25, 2022. 41 Copyright Notice 43 Copyright (c) 2021 IETF Trust and the persons identified as the 44 document authors. All rights reserved. 46 This document is subject to BCP 78 and the IETF Trust's Legal 47 Provisions Relating to IETF Documents 48 (https://trustee.ietf.org/license-info) in effect on the date of 49 publication of this document. Please review these documents 50 carefully, as they describe your rights and restrictions with respect 51 to this document. Code Components extracted from this document must 52 include Simplified BSD License text as described in Section 4.e of 53 the Trust Legal Provisions and are provided without warranty as 54 described in the Simplified BSD License. 56 Table of Contents 58 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 59 1.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 3 60 1.2. Requirements Language . . . . . . . . . . . . . . . . . . 3 61 2. Alternate Marking application to IPv6 . . . . . . . . . . . . 3 62 2.1. Controlled Domain . . . . . . . . . . . . . . . . . . . . 5 63 2.1.1. Alternate Marking Measurement Domain . . . . . . . . 6 64 3. Definition of the AltMark Option . . . . . . . . . . . . . . 7 65 3.1. Data Fields Format . . . . . . . . . . . . . . . . . . . 7 66 4. Use of the AltMark Option . . . . . . . . . . . . . . . . . . 8 67 5. Alternate Marking Method Operation . . . . . . . . . . . . . 10 68 5.1. Packet Loss Measurement . . . . . . . . . . . . . . . . . 10 69 5.2. Packet Delay Measurement . . . . . . . . . . . . . . . . 11 70 5.3. Flow Monitoring Identification . . . . . . . . . . . . . 13 71 5.4. Multipoint and Clustered Alternate Marking . . . . . . . 15 72 5.5. Data Collection and Calculation . . . . . . . . . . . . . 16 73 6. Security Considerations . . . . . . . . . . . . . . . . . . . 16 74 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 19 75 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 20 76 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 77 9.1. Normative References . . . . . . . . . . . . . . . . . . 20 78 9.2. Informative References . . . . . . . . . . . . . . . . . 20 79 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 22 81 1. Introduction 83 [RFC8321] and [RFC8889] describe a passive performance measurement 84 method, which can be used to measure packet loss, latency and jitter 85 on live traffic. Since this method is based on marking consecutive 86 batches of packets, the method is often referred to as the Alternate 87 Marking Method. 89 This document defines how the Alternate Marking Method can be used to 90 measure performance metrics in IPv6. The rationale is to apply the 91 Alternate Marking methodology to IPv6 and therefore allow detailed 92 packet loss, delay and delay variation measurements both hop-by-hop 93 and end-to-end to exactly locate the issues in an IPv6 network. 95 The Alternate Marking is an on-path telemetry technique and consists 96 of synchronizing the measurements in different points of a network by 97 switching the value of a marking bit and therefore dividing the 98 packet flow into batches. Each batch represents a measurable entity 99 recognizable by all network nodes along the path. By counting the 100 number of packets in each batch and comparing the values measured by 101 different nodes, it is possible to precisely measure the packet loss. 102 Similarly, the alternation of the values of the marking bits can be 103 used as a time reference to calculate the delay and delay variation. 104 The Alternate Marking operation is further described in Section 5. 106 The format of IPv6 addresses is defined in [RFC4291] while [RFC8200] 107 defines the IPv6 Header, including a 20-bit Flow Label and the IPv6 108 Extension Headers. 110 This document introduces a new TLV (type-length-value) that can be 111 encoded in the Options Headers (Hop-by-Hop or Destination) for the 112 purpose of the Alternate Marking Method application in an IPv6 113 domain. 115 The threat model for the application of the Alternate Marking Method 116 in an IPv6 domain is reported in Section 6. As with all on-path 117 telemetry techniques, the only definitive solution is that this 118 methodology MUST be applied in a controlled domain. 120 1.1. Terminology 122 This document uses the terms related to the Alternate Marking Method 123 as defined in [RFC8321] and [RFC8889]. 125 1.2. Requirements Language 127 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 128 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 129 "OPTIONAL" in this document are to be interpreted as described in BCP 130 14 [RFC2119] [RFC8174] when, and only when, they appear in all 131 capitals, as shown here. 133 2. Alternate Marking application to IPv6 135 The Alternate Marking Method requires a marking field. Several 136 alternatives could be considered such as IPv6 Extension Headers, IPv6 137 Address and Flow Label. But, it is necessary to analyze the 138 drawbacks for all the available possibilities, more specifically: 140 Reusing existing Extension Header for Alternate Marking leads to a 141 non-optimized implementation; 143 Using the IPv6 destination address to encode the Alternate Marking 144 processing is very expensive; 145 Using the IPv6 Flow Label for Alternate Marking conflicts with the 146 utilization of the Flow Label for load distribution purpose 147 ([RFC6438]). 149 In the end, a new Hop-by-Hop or a new Destination Option is the best 150 choice. 152 The approach for the Alternate Marking application to IPv6 specified 153 in this memo is compliant with [RFC8200]. It involves the following 154 operations: 156 o The source node is the only one that writes the Option Header to 157 mark alternately the flow (for both Hop-by-Hop and Destination 158 Option). The intermediate nodes and destination node MUST only 159 read the marking values of the option without modifying the Option 160 Header. 162 o In case of Hop-by-Hop Option Header carrying Alternate Marking 163 bits, it is not inserted or deleted, but can be read by any node 164 along the path. The intermediate nodes may be configured to 165 support this Option or not and the measurement can be done only 166 for the nodes configured to read the Option. As further discussed 167 in Section 4, the presence of the hop-by-hop option should not 168 affect the traffic throughput both on nodes that do not recognize 169 this option and on the nodes that support it. However, it is 170 worth mentioning that there is a difference between theory and 171 practice. Indeed, in a real implementation it can happen that 172 packets with hop-by-hop option could also be skipped or processed 173 in the slow path. While some proposals are trying to address this 174 problem and make Hop-by-Hop Options more practical 175 ([I-D.peng-v6ops-hbh], [I-D.hinden-6man-hbh-processing]), these 176 aspects are out of the scope for this document. 178 o In case of Destination Option Header carrying Alternate Marking 179 bits, it is not processed, inserted, or deleted by any node along 180 the path until the packet reaches the destination node. Note 181 that, if there is also a Routing Header (RH), any visited 182 destination in the route list can process the Option Header. 184 Hop-by-Hop Option Header is also useful to signal to routers on the 185 path to process the Alternate Marking. However, as said, routers 186 will only examine this option if properly configured. 188 The optimization of both implementation and scaling of the Alternate 189 Marking Method is also considered and a way to identify flows is 190 required. The Flow Monitoring Identification field (FlowMonID), as 191 introduced in Section 5.3, goes in this direction and it is used to 192 identify a monitored flow. 194 The FlowMonID is different from the Flow Label field of the IPv6 195 Header ([RFC6437]). The Flow Label field in the IPv6 header is used 196 by a source to label sequences of packets to be treated in the 197 network as a single flow and, as reported in [RFC6438], it can be 198 used for load-balancing/equal cost multi-path (LB/ECMP). The reuse 199 of Flow Label field for identifying monitored flows is not considered 200 because it may change the application intent and forwarding behavior. 201 Also, the Flow Label may be changed en route and this may also 202 invalidate the integrity of the measurement. Furthermore, since the 203 Flow Label is pseudo-random, there is always a finite probability of 204 collision. Those reasons make the definition of the FlowMonID 205 necessary for IPv6. Indeed, the FlowMonID is designed and only used 206 to identify the monitored flow. Flow Label and FlowMonID within the 207 same packet are totally disjoint, have different scope, are used to 208 identify flows based on different criteria, and are intended for 209 different use cases. 211 The rationale for the FlowMonID is further discussed in Section 5.3. 212 This 20 bit field allows easy and flexible identification of the 213 monitored flow and enables improved measurement correlation and finer 214 granularity since it can be used in combination with the traditional 215 TCP/IP 5-tuple to identify a flow. An important point that will be 216 discussed in Section 5.3 is the uniqueness of the FlowMonID and how 217 to allow disambiguation of the FlowMonID in case of collision. 219 The following section highlights an important requirement for the 220 application of the Alternate Marking to IPv6. The concept of the 221 controlled domain is explained and it is considered an essential 222 precondition, as also highlighted in Section 6. 224 2.1. Controlled Domain 226 [RFC8799] introduces the concept of specific limited domain solutions 227 and, in this regard, it is reported the IPv6 Application of the 228 Alternate Marking Method as an example. 230 IPv6 has much more flexibility than IPv4 and innovative applications 231 have been proposed, but for a number of reasons, such as the 232 policies, options supported, the style of network management and 233 security requirements, it is suggested to limit some of these 234 applications to a controlled domain. This is also the case of the 235 Alternate Marking application to IPv6 as assumed hereinafter. 237 Therefore, the IPv6 application of the Alternate Marking Method MUST 238 be deployed in a controlled domain. It is RECOMMENDED that an 239 implementation filters packets that carry Alternate Marking data and 240 are entering or leaving the controlled domains. 242 A controlled domain is a managed network where it is required to 243 select, monitor and control the access to the network by enforcing 244 policies at the domain boundaries in order to discard undesired 245 external packets entering the domain and check the internal packets 246 leaving the domain. It does not necessarily mean that a controlled 247 domain is a single administrative domain or a single organization. A 248 controlled domain can correspond to a single administrative domain or 249 can be composed by multiple administrative domains under a defined 250 network management. Indeed, some scenarios may imply that the 251 Alternate Marking Method involves more than one domain, but in these 252 cases, it is RECOMMENDED that the multiple domains create a whole 253 controlled domain while traversing the external domain by employing 254 IPsec [RFC4301] authentication and encryption or other VPN technology 255 that provides full packet confidentiality and integrity protection. 256 In a few words, it must be possible to control the domain boundaries 257 and eventually use specific precautions if the traffic traverse the 258 Internet. 260 The security considerations reported in Section 6 also highlight this 261 requirement. 263 2.1.1. Alternate Marking Measurement Domain 265 The Alternate Marking measurement domain can overlap with the 266 controlled domain or may be a subset of the controlled domain. The 267 typical scenarios for the application of the Alternate Marking Method 268 depend on the controlled domain boundaries, in particular: 270 the user equipment can be the starting or ending node, only in 271 case it is fully managed and if it belongs to the controlled 272 domain. In this case the user generated IPv6 packets contain the 273 Alternate Marking data. But, in practice, this is not common due 274 to the fact that the user equipment cannot be totally secured in 275 the majority of cases. 277 the CPE (Customer Premises Equipment) is most likely to be the 278 starting or ending node since it connects the user's premises with 279 the service provider's network and therefore belongs to the 280 operator's controlled domain. Typically the CPE encapsulates a 281 received packet in an outer IPv6 header which contains the 282 Alternate Marking data. The CPE can also be able to filter and 283 drop packets from outside of the domain with inconsistent fields 284 to make effective the relevant security rules at the domain 285 boundaries, for example a simple security check can be to insert 286 the Alternate Marking data if and only if the destination is 287 within the controlled domain. 289 3. Definition of the AltMark Option 291 The definition of a new TLV for the Options Extension Headers, 292 carrying the data fields dedicated to the Alternate Marking method, 293 is reported below. 295 3.1. Data Fields Format 297 The following figure shows the data fields format for enhanced 298 Alternate Marking TLV (AltMark). This AltMark data can be 299 encapsulated in the IPv6 Options Headers (Hop-by-Hop or Destination 300 Option). 302 0 1 2 3 303 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 1 304 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 305 | Option Type | Opt Data Len | 306 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 307 | FlowMonID |L|D| Reserved | 308 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 310 where: 312 o Option Type: 8-bit identifier of the type of Option that needs to 313 be allocated. Unrecognized Types MUST be ignored on processing. 314 For Hop-by-Hop Options Header or Destination Options Header, 315 [RFC8200] defines how to encode the three high-order bits of the 316 Option Type field. The two high-order bits specify the action 317 that must be taken if the processing IPv6 node does not recognize 318 the Option Type; for AltMark these two bits MUST be set to 00 319 (skip over this Option and continue processing the header). The 320 third-highest-order bit specifies whether the Option Data can 321 change en route to the packet's final destination; for AltMark the 322 value of this bit MUST be set to 0 (Option Data does not change en 323 route). In this way, since the three high-order bits of the 324 AltMark Option are set to 000, it means that nodes can simply skip 325 this Option if they do not recognize and that the data of this 326 Option do not change en route, indeed the source is the only one 327 that can write it. 329 o Opt Data Len: 4. It is the length of the Option Data Fields of 330 this Option in bytes. 332 o FlowMonID: 20-bit unsigned integer. The FlowMon identifier is 333 described in Section 5.3. As further discussed below, it has been 334 picked as 20 bits since it is a reasonable value and a good 335 compromise in relation to the chance of collision if it is set 336 pseudo randomly by the source node or set by a centralized 337 controller. 339 o L: Loss flag for Packet Loss Measurement as described in 340 Section 5.1; 342 o D: Delay flag for Single Packet Delay Measurement as described in 343 Section 5.2; 345 o Reserved: is reserved for future use. These bits MUST be set to 346 zero on transmission and ignored on receipt. 348 4. Use of the AltMark Option 350 The AltMark Option is the best way to implement the Alternate Marking 351 method and it is carried by the Hop-by-Hop Options header and the 352 Destination Options header. In case of Destination Option, it is 353 processed only by the source and destination nodes: the source node 354 inserts and the destination node processes it. While, in case of 355 Hop-by-Hop Option, it may be examined by any node along the path, if 356 explicitly configured to do so. 358 It is important to highlight that the Option Layout can be used both 359 as Destination Option and as Hop-by-Hop Option depending on the Use 360 Cases and it is based on the chosen type of performance measurement. 361 In general, it is needed to perform both end to end and hop by hop 362 measurements, and the Alternate Marking methodology allows, by 363 definition, both performance measurements. In many cases the end-to- 364 end measurement is not enough and it is required the hop-by-hop 365 measurement, so the most complete choice can be the Hop-by-Hop 366 Options Header. 368 IPv6, as specified in [RFC8200], allows nodes to optionally process 369 Hop-by-Hop headers. Specifically the Hop-by-Hop Options header is 370 not inserted or deleted, but may be examined or processed by any node 371 along a packet's delivery path, until the packet reaches the node (or 372 each of the set of nodes, in the case of multicast) identified in the 373 Destination Address field of the IPv6 header. Also, it is expected 374 that nodes along a packet's delivery path only examine and process 375 the Hop-by-Hop Options header if explicitly configured to do so. 377 Another scenario that can be mentioned is the presence of a Routing 378 Header, in particular it is possible to consider SRv6. A new type of 379 Routing Header, referred as Segment Routing Header (SRH), has been 380 defined in [RFC8754] for SRv6. Like any other use case of IPv6, Hop- 381 by-Hop and Destination Options are usable when SRv6 header is 382 present. Because SRv6 is implemented through a Segment Routing 383 Header (SRH), Destination Options before the Routing Header are 384 processed by each destination in the route list, that means, in case 385 of SRH, by every SR node that is identified by the SR path. More 386 details about the SRv6 application are described in 387 [I-D.fz-spring-srv6-alt-mark]. 389 In summary, it is possible to list the alternative possibilities: 391 o Destination Option not preceding a Routing Header => measurement 392 only by node in Destination Address. 394 o Hop-by-Hop Option => every router on the path with feature 395 enabled. 397 o Destination Option preceding a Routing Header => every destination 398 node in the route list. 400 In general, Hop-by-Hop and Destination Options are the most suitable 401 ways to implement Alternate Marking. 403 It is worth mentioning that new Hop-by-Hop Options are not strongly 404 recommended in [RFC7045] and [RFC8200], unless there is a clear 405 justification to standardize it, because nodes may be configured to 406 ignore the Options Header, drop or assign packets containing an 407 Options Header to a slow processing path. In case of the AltMark 408 data fields described in this document, the motivation to standardize 409 a new Hop-by-Hop Option is that it is needed for OAM (Operations, 410 Administration, and Maintenance). An intermediate node can read it 411 or not, but this does not affect the packet behavior. The source 412 node is the only one that writes the Hop-by-Hop Option to mark 413 alternately the flow, so, the performance measurement can be done for 414 those nodes configured to read this Option, while the others are 415 simply not considered for the metrics. 417 The Hop-by-Hop Option defined in this document is designed to take 418 advantage of the property of how Hop-by-Hop options are processed. 419 Nodes that do not support this Option SHOULD ignore them. This can 420 mean that, in this case, the performance measurement does not account 421 for all links and nodes along a path. The definition of the Hop-by- 422 Hop Options in this document is also designed to minimize throughput 423 impact both on nodes that do not recognize the Option and on node 424 that support it. Indeed, the three high-order bits of the Options 425 Header defined in this draft are 000 and, in theory, as per [RFC8200] 426 and [I-D.hinden-6man-hbh-processing], this means "skip if do not 427 recognize and data do not change en route". [RFC8200] also mentions 428 that the nodes only examine and process the Hop-by-Hop Options header 429 if explicitly configured to do so. For these reasons, this Hop-by- 430 Hop Option should not affect the throughput. However, in practice, 431 it is important to be aware that the things may be different in the 432 implementation and it can happen that packets with Hop-by-Hop are 433 forced onto the slow path, but this is a general issue, as also 434 explained in [I-D.hinden-6man-hbh-processing]. It is also worth 435 mentioning that the application to a controlled domain should avoid 436 the risk of arbitrary nodes dropping packets with Hop-by-Hop Options. 438 5. Alternate Marking Method Operation 440 This section describes how the method operates. [RFC8321] introduces 441 several applicable methods which are reported below, and a new field 442 is introduced to facilitate the deployment and improve the 443 scalability. 445 5.1. Packet Loss Measurement 447 The measurement of the packet loss is really straightforward in 448 comparison to the existing mechanisms, as detailed in [RFC8321]. The 449 packets of the flow are grouped into batches, and all the packets 450 within a batch are marked by setting the L bit (Loss flag) to a same 451 value. The source node can switch the value of the L bit between 0 452 and 1 after a fixed number of packets or according to a fixed timer, 453 and this depends on the implementation. The source node is the only 454 one that marks the packets to create the batches, while the 455 intermediate nodes only read the marking values and identify the 456 packet batches. By counting the number of packets in each batch and 457 comparing the values measured by different network nodes along the 458 path, it is possible to measure the packet loss occurred in any 459 single batch between any two nodes. Each batch represents a 460 measurable entity recognizable by all network nodes along the path. 462 Both fixed number of packets and fixed timer can be used by the 463 source node to create packet batches. But, as also explained in 464 [RFC8321], the timer-based batches are preferable because they are 465 more deterministic than the counter-based batches. There is no 466 definitive rule for counter-based batches, differently from timer- 467 based batches. Using a fixed timer for the switching offers better 468 control over the method, indeed the length of the batches can be 469 chosen large enough to simplify the collection and the comparison of 470 the measures taken by different network nodes. In the implementation 471 the counters can be sent out by each node to the controller that is 472 responsible for the calculation. It is also possible to exchange 473 this information by using other on-path techniques. But this is out 474 of scope for this document. 476 Packets with different L values may get swapped at batch boundaries, 477 and in this case, it is required that each marked packet can be 478 assigned to the right batch by each router. It is important to 479 mention that for the application of this method there are two 480 elements to consider: the clock error between network nodes and the 481 network delay. These can create offsets between the batches and out- 482 of-order of the packets. The mathematical formula on timing aspects, 483 explained in section 3.2 of [RFC8321], must be satisfied and it takes 484 into considerations the different causes of reordering such as clock 485 error and network delay. The assumption is to define the available 486 counting interval where to get stable counters and to avoid these 487 issues. Specifically, if the effects of network delay are ignored, 488 the condition to implement the methodology is that the clocks in 489 different nodes MUST be synchronized to the same clock reference with 490 an accuracy of +/- B/2 time units, where B is the fixed time duration 491 of the batch, which refers to the original marking interval at the 492 source node considering that this interval could fluctuate along the 493 path. In this way each marked packet can be assigned to the right 494 batch by each node. Usually the counters can be taken in the middle 495 of the batch period to be sure to take still counters. In a few 496 words this implies that the length of the batches MUST be chosen 497 large enough so that the method is not affected by those factors. 498 The length of the batches can be determined based on the specific 499 deployment scenario. 501 L bit=1 ----------+ +-----------+ +---------- 502 | | | | 503 L bit=0 +-----------+ +-----------+ 504 Batch n ... Batch 3 Batch 2 Batch 1 505 <---------> <---------> <---------> <---------> <---------> 507 Traffic Flow 508 ===========================================================> 509 L bit ...1111111111 0000000000 11111111111 00000000000 111111111... 510 ===========================================================> 512 Figure 1: Packet Loss Measurement and Single-Marking Methodology 513 using L bit 515 It is worth mentioning that the duration of the batches is considered 516 stable over time in the previous figure. In theory, it is possible 517 to change the length of batches over time and among different flows 518 for more flexibility. But, in practice, it could complicate the 519 correlation of the information. 521 5.2. Packet Delay Measurement 523 The same principle used to measure packet loss can be applied also to 524 one-way delay measurement. Delay metrics MAY be calculated using the 525 two possibilities: 527 1. Single-Marking Methodology: This approach uses only the L bit to 528 calculate both packet loss and delay. In this case, the D flag 529 MUST be set to zero on transmit and ignored by the monitoring 530 points. The alternation of the values of the L bit can be used 531 as a time reference to calculate the delay. Whenever the L bit 532 changes and a new batch starts, a network node can store the 533 timestamp of the first packet of the new batch, that timestamp 534 can be compared with the timestamp of the first packet of the 535 same batch on a second node to compute packet delay. But this 536 measurement is accurate only if no packet loss occurs and if 537 there is no packet reordering at the edges of the batches. A 538 different approach can also be considered and it is based on the 539 concept of the mean delay. The mean delay for each batch is 540 calculated by considering the average arrival time of the packets 541 for the relative batch. There are limitations also in this case 542 indeed, each node needs to collect all the timestamps and 543 calculate the average timestamp for each batch. In addition, the 544 information is limited to a mean value. 546 2. Double-Marking Methodology: This approach is more complete and 547 uses the L bit only to calculate packet loss and the D bit (Delay 548 flag) is fully dedicated to delay measurements. The idea is to 549 use the first marking with the L bit to create the alternate flow 550 and, within the batches identified by the L bit, a second marking 551 is used to select the packets for measuring delay. The D bit 552 creates a new set of marked packets that are fully identified 553 over the network, so that a network node can store the timestamps 554 of these packets; these timestamps can be compared with the 555 timestamps of the same packets on a second node to compute packet 556 delay values for each packet. The most efficient and robust mode 557 is to select a single double-marked packet for each batch, in 558 this way there is no time gap to consider between the double- 559 marked packets to avoid their reorder. Regarding the rule for 560 the selection of the packet to be double-marked, the same 561 considerations in Section 5.1 apply also here and the double- 562 marked packet can be chosen within the available counting 563 interval that is not affected by factors such as clock errors. 564 If a double-marked packet is lost, the delay measurement for the 565 considered batch is simply discarded, but this is not a big 566 problem because it is easy to recognize the problematic batch and 567 skip the measurement just for that one. So in order to have more 568 information about the delay and to overcome out-of-order issues 569 this method is preferred. 571 In summary the approach with double marking is better than the 572 approach with single marking. Moreover, the two approaches provide 573 slightly different pieces of information and the data consumer can 574 combine them to have a more robust data set. 576 Similar to what said in Section 5.1 for the packet counters, in the 577 implementation the timestamps can be sent out to the controller that 578 is responsible for the calculation or could also be exchanged using 579 other on-path techniques. But this is out of scope for this 580 document. 582 L bit=1 ----------+ +-----------+ +---------- 583 | | | | 584 L bit=0 +-----------+ +-----------+ 586 D bit=1 + + + + + 587 | | | | | 588 D bit=0 ------+----------+----------+----------+------------+----- 590 Traffic Flow 591 ===========================================================> 592 L bit ...1111111111 0000000000 11111111111 00000000000 111111111... 594 D bit ...0000010000 0000010000 00000100000 00001000000 000001000... 595 ===========================================================> 597 Figure 2: Double-Marking Methodology using L bit and D bit 599 Likewise to packet delay measurement (both for Single Marking and 600 Double Marking), the method can also be used to measure the inter- 601 arrival jitter. 603 5.3. Flow Monitoring Identification 605 The Flow Monitoring Identification (FlowMonID) identifies the flow to 606 be measured and is required for some general reasons: 608 First, it helps to reduce the per node configuration. Otherwise, 609 each node needs to configure an access-control list (ACL) for each 610 of the monitored flows. Moreover, using a flow identifier allows 611 a flexible granularity for the flow definition, indeed, it can be 612 used together with other identifiers (e.g. 5-tuple). 614 Second, it simplifies the counters handling. Hardware processing 615 of flow tuples (and ACL matching) is challenging and often incurs 616 into performance issues, especially in tunnel interfaces. 618 Third, it eases the data export encapsulation and correlation for 619 the collectors. 621 The FlowMonID MUST only be used as a monitored flow identifier in 622 order to determine a monitored flow within the measurement domain. 623 This entails not only an easy identification but improved correlation 624 as well. 626 The value of 20 bits has been selected for the FlowMonID since it is 627 a good compromise and implies a low rate of ambiguous FlowMonIDs that 628 can be considered acceptable in most of the applications. The 629 disambiguation issue is more visible when the FlowMonID is pseudo 630 randomly generated but, it can be solved by tagging the FlowMonID 631 with additional flow information. In particular, it is RECOMMENDED 632 to consider the 3-tuple FlowMonID, source and destination addresses: 634 o If the 20 bit FlowMonID is set independently and pseudo randomly 635 in a distributed way there is a chance of collision. Indeed, by 636 using the well-known birthday problem in probability theory, if 637 the 20 bit FlowMonID is set independently and pseudo randomly 638 without any additional input entropy, there is a 50% chance of 639 collision for 1206 flows. So, for more entropy, FlowMonID is 640 combined with source and destination addresses. Since there is a 641 1% chance of collision for 145 flows, it is possible to monitor 642 145 concurrent flows per host pairs with a 1% chance of collision. 644 o If the 20 bits FlowMonID is set in a centralized way, the 645 controller can instruct the nodes properly in order to guarantee 646 the uniqueness of the FlowMonID. With 20 bits, the number of 647 combinations is 1048576, and the controller should ensure that all 648 the FlowMonID values are used without any collision. Therefore, 649 by considering source and destination addresses together with the 650 FlowMonID, it can be possible to monitor 1048576 concurrent flows 651 per host pairs. 653 A consistent approach MUST be used in the Alternate Marking 654 deployment to avoid the mixture of different ways of identifying. 655 All the nodes along the path and involved into the measurement SHOULD 656 use the same mode for identification. As mentioned, it is 657 RECOMMENDED to use the FlowMonID for identification purpose in 658 combination with source and destination addresses to identify a flow. 659 By considering source and destination addresses together with the 660 FlowMonID it can be possible to monitor 145 concurrent flows per host 661 pairs with a 1% chance of collision in case of pseudo randomly 662 generated FlowMonID, or 1048576 concurrent flows per host pairs in 663 case of centralized controller. It is worth mentioning that the 664 solution with the centralized control allows finer granularity and 665 therefore adds even more flexibility to the flow identification. 667 The FlowMonID field is set at the source node, which is the ingress 668 point of the measurement domain, and can be set in two ways: 670 a. It can be algorithmically generated by the source node, that can 671 set it pseudo-randomly with some chance of collision. This 672 approach cannot guarantee the uniqueness of FlowMonID but, 673 considering the recommendation to use FlowMonID with source and 674 destination addresses the conflict probability is reduced due to 675 the FlowMonID space available for each endpoint pair (i.e. 145 676 flows with 1% chance of collision). 678 b. It can be assigned by the central controller. Since the 679 controller knows the network topology, it can set the value 680 properly to avoid or minimize ambiguity and guarantee the 681 uniqueness. In this regard, the controller can simply verify 682 that there is no ambiguity between different pseudo-randomly 683 generated FlowMonIDs on the same path. The conflict probability 684 is really small given that the FlowMonID is coupled with source 685 and destination addresses and up to 1048576 flows can be 686 monitored for each endpoint pair. 688 If the FlowMonID is set by the source node, the intermediate nodes 689 can read the FlowMonIDs from the packets in flight and act 690 accordingly. While, if the FlowMonID is set by the controller, both 691 possibilities are feasible for the intermediate nodes which can learn 692 by reading the packets or can be instructed by the controller. 694 When all values in the FlowMonID space are consumed, the centralized 695 controller can keep track and reassign the values that are not used 696 any more by old flows, while if the FlowMonID is pseudo randomly 697 generated by the source, conflicts and collisions are possible. 699 5.4. Multipoint and Clustered Alternate Marking 701 The Alternate Marking method can also be extended to any kind of 702 multipoint to multipoint paths, and the network clustering approach 703 allows a flexible and optimized performance measurement, as described 704 in [RFC8889]. 706 The Cluster is the smallest identifiable subnetwork of the entire 707 Network graph that still satisfies the condition that the number of 708 packets that goes in is the same that goes out. With network 709 clustering, it is possible to use the partition of the network into 710 clusters at different levels in order to perform the needed degree of 711 detail. So, for Multipoint Alternate Marking, FlowMonID can identify 712 in general a multipoint-to-multipoint flow and not only a point-to- 713 point flow. 715 5.5. Data Collection and Calculation 717 The nodes enabled to perform performance monitoring collect the value 718 of the packet counters and timestamps. There are several 719 alternatives to implement Data Collection and Calculation, but this 720 is not specified in this document. 722 There are documents on the control plane mechanisms of Alternate 723 Marking, e.g. [I-D.ietf-idr-sr-policy-ifit], 724 [I-D.chen-pce-pcep-ifit]. 726 6. Security Considerations 728 This document aims to apply a method to perform measurements that 729 does not directly affect Internet security nor applications that run 730 on the Internet. However, implementation of this method must be 731 mindful of security and privacy concerns. 733 There are two types of security concerns: potential harm caused by 734 the measurements and potential harm to the measurements. 736 Harm caused by the measurement: Alternate Marking implies 737 modifications on the fly to an Option Header of IPv6 packets by the 738 source node, but this must be performed in a way that does not alter 739 the quality of service experienced by the packets and that preserves 740 stability and performance of routers doing the measurements. As 741 already discussed in Section 4, it is RECOMMENDED that the AltMark 742 Option does not affect the throughput and therefore the user 743 experience. 745 Harm to the measurement: Alternate Marking measurements could be 746 harmed by routers altering the fields of the AltMark Option (e.g. 747 marking of the packets, FlowMonID) or by a malicious attacker adding 748 AltMark Option to the packets in order to consume the resources of 749 network devices and entities involved. As described above, the 750 source node is the only one that writes the Option Header while the 751 intermediate nodes and destination node only read it without 752 modifying the Option Header. But, for example, an on-path attacker 753 can modify the flags, whether intentionally or accidentally, or 754 deliberately insert a new option to the packet flow or delete the 755 option from the packet flow. The consequent effect could be to give 756 the appearance of loss or delay or invalidate the measurement by 757 modifying option identifiers, such as FlowMonID. The malicious 758 implication can be to cause actions from the network administrator 759 where an intervention is not necessary or to hide real issues in the 760 network. Since the measurement itself may be affected by network 761 nodes intentionally altering the bits of the AltMark Option or 762 injecting Options headers as a means for Denial of Service (DoS), the 763 Alternate Marking MUST be applied in the context of a controlled 764 domain, where the network nodes are locally administered and this 765 type of attack can be avoided. For this reason, the implementation 766 of the method is not done on the end node if it is not fully managed 767 and does not belong to the controlled domain. Packets generated 768 outside the controlled domain may consume router resources by 769 maliciously using the HbH Option, but this can be mitigated by 770 filtering these packets at the controlled domain boundary. This can 771 be done because, if the end node does not belong to the controlled 772 domain, it is not supposed to add the AltMark HbH Option, and it can 773 be easily recognized. 775 The flow identifier (FlowMonID) composes the AltMark Option together 776 with the two marking bits (L and D). As explained in Section 5.3, 777 there is a chance of collision if the FlowMonID is set pseudo 778 randomly and a solution exists. In general this may not be a problem 779 and a low rate of ambiguous FlowMonIDs can be acceptable, since this 780 does not cause significant harm to the operators or their clients and 781 this harm may not justify the complications of avoiding it. But, for 782 large scale measurements, a big number of flows could be monitored 783 and the probability of a collision is higher, thus the disambiguation 784 of the FlowMonID field can be considered. 786 The privacy concerns also need to be analyzed even if the method only 787 relies on information contained in the Option Header without any 788 release of user data. Indeed, from a confidentiality perspective, 789 although AltMark Option does not contain user data, the metadata can 790 be used for network reconnaissance to compromise the privacy of users 791 by allowing attackers to collect information about network 792 performance and network paths. AltMark Option contains two kinds of 793 metadata: the marking bits (L and D bits) and the flow identifier 794 (FlowMonID). 796 The marking bits are the small information that is exchanged 797 between the network nodes. Therefore, due to this intrinsic 798 characteristic, network reconnaissance through passive 799 eavesdropping on data-plane traffic is difficult. Indeed, an 800 attacker cannot gain information about network performance from a 801 single monitoring point. The only way for an attacker can be to 802 eavesdrop on multiple monitoring points at the same time, because 803 they have to do the same kind of calculation and aggregation as 804 Alternate Marking requires. 806 The FlowMonID field is used in the AltMark Option as the 807 identifier of the monitored flow. It represents a more sensitive 808 information for network reconnaissance and may allow a flow 809 tracking type of attack because an attacker could collect 810 information about network paths. 812 Furthermore, in a pervasive surveillance attack, the information that 813 can be derived over time is more. But, as further described 814 hereinafter, the application of the Alternate Marking to a controlled 815 domain helps to mitigate all the above aspects of privacy concerns. 817 At the management plane, attacks can be set up by misconfiguring or 818 by maliciously configuring AltMark Option. Thus, AltMark Option 819 configuration MUST be secured in a way that authenticates authorized 820 users and verifies the integrity of configuration procedures. 821 Solutions to ensure the integrity of AltMark Option are outside the 822 scope of this document. Also, attacks on the reporting of the 823 statistics between the monitoring points and the network management 824 system (e.g. centralized controller) can interfere with the proper 825 functioning of the system. Hence, the channels used to report back 826 flow statistics MUST be secured. 828 As stated above, the precondition for the application of the 829 Alternate Marking is that it MUST be applied in specific controlled 830 domains, thus confining the potential attack vectors within the 831 network domain. [RFC8799] analyzes and discusses the trend towards 832 network behaviors that can be applied only within a limited domain. 833 This is due to the specific set of requirements especially related to 834 security, network management, policies and options supported which 835 may vary between such limited domains. A limited administrative 836 domain provides the network administrator with the means to select, 837 monitor and control the access to the network, making it a trusted 838 domain. In this regard it is expected to enforce policies at the 839 domain boundaries to filter both external packets with AltMark Option 840 entering the domain and internal packets with AltMark Option leaving 841 the domain. Therefore, the trusted domain is unlikely subject to 842 hijacking of packets since packets with AltMark Option are processed 843 and used only within the controlled domain. 845 As stated, the application to a controlled domain ensures the control 846 over the packets entering and leaving the domain, but despite that, 847 leakages may happen for different reasons, such as a failure or a 848 fault. In this case, nodes outside the domain MUST simply ignore 849 packets with AltMark Option since they should not process it. 851 Additionally, it is to be noted that the AltMark Option is carried by 852 the Options Header and it may have some impact on the packet sizes 853 for the monitored flow and on the path MTU, since some packets might 854 exceed the MTU. However, the relative small size (48 bit in total) 855 of these Option Headers and its application to a controlled domain 856 help to mitigate the problem. 858 It is worth mentioning that the security concerns may change based on 859 the specific deployment scenario and related threat analysis, which 860 can lead to specific security solutions that are beyond the scope of 861 this document. As an example, the AltMark Option can be used as Hop- 862 by-Hop or Destination Option and, in case of Destination Option, 863 multiple administrative domains may be traversed by the AltMark 864 Option that is not confined to a single administrative domain. In 865 this case, the user, aware of the kind of risks, may still want to 866 use Alternate Marking for telemetry and test purposes but the 867 controlled domain must be composed by more than one administrative 868 domains. To this end, the inter-domain links need to be secured 869 (e.g., by IPsec, VPNs) in order to avoid external threats and realize 870 the whole controlled domain. 872 It might be theoretically possible to modulate the marking or the 873 other fields of the AltMark Option to serve as a covert channel to be 874 used by an on-path observer. This may affect both the data and 875 management plane, but, here too, the application to a controlled 876 domain helps to reduce the effects. 878 The Alternate Marking application described in this document relies 879 on a time synchronization protocol. Thus, by attacking the time 880 protocol, an attacker can potentially compromise the integrity of the 881 measurement. A detailed discussion about the threats against time 882 protocols and how to mitigate them is presented in [RFC7384]. 883 Network Time Security (NTS), described in [RFC8915], is a mechanism 884 that can be employed. Also, the time, which is distributed to the 885 network nodes through the time protocol, is centrally taken from an 886 external accurate time source, such as an atomic clock or a GPS 887 clock. By attacking the time source it can be possible to compromise 888 the integrity of the measurement as well. There are security 889 measures that can be taken to mitigate the GPS spoofing attacks and a 890 network administrator should certainly employ solutions to secure the 891 network domain. 893 7. IANA Considerations 895 The Option Type should be assigned in IANA's "Destination Options and 896 Hop-by-Hop Options" registry. 898 This draft requests the following IPv6 Option Type assignment from 899 the Destination Options and Hop-by-Hop Options sub-registry of 900 Internet Protocol Version 6 (IPv6) Parameters 901 (https://www.iana.org/assignments/ipv6-parameters/). 903 Hex Value Binary Value Description Reference 904 act chg rest 905 ---------------------------------------------------------------- 906 TBD 00 0 tbd AltMark [This draft] 908 8. Acknowledgements 910 The authors would like to thank Bob Hinden, Ole Troan, Martin Duke, 911 Lars Eggert, Roman Danyliw, Alvaro Retana, Eric Vyncke, Warren 912 Kumari, Benjamin Kaduk, Stewart Bryant, Christopher Wood, Yoshifumi 913 Nishida, Tom Herbert, Stefano Previdi, Brian Carpenter, Greg Mirsky, 914 Ron Bonica for the precious comments and suggestions. 916 9. References 918 9.1. Normative References 920 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 921 Requirement Levels", BCP 14, RFC 2119, 922 DOI 10.17487/RFC2119, March 1997, 923 . 925 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 926 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 927 May 2017, . 929 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 930 (IPv6) Specification", STD 86, RFC 8200, 931 DOI 10.17487/RFC8200, July 2017, 932 . 934 9.2. Informative References 936 [I-D.chen-pce-pcep-ifit] 937 Yuan, H., Zhou, T., Li, W., Fioccola, G., and Y. Wang, 938 "Path Computation Element Communication Protocol (PCEP) 939 Extensions to Enable IFIT", draft-chen-pce-pcep-ifit-04 940 (work in progress), July 2021. 942 [I-D.fz-spring-srv6-alt-mark] 943 Fioccola, G., Zhou, T., and M. Cociglio, "Segment Routing 944 Header encapsulation for Alternate Marking Method", draft- 945 fz-spring-srv6-alt-mark-01 (work in progress), July 2021. 947 [I-D.hinden-6man-hbh-processing] 948 Hinden, R. M. and G. Fairhurst, "IPv6 Hop-by-Hop Options 949 Processing Procedures", draft-hinden-6man-hbh- 950 processing-01 (work in progress), June 2021. 952 [I-D.ietf-idr-sr-policy-ifit] 953 Qin, F., Yuan, H., Zhou, T., Fioccola, G., and Y. Wang, 954 "BGP SR Policy Extensions to Enable IFIT", draft-ietf-idr- 955 sr-policy-ifit-02 (work in progress), July 2021. 957 [I-D.peng-v6ops-hbh] 958 Peng, S., Li, Z., Xie, C., Qin, Z., and G. Mishra, 959 "Processing of the Hop-by-Hop Options Header", draft-peng- 960 v6ops-hbh-06 (work in progress), August 2021. 962 [RFC4291] Hinden, R. and S. Deering, "IP Version 6 Addressing 963 Architecture", RFC 4291, DOI 10.17487/RFC4291, February 964 2006, . 966 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 967 Internet Protocol", RFC 4301, DOI 10.17487/RFC4301, 968 December 2005, . 970 [RFC6437] Amante, S., Carpenter, B., Jiang, S., and J. Rajahalme, 971 "IPv6 Flow Label Specification", RFC 6437, 972 DOI 10.17487/RFC6437, November 2011, 973 . 975 [RFC6438] Carpenter, B. and S. Amante, "Using the IPv6 Flow Label 976 for Equal Cost Multipath Routing and Link Aggregation in 977 Tunnels", RFC 6438, DOI 10.17487/RFC6438, November 2011, 978 . 980 [RFC7045] Carpenter, B. and S. Jiang, "Transmission and Processing 981 of IPv6 Extension Headers", RFC 7045, 982 DOI 10.17487/RFC7045, December 2013, 983 . 985 [RFC7384] Mizrahi, T., "Security Requirements of Time Protocols in 986 Packet Switched Networks", RFC 7384, DOI 10.17487/RFC7384, 987 October 2014, . 989 [RFC8321] Fioccola, G., Ed., Capello, A., Cociglio, M., Castaldelli, 990 L., Chen, M., Zheng, L., Mirsky, G., and T. Mizrahi, 991 "Alternate-Marking Method for Passive and Hybrid 992 Performance Monitoring", RFC 8321, DOI 10.17487/RFC8321, 993 January 2018, . 995 [RFC8754] Filsfils, C., Ed., Dukes, D., Ed., Previdi, S., Leddy, J., 996 Matsushima, S., and D. Voyer, "IPv6 Segment Routing Header 997 (SRH)", RFC 8754, DOI 10.17487/RFC8754, March 2020, 998 . 1000 [RFC8799] Carpenter, B. and B. Liu, "Limited Domains and Internet 1001 Protocols", RFC 8799, DOI 10.17487/RFC8799, July 2020, 1002 . 1004 [RFC8889] Fioccola, G., Ed., Cociglio, M., Sapio, A., and R. Sisto, 1005 "Multipoint Alternate-Marking Method for Passive and 1006 Hybrid Performance Monitoring", RFC 8889, 1007 DOI 10.17487/RFC8889, August 2020, 1008 . 1010 [RFC8915] Franke, D., Sibold, D., Teichel, K., Dansarie, M., and R. 1011 Sundblad, "Network Time Security for the Network Time 1012 Protocol", RFC 8915, DOI 10.17487/RFC8915, September 2020, 1013 . 1015 Authors' Addresses 1017 Giuseppe Fioccola 1018 Huawei 1019 Riesstrasse, 25 1020 Munich 80992 1021 Germany 1023 Email: giuseppe.fioccola@huawei.com 1025 Tianran Zhou 1026 Huawei 1027 156 Beiqing Rd. 1028 Beijing 100095 1029 China 1031 Email: zhoutianran@huawei.com 1033 Mauro Cociglio 1034 Telecom Italia 1035 Via Reiss Romoli, 274 1036 Torino 10148 1037 Italy 1039 Email: mauro.cociglio@telecomitalia.it 1041 Fengwei Qin 1042 China Mobile 1043 32 Xuanwumenxi Ave. 1044 Beijing 100032 1045 China 1047 Email: qinfengwei@chinamobile.com 1048 Ran Pang 1049 China Unicom 1050 9 Shouti South Rd. 1051 Beijing 100089 1052 China 1054 Email: pangran@chinaunicom.cn