idnits 2.17.00 (12 Aug 2021) /tmp/idnits21332/draft-fioccola-rfc8321bis-01.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 : ---------------------------------------------------------------------------- ** The abstract seems to contain references ([RFC8321]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (December 9, 2021) is 163 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) -- Possible downref: Non-RFC (?) normative reference: ref. 'IEEE-1588' == Outdated reference: A later version (-04) exists of draft-fioccola-rfc8889bis-00 == Outdated reference: A later version (-09) exists of draft-zhou-ippm-enhanced-alternate-marking-07 Summary: 1 error (**), 0 flaws (~~), 3 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group G. Fioccola, Ed. 3 Internet-Draft Huawei Technologies 4 Obsoletes: 8321 (if approved) M. Cociglio 5 Intended status: Standards Track Telecom Italia 6 Expires: June 12, 2022 G. Mirsky 7 Ericsson 8 T. Mizrahi 9 T. Zhou 10 Huawei Technologies 11 December 9, 2021 13 Alternate-Marking Method 14 draft-fioccola-rfc8321bis-01 16 Abstract 18 This document describes the Alternate-Marking technique to perform 19 packet loss, delay, and jitter measurements on live traffic. This 20 technology can be applied in various situations and for different 21 protocols. It could be considered Passive or Hybrid depending on the 22 application. This document obsoletes [RFC8321]. 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 June 12, 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. Requirements Language . . . . . . . . . . . . . . . . . . 4 60 2. Overview of the Method . . . . . . . . . . . . . . . . . . . 4 61 3. Detailed Description of the Method . . . . . . . . . . . . . 5 62 3.1. Packet Loss Measurement . . . . . . . . . . . . . . . . . 5 63 3.1.1. Coloring the Packets . . . . . . . . . . . . . . . . 10 64 3.1.2. Counting the Packets . . . . . . . . . . . . . . . . 10 65 3.1.3. Collecting Data and Calculating Packet Loss . . . . . 11 66 3.2. Timing Aspects . . . . . . . . . . . . . . . . . . . . . 12 67 3.3. One-Way Delay Measurement . . . . . . . . . . . . . . . . 13 68 3.3.1. Single-Marking Methodology . . . . . . . . . . . . . 14 69 3.3.2. Double-Marking Methodology . . . . . . . . . . . . . 16 70 3.4. Delay Variation Measurement . . . . . . . . . . . . . . . 17 71 4. Considerations . . . . . . . . . . . . . . . . . . . . . . . 17 72 4.1. Synchronization . . . . . . . . . . . . . . . . . . . . . 17 73 4.2. Data Correlation . . . . . . . . . . . . . . . . . . . . 18 74 4.3. Packet Reordering . . . . . . . . . . . . . . . . . . . . 19 75 4.4. Packet Fragmentation . . . . . . . . . . . . . . . . . . 20 76 5. Results of the Alternate Marking Experiment . . . . . . . . . 20 77 5.1. Controlled Domain requirement . . . . . . . . . . . . . . 22 78 6. Compliance with Guidelines from RFC 6390 . . . . . . . . . . 22 79 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 24 80 8. Security Considerations . . . . . . . . . . . . . . . . . . . 24 81 9. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 25 82 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 26 83 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 26 84 11.1. Normative References . . . . . . . . . . . . . . . . . . 26 85 11.2. Informative References . . . . . . . . . . . . . . . . . 26 86 Appendix A. Changes Log . . . . . . . . . . . . . . . . . . . . 28 87 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 29 89 1. Introduction 91 Nowadays, most Service Providers' networks carry traffic with 92 contents that are highly sensitive to packet loss [RFC7680], delay 93 [RFC7679], and jitter [RFC3393]. 95 In view of this scenario, Service Providers need methodologies and 96 tools to monitor and measure network performance with an adequate 97 accuracy, in order to constantly control the quality of experience 98 perceived by their customers. Performance monitoring also provides 99 useful information for improving network management (e.g., isolation 100 of network problems, troubleshooting, etc.). 102 A lot of work related to Operations, Administration, and Maintenance 103 (OAM), which also includes performance monitoring techniques, has 104 been done by Standards Developing Organizations (SDOs): [RFC7276] 105 provides a good overview of existing OAM mechanisms defined in the 106 IETF, ITU-T, and IEEE. In the IETF, a lot of work has been done on 107 fault detection and connectivity verification, while a minor effort 108 has been thus far dedicated to performance monitoring. The IPPM WG 109 has defined standard metrics to measure network performance; however, 110 the methods developed in this WG mainly refer to focus on Active 111 measurement techniques. More recently, the MPLS WG has defined 112 mechanisms for measuring packet loss, one-way and two-way delay, and 113 delay variation in MPLS networks [RFC6374], but their applicability 114 to Passive measurements has some limitations, especially for pure 115 connection-less networks. 117 The lack of adequate tools to measure packet loss with the desired 118 accuracy drove an effort to design a new method for the performance 119 monitoring of live traffic, which is easy to implement and deploy. 120 The effort led to the method described in this document: basically, 121 it is a Passive performance monitoring technique, potentially 122 applicable to any kind of packet-based traffic, including Ethernet, 123 IP, and MPLS, both unicast and multicast. The method addresses 124 primarily packet loss measurement, but it can be easily extended to 125 one-way or two-way delay and delay variation measurements as well. 127 The method has been explicitly designed for Passive measurements, but 128 it can also be used with Active probes. Passive measurements are 129 usually more easily understood by customers and provide much better 130 accuracy, especially for packet loss measurements. 132 RFC 7799 [RFC7799] defines Passive and Hybrid Methods of Measurement. 133 In particular, Passive Methods of Measurement are based solely on 134 observations of an undisturbed and unmodified packet stream of 135 interest; Hybrid Methods are Methods of Measurement that use a 136 combination of Active Methods and Passive Methods. 138 Taking into consideration these definitions, the Alternate-Marking 139 Method could be considered Hybrid or Passive, depending on the case. 140 In the case where the marking method is obtained by changing existing 141 field values of the packets the technique is Hybrid. In the case 142 where the marking field is dedicated, reserved, and included in the 143 protocol specification, the Alternate-Marking technique can be 144 considered as Passive. 146 1.1. Requirements Language 148 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 149 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 150 "OPTIONAL" in this document are to be interpreted as described in BCP 151 14 [RFC2119] [RFC8174] when, and only when, they appear in all 152 capitals, as shown here. 154 2. Overview of the Method 156 In order to perform packet loss measurements on a production traffic 157 flow, different approaches exist. The most intuitive one consists in 158 numbering the packets so that each router that receives the flow can 159 immediately detect a packet that is missing. This approach, though 160 very simple in theory, is not simple to achieve: it requires the 161 insertion of a sequence number into each packet, and the devices must 162 be able to extract the number and check it in real time. Such a task 163 can be difficult to implement on live traffic: if UDP is used as the 164 transport protocol, the sequence number is not available; on the 165 other hand, if a higher-layer sequence number (e.g., in the RTP 166 header) is used, extracting that information from each packet and 167 processing it in real time could overload the device. 169 An alternate approach is to count the number of packets sent on one 170 end, count the number of packets received on the other end, and 171 compare the two values. This operation is much simpler to implement, 172 but it requires the devices performing the measurement to be in sync: 173 in order to compare two counters, it is required that they refer 174 exactly to the same set of packets. Since a flow is continuous and 175 cannot be stopped when a counter has to be read, it can be difficult 176 to determine exactly when to read the counter. A possible solution 177 to overcome this problem is to virtually split the flow in 178 consecutive blocks by periodically inserting a delimiter so that each 179 counter refers exactly to the same block of packets. The delimiter 180 could be, for example, a special packet inserted artificially into 181 the flow. However, delimiting the flow using specific packets has 182 some limitations. First, it requires generating additional packets 183 within the flow and requires the equipment to be able to process 184 those packets. In addition, the method is vulnerable to out-of-order 185 reception of delimiting packets and, to a lesser extent, to their 186 loss. 188 The method proposed in this document follows the second approach, but 189 it doesn't use additional packets to virtually split the flow in 190 blocks. Instead, it "marks" the packets so that the packets 191 belonging to the same block will have the same color, whilst 192 consecutive blocks will have different colors. Each change of color 193 represents a sort of auto-synchronization signal that guarantees the 194 consistency of measurements taken by different devices along the 195 path. 197 Figure 1 represents a very simple network and shows how the method 198 can be used to measure packet loss on different network segments: by 199 enabling the measurement on several interfaces along the path, it is 200 possible to perform link monitoring, node monitoring, or end-to-end 201 monitoring. The method is flexible enough to measure packet loss on 202 any segment of the network and can be used to isolate the faulty 203 element. 205 Traffic Flow 206 ========================================================> 207 +------+ +------+ +------+ +------+ 208 ---<> R1 <>-----<> R2 <>-----<> R3 <>-----<> R4 <>--- 209 +------+ +------+ +------+ +------+ 210 . . . . . . 211 . . . . . . 212 . <------> <-------> . 213 . Node Packet Loss Link Packet Loss . 214 . . 215 <---------------------------------------------------> 216 End-to-End Packet Loss 218 Figure 1: Available Measurements 220 3. Detailed Description of the Method 222 This section describes, in detail, how the method operates. A 223 special emphasis is given to the measurement of packet loss, which 224 represents the core application of the method, but applicability to 225 delay and jitter measurements is also considered. 227 3.1. Packet Loss Measurement 229 The basic idea is to virtually split traffic flows into consecutive 230 blocks: each block represents a measurable entity unambiguously 231 recognizable by all network devices along the path. By counting the 232 number of packets in each block and comparing the values measured by 233 different network devices along the path, it is possible to measure 234 if packet loss occurred in any single block between any two points. 236 As discussed in the previous section, a simple way to create the 237 blocks is to "color" the traffic (two colors are sufficient), so that 238 packets belonging to different consecutive blocks will have different 239 colors. Whenever the color changes, the previous block terminates 240 and the new one begins. Hence, all the packets belonging to the same 241 block will have the same color and packets of different consecutive 242 blocks will have different colors. The number of packets in each 243 block depends on the criterion used to create the blocks: 245 o if the color is switched after a fixed number of packets, then 246 each block will contain the same number of packets (except for any 247 losses); and 249 o if the color is switched according to a fixed timer, then the 250 number of packets may be different in each block depending on the 251 packet rate. 253 The rest of the document assumes that the blocks are created 254 according to a fixed timer. The switching after a fixed number of 255 packets is an additional possibility but its detailed specification 256 is out of scope. 258 The following figure shows how a flow looks like when it is split in 259 traffic blocks with colored packets. 261 A: packet with A coloring 262 B: packet with B coloring 264 | | | | | 265 | | Traffic Flow | | 266 -------------------------------------------------------------------> 267 BBBBBBB AAAAAAAAAAA BBBBBBBBBBB AAAAAAAAAAA BBBBBBBBBBB AAAAAAA 268 -------------------------------------------------------------------> 269 ... | Block 5 | Block 4 | Block 3 | Block 2 | Block 1 270 | | | | | 272 Figure 2: Traffic Coloring 274 Figure 3 shows how the method can be used to measure link packet loss 275 between two adjacent nodes. 277 Referring to the figure, let's assume we want to monitor the packet 278 loss on the link between two routers: router R1 and router R2. 279 According to the method, the traffic is colored alternatively with 280 two different colors: A and B. Whenever the color changes, the 281 transition generates a sort of square-wave signal, as depicted in the 282 following figure. 284 Color A ----------+ +-----------+ +---------- 285 | | | | 286 Color B +-----------+ +-----------+ 287 Block n ... Block 3 Block 2 Block 1 288 <---------> <---------> <---------> <---------> <---------> 290 Traffic Flow 291 ===========================================================> 292 Color ...AAAAAAAAAAA BBBBBBBBBBB AAAAAAAAAAA BBBBBBBBBBB AAAAAAA... 293 ===========================================================> 295 Figure 3: Computation of Link Packet Loss 297 Traffic coloring can be done by R1 itself if the traffic is not 298 already colored. R1 needs two counters, C(A)R1 and C(B)R1, on its 299 egress interface: C(A)R1 counts the packets with color A and C(B)R1 300 counts those with color B. As long as traffic is colored as A, only 301 counter C(A)R1 will be incremented, while C(B)R1 is not incremented; 302 conversely, when the traffic is colored as B, only C(B)R1 is 303 incremented. C(A)R1 and C(B)R1 can be used as reference values to 304 determine the packet loss from R1 to any other measurement point down 305 the path. Router R2, similarly, will need two counters on its 306 ingress interface, C(A)R2 and C(B)R2, to count the packets received 307 on that interface and colored with A and B, respectively. When an A 308 block ends, it is possible to compare C(A)R1 and C(A)R2 and calculate 309 the packet loss within the block; similarly, when the successive B 310 block terminates, it is possible to compare C(B)R1 with C(B)R2, and 311 so on, for every successive block. 313 Likewise, by using two counters on the R2 egress interface, it is 314 possible to count the packets sent out of the R2 interface and use 315 them as reference values to calculate the packet loss from R2 to any 316 measurement point down R2. 318 Using a fixed timer for color switching offers better control over 319 the method: the (time) length of the blocks can be chosen large 320 enough to simplify the collection and the comparison of measures 321 taken by different network devices. It's preferable to read the 322 value of the counters not immediately after the color switch: some 323 packets could arrive out of order and increment the counter 324 associated with the previous block (color), so it is worth waiting 325 for some time. A safe choice is to wait L/2 time units (where L is 326 the duration for each block) after the color switch, to read the 327 still counter of the previous color, so the possibility of reading a 328 running counter instead of a still one is minimized. The drawback is 329 that the longer the duration of the block, the less frequent the 330 measurement can be taken. 332 The following table shows how the counters can be used to calculate 333 the packet loss between R1 and R2. The first column lists the 334 sequence of traffic blocks, while the other columns contain the 335 counters of A-colored packets and B-colored packets for R1 and R2. 336 In this example, we assume that the values of the counters are reset 337 to zero whenever a block ends and its associated counter has been 338 read: with this assumption, the table shows only relative values, 339 which is the exact number of packets of each color within each block. 340 If the values of the counters were not reset, the table would contain 341 cumulative values, but the relative values could be determined simply 342 by the difference from the value of the previous block of the same 343 color. 345 The color is switched on the basis of a fixed timer (not shown in the 346 table), so the number of packets in each block is different. 348 +-------+--------+--------+--------+--------+------+ 349 | Block | C(A)R1 | C(B)R1 | C(A)R2 | C(B)R2 | Loss | 350 +-------+--------+--------+--------+--------+------+ 351 | 1 | 375 | 0 | 375 | 0 | 0 | 352 | 2 | 0 | 388 | 0 | 388 | 0 | 353 | 3 | 382 | 0 | 381 | 0 | 1 | 354 | 4 | 0 | 377 | 0 | 374 | 3 | 355 | ... | ... | ... | ... | ... | ... | 356 | 2n | 0 | 387 | 0 | 387 | 0 | 357 | 2n+1 | 379 | 0 | 377 | 0 | 2 | 358 +-------+--------+--------+--------+--------+------+ 360 Figure 4: Evaluation of Counters for Packet Loss Measurements 362 During an A block (blocks 1, 3, and 2n+1), all the packets are 363 A-colored; therefore, the C(A) counters are incremented to the number 364 seen on the interface, while C(B) counters are zero. Conversely, 365 during a B block (blocks 2, 4, and 2n), all the packets are 366 B-colored: C(A) counters are zero, while C(B) counters are 367 incremented. 369 When a block ends (because of color switching), the relative counters 370 stop incrementing; it is possible to read them, compare the values 371 measured on routers R1 and R2, and calculate the packet loss within 372 that block. 374 For example, looking at the table above, during the first block 375 (A-colored), C(A)R1 and C(A)R2 have the same value (375), which 376 corresponds to the exact number of packets of the first block (no 377 loss). Also, during the second block (B-colored), R1 and R2 counters 378 have the same value (388), which corresponds to the number of packets 379 of the second block (no loss). During the third and fourth blocks, 380 R1 and R2 counters are different, meaning that some packets have been 381 lost: in the example, one single packet (382-381) was lost during 382 block three, and three packets (377-374) were lost during block four. 384 The method applied to R1 and R2 can be extended to any other router 385 and applied to more complex networks, as far as the measurement is 386 enabled on the path followed by the traffic flow(s) being observed. 388 It's worth mentioning two different strategies that can be used when 389 implementing the method: 391 o flow-based: the flow-based strategy is used when only a limited 392 number of traffic flows need to be monitored. According to this 393 strategy, only a subset of the flows is colored. Counters for 394 packet loss measurements can be instantiated for each single flow, 395 or for the set as a whole, depending on the desired granularity. 396 A relevant problem with this approach is the necessity to know in 397 advance the path followed by flows that are subject to 398 measurement. Path rerouting and traffic load-balancing increase 399 the issue complexity, especially for unicast traffic. The problem 400 is easier to solve for multicast traffic, where load-balancing is 401 seldom used and static joins are frequently used to force traffic 402 forwarding and replication. 404 o link-based: measurements are performed on all the traffic on a 405 link-by-link basis. The link could be a physical link or a 406 logical link. Counters could be instantiated for the traffic as a 407 whole or for each traffic class (in case it is desired to monitor 408 each class separately), but in the second case, two counters are 409 needed for each class. 411 As mentioned, the flow-based measurement requires the identification 412 of the flow to be monitored and the discovery of the path followed by 413 the selected flow. It is possible to monitor a single flow or 414 multiple flows grouped together, but in this case, measurement is 415 consistent only if all the flows in the group follow the same path. 416 Moreover, if a measurement is performed by grouping many flows, it is 417 not possible to determine exactly which flow was affected by packet 418 loss. In order to have measures per single flow, it is necessary to 419 configure counters for each specific flow. Once the flow(s) to be 420 monitored has been identified, it is necessary to configure the 421 monitoring on the proper nodes. Configuring the monitoring means 422 configuring the rule to intercept the traffic and configuring the 423 counters to count the packets. To have just an end-to-end 424 monitoring, it is sufficient to enable the monitoring on the first- 425 and last-hop routers of the path: the mechanism is completely 426 transparent to intermediate nodes and independent from the path 427 followed by traffic flows. On the contrary, to monitor the flow on a 428 hop-by-hop basis along its whole path, it is necessary to enable the 429 monitoring on every node from the source to the destination. In case 430 the exact path followed by the flow is not known a priori (i.e., the 431 flow has multiple paths to reach the destination), it is necessary to 432 enable the monitoring system on every path: counters on interfaces 433 traversed by the flow will report packet count, whereas counters on 434 other interfaces will be null. 436 3.1.1. Coloring the Packets 438 The coloring operation is fundamental in order to create packet 439 blocks. This implies choosing where to activate the coloring and how 440 to color the packets. 442 In case of flow-based measurements, the flow to monitor can be 443 defined by a set of selection rules (e.g., header fields) used to 444 match a subset of the packets; in this way, it is possible to control 445 the number of involved nodes, the path followed by the packets, and 446 the size of the flows. It is possible, in general, to have multiple 447 coloring nodes or a single coloring node that is easier to manage and 448 doesn't raise any risk of conflict. Coloring in multiple nodes can 449 be done, and the requirement is that the coloring must change 450 periodically between the nodes according to the timing considerations 451 in Section 3.2; so every node that is designated as a measurement 452 point along the path should be able to identify unambiguously the 453 colored packets. Furthermore, [I-D.fioccola-rfc8889bis] generalizes 454 the coloring for multipoint-to-multipoint flow. In addition, it can 455 be advantageous to color the flow as close as possible to the source 456 because it allows an end-to-end measure if a measurement point is 457 enabled on the last-hop router as well. 459 For link-based measurements, all traffic needs to be colored when 460 transmitted on the link. If the traffic had already been colored, 461 then it has to be re-colored because the color must be consistent on 462 the link. This means that each hop along the path must (re-)color 463 the traffic; the color is not required to be consistent along 464 different links. 466 Traffic coloring can be implemented by setting a specific bit in the 467 packet header and changing the value of that bit periodically. How 468 to choose the marking field depends on the application and is out of 469 scope here. 471 3.1.2. Counting the Packets 473 For flow-based measurements, assuming that the coloring of the 474 packets is performed only by the source nodes, the nodes between 475 source and destination (included) have to count the colored packets 476 that they receive and forward: this operation can be enabled on every 477 router along the path or only on a subset, depending on which network 478 segment is being monitored (a single link, a particular metro area, 479 the backbone, or the whole path). Since the color switches 480 periodically between two values, two counters (one for each value) 481 are needed: one counter for packets with color A and one counter for 482 packets with color B. For each flow (or group of flows) being 483 monitored and for every interface where the monitoring is Active, two 484 counters are needed. For example, in order to separately monitor 485 three flows on a router with four interfaces involved, 24 counters 486 are needed (two counters for each of the three flows on each of the 487 four interfaces). Furthermore, [I-D.fioccola-rfc8889bis] generalizes 488 the counting for multipoint-to-multipoint flow. 490 In case of link-based measurements, the behavior is similar except 491 that coloring and counting operations are performed on a link-by-link 492 basis at each endpoint of the link. 494 Another important aspect to take into consideration is when to read 495 the counters: in order to count the exact number of packets of a 496 block, the routers must perform this operation when that block has 497 ended; in other words, the counter for color A must be read when the 498 current block has color B, in order to be sure that the value of the 499 counter is stable. This task can be accomplished in two ways. The 500 general approach suggests reading the counters periodically, many 501 times during a block duration, and comparing these successive 502 readings: when the counter stops incrementing, it means that the 503 current block has ended, and its value can be elaborated safely. 504 Alternatively, if the coloring operation is performed on the basis of 505 a fixed timer, it is possible to configure the reading of the 506 counters according to that timer: for example, reading the counter 507 for color A every period in the middle of the subsequent block with 508 color B is a safe choice. A sufficient margin should be considered 509 between the end of a block and the reading of the counter, in order 510 to take into account any out-of-order packets. 512 3.1.3. Collecting Data and Calculating Packet Loss 514 The nodes enabled to perform performance monitoring collect the value 515 of the counters, but they are not able to directly use this 516 information to measure packet loss, because they only have their own 517 samples. For this reason, an external Network Management System 518 (NMS) can be used to collect and elaborate data and to perform packet 519 loss calculation. The NMS compares the values of counters from 520 different nodes and can calculate if some packets were lost (even a 521 single packet) and where those packets were lost. 523 The value of the counters needs to be transmitted to the NMS as soon 524 as it has been read. This can be accomplished by using SNMP or FTP 525 and can be done in Push Mode or Polling Mode. In the first case, 526 each router periodically sends the information to the NMS; in the 527 latter case, it is the NMS that periodically polls routers to collect 528 information. In any case, the NMS has to collect all the relevant 529 values from all the routers within one cycle of the timer. 531 It would also be possible to use a protocol to exchange values of 532 counters between the two endpoints in order to let them perform the 533 packet loss calculation for each traffic direction. 535 3.2. Timing Aspects 537 This document introduces two color-switching methods: one is based on 538 a fixed number of packets, and the other is based on a fixed timer. 539 But the method based on a fixed timer is preferable because it is 540 more deterministic, and it is considered in the document. 542 In general, clocks in network devices are not accurate and for this 543 reason, there is a clock error between the measurement points R1 and 544 R2. But, to implement the methodology, they must be synchronized to 545 the same clock reference with an accuracy of +/- L/2 time units, 546 where L is the fixed time duration of the block. So each colored 547 packet can be assigned to the right batch by each router. This is 548 because the minimum time distance between two packets of the same 549 color but that belong to different batches is L time units. 551 In practice, in addition to clock errors, the delay between 552 measurement points also affects the implementation of the methodology 553 because each packet can be delayed differently, and this can produce 554 out of order at batch boundaries. This means that, without 555 considering clock error, we wait L/2 after color switching to be sure 556 to take a still counter. 558 In summary, we need to take into account two contributions: clock 559 error between network devices and the interval we need to wait to 560 avoid packets being out of order because of network delay. 562 The following figure explains both issues. 564 ...BBBBBBBBB | AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA | BBBBBBBBB... 565 |<======================================>| 566 | L | 567 ...=========>|<==================><==================>|<==========... 568 | L/2 L/2 | 569 |<===>| |<===>| 570 d | | d 571 |<==========================>| 572 available counting interval 574 Figure 5: Timing Aspects 576 It is assumed that all network devices are synchronized to a common 577 reference time with an accuracy of +/- A/2. Thus, the difference 578 between the clock values of any two network devices is bounded by A. 580 The network delay between the network devices can be represented as a 581 data set and 99.7% of the samples are within 3 standard deviation of 582 the average. 584 The guard band d is given by: 586 d = A + D_avg + 3*D_stddev, 588 where A is the clock accuracy, D_avg is the average value of the 589 network delay between the network devices, and D_stddev is the 590 standard deviation of the delay. 592 The available counting interval is L - 2d that must be > 0. 594 The condition that must be satisfied and is a requirement on the 595 synchronization accuracy is: 597 d < L/2. 599 3.3. One-Way Delay Measurement 601 The same principle used to measure packet loss can be applied also to 602 one-way delay measurement. There are three alternatives, as 603 described hereinafter. 605 Note that, for all the one-way delay alternatives described in the 606 next sections, by summing the one-way delays of the two directions of 607 a path, it is always possible to measure the two-way delay (round- 608 trip "virtual" delay). 610 3.3.1. Single-Marking Methodology 612 The alternation of colors can be used as a time reference to 613 calculate the delay. Whenever the color changes (which means that a 614 new block has started), a network device can store the timestamp of 615 the first packet of the new block; that timestamp can be compared 616 with the timestamp of the same packet on a second router to compute 617 packet delay. When looking at Figure 2, R1 stores the timestamp 618 TS(A1)R1 when it sends the first packet of block 1 (A-colored), the 619 timestamp TS(B2)R1 when it sends the first packet of block 2 620 (B-colored), and so on for every other block. R2 performs the same 621 operation on the receiving side, recording TS(A1)R2, TS(B2)R2, and so 622 on. Since the timestamps refer to specific packets (the first packet 623 of each block), we are sure that timestamps compared to compute delay 624 refer to the same packets. By comparing TS(A1)R1 with TS(A1)R2 (and 625 similarly TS(B2)R1 with TS(B2)R2, and so on), it is possible to 626 measure the delay between R1 and R2. In order to have more 627 measurements, it is possible to take and store more timestamps, 628 referring to other packets within each block. 630 In order to coherently compare timestamps collected on different 631 routers, the clocks on the network nodes must be in sync. 632 Furthermore, a measurement is valid only if no packet loss occurs and 633 if packet misordering can be avoided; otherwise, the first packet of 634 a block on R1 could be different from the first packet of the same 635 block on R2 (for instance, if that packet is lost between R1 and R2 636 or it arrives after the next one). Since packet misordering is 637 generally undetectable it is not possible to check whether the first 638 packet on R1 is the same on R2 and this is part of the intrinsic 639 error in this measurement. 641 The following table shows how timestamps can be used to calculate the 642 delay between R1 and R2. The first column lists the sequence of 643 blocks, while other columns contain the timestamp referring to the 644 first packet of each block on R1 and R2. The delay is computed as a 645 difference between timestamps. For the sake of simplicity, all the 646 values are expressed in milliseconds. 648 +-------+---------+---------+---------+---------+-------------+ 649 | Block | TS(A)R1 | TS(B)R1 | TS(A)R2 | TS(B)R2 | Delay R1-R2 | 650 +-------+---------+---------+---------+---------+-------------+ 651 | 1 | 12.483 | - | 15.591 | - | 3.108 | 652 | 2 | - | 6.263 | - | 9.288 | 3.025 | 653 | 3 | 27.556 | - | 30.512 | - | 2.956 | 654 | 4 | - | 18.113 | - | 21.269 | 3.156 | 655 | ... | ... | ... | ... | ... | ... | 656 | 2n | 77.463 | - | 80.501 | - | 3.038 | 657 | 2n+1 | - | 24.333 | - | 27.433 | 3.100 | 658 +-------+---------+---------+---------+---------+-------------+ 660 Figure 6: Evaluation of Timestamps for Delay Measurements 662 The first row shows timestamps taken on R1 and R2, respectively, and 663 refers to the first packet of block 1 (which is A-colored). Delay 664 can be computed as a difference between the timestamp on R2 and the 665 timestamp on R1. Similarly, the second row shows timestamps (in 666 milliseconds) taken on R1 and R2 and refers to the first packet of 667 block 2 (which is B-colored). By comparing timestamps taken on 668 different nodes in the network and referring to the same packets 669 (identified using the alternation of colors), it is possible to 670 measure delay on different network segments. 672 For the sake of simplicity, in the above example, a single 673 measurement is provided within a block, taking into account only the 674 first packet of each block. The number of measurements can be easily 675 increased by considering multiple packets in the block: for instance, 676 a timestamp could be taken every N packets, thus generating multiple 677 delay measurements. Taking this to the limit, in principle, the 678 delay could be measured for each packet by taking and comparing the 679 corresponding timestamps (possible but impractical from an 680 implementation point of view). 682 3.3.1.1. Mean Delay 684 As mentioned before, the method previously exposed for measuring the 685 delay is sensitive to out-of-order reception of packets. In order to 686 overcome this problem, a different approach has been considered: it 687 is based on the concept of mean delay. The mean delay is calculated 688 by considering the average arrival time of the packets within a 689 single block. The network device locally stores a timestamp for each 690 packet received within a single block: summing all the timestamps and 691 dividing by the total number of packets received, the average arrival 692 time for that block of packets can be calculated. By subtracting the 693 average arrival times of two adjacent devices, it is possible to 694 calculate the mean delay between those nodes. When computing the 695 mean delay, the measurement error could be augmented by accumulating 696 the measurement error of a lot of packets. This method is robust to 697 out-of-order packets and also to packet loss (only a small error is 698 introduced). Moreover, it greatly reduces the number of timestamps 699 (only one per block for each network device) that have to be 700 collected by the management system. On the other hand, it only gives 701 one measure for the duration of the block, and it doesn't give the 702 minimum, maximum, and median delay values [RFC6703]. This limitation 703 could be overcome by reducing the duration of the block (for 704 instance, from 5 minutes to a few seconds), which implies a highly 705 optimized implementation of the method. 707 3.3.2. Double-Marking Methodology 709 The Single-Marking methodology for one-way delay measurement is 710 sensitive to out-of-order reception of packets. The first approach 711 to overcome this problem has been described before and is based on 712 the concept of mean delay. But the limitation of mean delay is that 713 it doesn't give information about the delay value's distribution for 714 the duration of the block. Additionally, it may be useful to have 715 not only the mean delay but also the minimum, maximum, and median 716 delay values and, in wider terms, to know more about the statistic 717 distribution of delay values. So, in order to have more information 718 about the delay and to overcome out-of-order issues, a different 719 approach can be introduced; it is based on a Double-Marking 720 methodology. 722 Basically, the idea is to use the first marking to create the 723 alternate flow and, within this colored flow, a second marking to 724 select the packets for measuring delay/jitter. The first marking is 725 needed for packet loss and mean delay measurement. The second 726 marking creates a new set of marked packets that are fully identified 727 over the network, so that a network device can store the timestamps 728 of these packets; these timestamps can be compared with the 729 timestamps of the same packets on a second router to compute packet 730 delay values for each packet. The number of measurements can be 731 easily increased by changing the frequency of the second marking. 732 But the frequency of the second marking must not be too high in order 733 to avoid out-of-order issues. Between packets with the second 734 marking, there should be a security time gap (e.g., this gap could 735 be, at the minimum, the mean network delay calculated with the 736 previous methodology) to avoid out-of-order issues and also to have a 737 number of measurement packets that are rate independent. If a 738 second-marking packet is lost, the delay measurement for the 739 considered block is corrupted and should be discarded. 741 Mean delay is calculated on all the packets of a sample and is a 742 simple computation to be performed for a Single-Marking Method. In 743 some cases, the mean delay measure is not sufficient to characterize 744 the sample, and more statistics of delay extent data are needed, 745 e.g., percentiles, variance, and median delay values. The 746 conventional range (maximum-minimum) should be avoided for several 747 reasons, including stability of the maximum delay due to the 748 influence by outliers. RFC 5481 [RFC5481], Section 6.5 highlights 749 how the 99.9th percentile of delay and delay variation is more 750 helpful to performance planners. To overcome this drawback, the idea 751 is to couple the mean delay measure for the entire batch with a 752 Double-Marking Method, where a subset of batch packets is selected 753 for extensive delay calculation by using a second marking. In this 754 way, it is possible to perform a detailed analysis on these double- 755 marked packets. Please note that there are classic algorithms for 756 median and variance calculation, but they are out of the scope of 757 this document. The comparison between the mean delay for the entire 758 batch and the mean delay on these double-marked packets gives useful 759 information since it is possible to understand if the Double-Marking 760 measurements are actually representative of the delay trends. 762 3.4. Delay Variation Measurement 764 Similar to one-way delay measurement (both for Single Marking and 765 Double Marking), the method can also be used to measure the inter- 766 arrival jitter. We refer to the definition in RFC 3393 [RFC3393]. 767 The alternation of colors, for a Single-Marking Method, can be used 768 as a time reference to measure delay variations. In case of Double 769 Marking, the time reference is given by the second-marked packets. 770 Considering the example depicted in Figure 2, R1 stores the timestamp 771 TS(A)R1 whenever it sends the first packet of a block, and R2 stores 772 the timestamp TS(B)R2 whenever it receives the first packet of a 773 block. The inter-arrival jitter can be easily derived from one-way 774 delay measurement, by evaluating the delay variation of consecutive 775 samples. 777 The concept of mean delay can also be applied to delay variation, by 778 evaluating the average variation of the interval between consecutive 779 packets of the flow from R1 to R2. 781 4. Considerations 783 This section highlights some considerations about the methodology. 785 4.1. Synchronization 787 The Alternate-Marking technique does not require a strong 788 synchronization, especially for packet loss and two-way delay 789 measurement. Only one-way delay measurement requires network devices 790 to have synchronized clocks. 792 Color switching is the reference for all the network devices, and the 793 only requirement to be achieved is that all network devices have to 794 recognize the right batch along the path. 796 Section 3.2 specifies the level of synchronization accuracy so that 797 all network devices consistently match the color bit to the correct 798 block. 800 This synchronization requirement can be satisfied even with a 801 relatively inaccurate synchronization method. This is true for 802 packet loss and two-way delay measurement, but not for one-way delay 803 measurement, where clock synchronization must be accurate. 805 Therefore, a system that uses only packet loss and two-way delay 806 measurement does not require synchronization. This is because the 807 value of the clocks of network devices does not affect the 808 computation of the two-way delay measurement. 810 4.2. Data Correlation 812 Data correlation is the mechanism to compare counters and timestamps 813 for packet loss, delay, and delay variation calculation. It could be 814 performed in several ways depending on the Alternate-Marking 815 application and use case. Some possibilities are to: 817 o use a centralized solution using NMS to correlate data; and 819 o define a protocol-based distributed solution by introducing a new 820 protocol or by extending the existing protocols (e.g., see RFC 821 6374 [RFC6374] or the Two-Way Active Measurement Protocol (TWAMP) 822 as defined in RFC 5357 [RFC5357] or the One-Way Active Measurement 823 Protocol (OWAMP) as defined in RFC 4656 [RFC4656]) in order to 824 communicate the counters and timestamps between nodes. 826 In the following paragraphs, an example data correlation mechanism is 827 explained and could be used independently of the adopted solutions. 829 When data is collected on the upstream and downstream nodes, e.g., 830 packet counts for packet loss measurement or timestamps for packet 831 delay measurement, and is periodically reported to or pulled by other 832 nodes or an NMS, a certain data correlation mechanism SHOULD be in 833 use to help the nodes or NMS tell whether any two or more packet 834 counts are related to the same block of markers or if any two 835 timestamps are related to the same marked packet. 837 The Alternate-Marking Method described in this document literally 838 splits the packets of the measured flow into different measurement 839 blocks; in addition, a Block Number (BN) could be assigned to each 840 such measurement block. The BN is generated each time a node reads 841 the data (packet counts or timestamps) and is associated with each 842 packet count and timestamp reported to or pulled by other nodes or 843 NMSs. The value of a BN could be calculated as the modulo of the 844 local time (when the data are read) and the interval of the marking 845 time period. 847 When the nodes or NMS see, for example, the same BNs associated with 848 two packet counts from an upstream and a downstream node, 849 respectively, it considers that these two packet counts correspond to 850 the same block, i.e., these two packet counts belong to the same 851 block of markers from the upstream and downstream nodes. The 852 assumption of this BN mechanism is that the measurement nodes are 853 time synchronized. This requires the measurement nodes to have a 854 certain time synchronization capability (e.g., the Network Time 855 Protocol (NTP) [RFC5905] or the IEEE 1588 Precision Time Protocol 856 (PTP) [IEEE-1588]). Synchronization aspects are further discussed in 857 Section 3.2. 859 4.3. Packet Reordering 861 Due to ECMP, packet reordering is very common in an IP network. The 862 accuracy of a marking-based PM, especially packet loss measurement, 863 may be affected by packet reordering. Take a look at the following 864 example: 866 Block : 1 | 2 | 3 | 4 | 5 |... 867 --------|---------|---------|---------|---------|---------|--- 868 Node R1 : AAAAAAA | BBBBBBB | AAAAAAA | BBBBBBB | AAAAAAA |... 869 Node R2 : AAAAABB | AABBBBA | AAABAAA | BBBBBBA | ABAAABA |... 871 Figure 7: Packet Reordering 873 In Figure 7, the packet stream for Node R1 isn't being reordered and 874 can be safely assigned to interval blocks, but the packet stream for 875 Node R2 is being reordered; so, looking at the packet with the marker 876 of "B" in block 3, there is no safe way to tell whether the packet 877 belongs to block 2 or block 4. 879 In general, there is the need to assign packets with the marker of 880 "B" or "A" to the right interval blocks. Most of the packet 881 reordering occurs at the edge of adjacent blocks, and they are easy 882 to handle if the interval of each block is sufficiently large. Then, 883 it can be assumed that the packets with different markers belong to 884 the block that they are closer to. If the interval is small, it is 885 difficult and sometimes impossible to determine to which block a 886 packet belongs. 888 Section 3.2 provides a guidance on how to choose a proper interval 889 and mitigate packet reordering issues. 891 4.4. Packet Fragmentation 893 Fragmentation can be managed with the Alternate-Marking Method and in 894 particular it is possible to give the following guidance: 896 Marking nodes MUST mark all fragments if there are flag bits to 897 use (i.e. it is in the specific encapsulation), as if they were 898 separate packets. 900 Nodes that fragment packets within the measurement domain MUST NOT 901 replicate marks, but SHOULD mark the first fragment if they have 902 the capability to do so. 904 Measurement points MAY simply ignore unmarked fragments and count 905 marked fragments as full packets. However, if resources allow, 906 measurement points MAY make note of both marked and unmarked 907 initial fragments and only increment the corresponding counter if 908 (a) other fragments are also marked, or (b) it observes all other 909 fragments and they are unmarked. 911 The proposed approach allows the marking node to mark all the 912 fragments except in the case of fragmentation within the network 913 domain, in that event it is suggested to mark only the first 914 fragment. In addition it could be possible to take the counters 915 properly in order to keep track of both marked and unmarked 916 fragments. 918 5. Results of the Alternate Marking Experiment 920 The methodology described in the previous sections can be applied to 921 various performance measurement problems, as explained in [RFC8321]. 922 The only requirement is to select and mark the flow to be monitored; 923 in this way, packets are batched by the sender, and each batch is 924 alternately marked such that it can be easily recognized by the 925 receiver. 927 Either one or two flag bits might be available for marking in 928 different deployments: 930 One flag: packet loss measurement SHOULD be done as described in 931 Section 3.1, while delay measurement MAY be done according to the 932 single-marking method described in Section 3.3.1. Mean delay 933 (Section 3.3.1.1) is NOT RECOMMENDED since it implies more 934 computational load. 936 Two flags: packet loss measurement SHOULD be done as described in 937 Section 3.1, while delay measurement SHOULD be done according to 938 double-marking method Section 3.3.2. In this case single-marking 939 MAY also be used in combination with double-marking and the two 940 approaches provide slightly different pieces of information that 941 can be combined to have a more robust data set. 943 The experiment with Alternate Marking methodologies confirmed the 944 following benefits: 946 o easy implementation: it can be implemented by using features 947 already available on major routing platforms, or by applying an 948 optimized implementation of the method for both legacy and newest 949 technologies; 951 o low computational effort: the additional load on processing is 952 negligible; 954 o accurate loss and delay measurements: single packet loss 955 granularity is achieved with a Passive measurement; 957 o potential applicability to any kind of packet-based or frame-based 958 traffic: Ethernet, IP, MPLS, etc., and both unicast and multicast; 960 o robustness: the method can easily tolerate out-of-order packets, 961 and it's not based on "special" packets whose loss could have a 962 negative impact; 964 o flexibility: all the timestamp formats are allowed, because they 965 are managed out of band. The format (the Network Time Protocol 966 (NTP) [RFC5905] or the IEEE 1588 Precision Time Protocol (PTP) 967 [IEEE-1588]) depends on the precision you want; and 969 o no interoperability issues: the features required are available on 970 all current routing platforms. Both a centralized or distributed 971 solution can be used to harvest data from the routers. 973 It is worth mentioning some related work: in particular 974 [IEEE-Network-PNPM] explains the Alternate-Marking method together 975 with new mechanisms based on hashing techniques as also further 976 described in [I-D.mizrahi-ippm-marking]; while 977 [I-D.zhou-ippm-enhanced-alternate-marking] extends the Alternate- 978 Marking Data Fields, to provide enhanced capabilities and allow 979 advanced functionalities. 981 5.1. Controlled Domain requirement 983 The Alternate Marking Method is an example of a solution limited to a 984 controlled domain [RFC8799]. 986 A controlled domain is a managed network that selects, monitors, and 987 controls access by enforcing policies at the domain boundaries, in 988 order to discard undesired external packets entering the domain and 989 check internal packets leaving the domain. It does not necessarily 990 mean that a controlled domain is a single administrative domain or a 991 single organization. A controlled domain can correspond to a single 992 administrative domain or multiple administrative domains under a 993 defined network management. It must be possible to control the 994 domain boundaries, and use specific precautions if traffic traverses 995 the Internet. 997 For security reasons, the Alternate Marking Method is RECOMMENDED 998 only for controlled domains. 1000 6. Compliance with Guidelines from RFC 6390 1002 RFC 6390 [RFC6390] defines a framework and a process for developing 1003 Performance Metrics for protocols above and below the IP layer (such 1004 as IP-based applications that operate over reliable or datagram 1005 transport protocols). 1007 This document doesn't aim to propose a new Performance Metric but 1008 rather a new Method of Measurement for a few Performance Metrics that 1009 have already been standardized. Nevertheless, it's worth applying 1010 guidelines from [RFC6390] to the present document, in order to 1011 provide a more complete and coherent description of the proposed 1012 method. We used a combination of the Performance Metric Definition 1013 template defined in Section 5.4 of [RFC6390] and the Dependencies 1014 laid out in Section 5.5 of that document. 1016 o Metric Name / Metric Description: as already stated, this document 1017 doesn't propose any new Performance Metrics. On the contrary, it 1018 describes a novel method for measuring packet loss [RFC7680]. The 1019 same concept, with small differences, can also be used to measure 1020 delay [RFC7679] and jitter [RFC3393]. The document mainly 1021 describes the applicability to packet loss measurement. 1023 o Method of Measurement or Calculation: according to the method 1024 described in the previous sections, the number of packets lost is 1025 calculated by subtracting the value of the counter on the source 1026 node from the value of the counter on the destination node. Both 1027 counters must refer to the same color. The calculation is 1028 performed when the value of the counters is in a steady state. 1030 The steady state is an intrinsic characteristic of the marking 1031 method counters because the alternation of color makes the 1032 counters associated with each color still one at a time for the 1033 duration of a marking period. 1035 o Units of Measurement: the method calculates and reports the exact 1036 number of packets sent by the source node and not received by the 1037 destination node. 1039 o Measurement Point(s) with Potential Measurement Domain: the 1040 measurement can be performed between adjacent nodes, on a per-link 1041 basis, or along a multi-hop path, provided that the traffic under 1042 measurement follows that path. In case of a multi-hop path, the 1043 measurements can be performed both end-to-end and hop-by-hop. 1045 o Measurement Timing: the method has a constraint on the frequency 1046 of measurements. This is detailed in Section 3.2, where it is 1047 specified that the marking period and the guard band interval are 1048 strictly related each other to avoid out-of-order issues. That is 1049 because, in order to perform a measurement, the counter must be in 1050 a steady state, and this happens when the traffic is being colored 1051 with the alternate color. 1053 o Implementation: the method uses one or two marking bits to color 1054 the packets; this enables the use of policy configurations on the 1055 router to color the packets and accordingly configure the counter 1056 for each color. The path followed by traffic being measured 1057 should be known in advance in order to configure the counters 1058 along the path and be able to compare the correct values. 1060 o Verification: both in the lab and in the operational network, the 1061 methodology has been tested and experimented for packet loss and 1062 delay measurements by using traffic generators together with 1063 precision test instruments and network emulators. 1065 o Use and Applications: the method can be used to measure packet 1066 loss with high precision on live traffic; moreover, by combining 1067 end-to-end and per-link measurements, the method is useful to 1068 pinpoint the single link that is experiencing loss events. 1070 o Reporting Model: the value of the counters has to be sent to a 1071 centralized management system that performs the calculations; such 1072 samples must contain a reference to the time interval they refer 1073 to, so that the management system can perform the correct 1074 correlation; the samples have to be sent while the corresponding 1075 counter is in a steady state (within a time interval); otherwise, 1076 the value of the sample should be stored locally. 1078 o Dependencies: the values of the counters have to be correlated to 1079 the time interval they refer to. 1081 o Organization of Results: the Method of Measurement produces 1082 singletons. 1084 o Parameters: currently, the main parameter of the method is the 1085 time interval used to alternate the colors and read the counters. 1087 7. IANA Considerations 1089 This document has no IANA actions. 1091 8. Security Considerations 1093 This document specifies a method to perform measurements in the 1094 context of a Service Provider's network and has not been developed to 1095 conduct Internet measurements, so it does not directly affect 1096 Internet security nor applications that run on the Internet. 1097 However, implementation of this method must be mindful of security 1098 and privacy concerns. 1100 There are two types of security concerns: potential harm caused by 1101 the measurements and potential harm to the measurements. 1103 o Harm caused by the measurement: the measurements described in this 1104 document are Passive, so there are no new packets injected into 1105 the network causing potential harm to the network itself and to 1106 data traffic. Nevertheless, the method implies modifications on 1107 the fly to a header or encapsulation of the data packets: this 1108 must be performed in a way that doesn't alter the quality of 1109 service experienced by packets subject to measurements and that 1110 preserves stability and performance of routers doing the 1111 measurements. One of the main security threats in OAM protocols 1112 is network reconnaissance; an attacker can gather information 1113 about the network performance by passively eavesdropping on OAM 1114 messages. The advantage of the methods described in this document 1115 is that the marking bits are the only information that is 1116 exchanged between the network devices. Therefore, Passive 1117 eavesdropping on data-plane traffic does not allow attackers to 1118 gain information about the network performance. 1120 o Harm to the Measurement: the measurements could be harmed by 1121 routers altering the marking of the packets or by an attacker 1122 injecting artificial traffic. Authentication techniques, such as 1123 digital signatures, may be used where appropriate to guard against 1124 injected traffic attacks. Since the measurement itself may be 1125 affected by routers (or other network devices) along the path of 1126 IP packets intentionally altering the value of marking bits of 1127 packets, as mentioned above, the mechanism specified in this 1128 document can be applied just in the context of a controlled 1129 domain; thus, the routers (or other network devices) are locally 1130 administered and this type of attack can be avoided. In addition, 1131 an attacker can't gain information about network performance from 1132 a single monitoring point; it must use synchronized monitoring 1133 points at multiple points on the path, because they have to do the 1134 same kind of measurement and aggregation that Service Providers 1135 using Alternate Marking must do. 1137 Attacks on the data collection and reporting of the statistics 1138 between the monitoring points and the network management system can 1139 interfere with the proper functioning of the system. Hence, the 1140 channels used to report back flow statistics MUST be secured. 1142 The privacy concerns of network measurement are limited because the 1143 method only relies on information contained in the header or 1144 encapsulation without any release of user data. Although information 1145 in the header or encapsulation is metadata that can be used to 1146 compromise the privacy of users, the limited marking technique in 1147 this document seems unlikely to substantially increase the existing 1148 privacy risks from header or encapsulation metadata. It might be 1149 theoretically possible to modulate the marking to serve as a covert 1150 channel, but it would have a very low data rate if it is to avoid 1151 adversely affecting the measurement systems that monitor the marking. 1153 Delay attacks are another potential threat in the context of this 1154 document. Delay measurement is performed using a specific packet in 1155 each block, marked by a dedicated color bit. Therefore, a 1156 man-in-the-middle attacker can selectively induce synthetic delay 1157 only to delay-colored packets, causing systematic error in the delay 1158 measurements. As discussed in previous sections, the methods 1159 described in this document rely on an underlying time synchronization 1160 protocol. Thus, by attacking the time protocol, an attacker can 1161 potentially compromise the integrity of the measurement. A detailed 1162 discussion about the threats against time protocols and how to 1163 mitigate them is presented in RFC 7384 [RFC7384]. 1165 9. Contributors 1167 Xiao Min 1168 ZTE Corp. 1169 Email: xiao.min2@zte.com.cn 1171 10. Acknowledgements 1173 The authors would like to thank Alessandro Capello, Luca Castaldelli, 1174 Mach Chen and Lianshu Zheng for their contribution to the definition 1175 and the implementation of the method. 1177 The authors would also thank Martin Duke and Tommy Pauly for their 1178 assistance and their detailed and precious reviews. 1180 11. References 1182 11.1. Normative References 1184 [IEEE-1588] 1185 IEEE, "IEEE Standard for a Precision Clock Synchronization 1186 Protocol for Networked Measurement and Control Systems", 1187 IEEE Std 1588-2008. 1189 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1190 Requirement Levels", BCP 14, RFC 2119, 1191 DOI 10.17487/RFC2119, March 1997, 1192 . 1194 [RFC5905] Mills, D., Martin, J., Ed., Burbank, J., and W. Kasch, 1195 "Network Time Protocol Version 4: Protocol and Algorithms 1196 Specification", RFC 5905, DOI 10.17487/RFC5905, June 2010, 1197 . 1199 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 1200 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 1201 May 2017, . 1203 11.2. Informative References 1205 [I-D.fioccola-rfc8889bis] 1206 Fioccola, G., Cociglio, M., Sapio, A., and R. Sisto, 1207 "Multipoint Alternate-Marking Method", draft-fioccola- 1208 rfc8889bis-00 (work in progress), November 2021. 1210 [I-D.mizrahi-ippm-marking] 1211 Mizrahi, T., Fioccola, G., Cociglio, M., Chen, M., and G. 1212 Mirsky, "Marking Methods for Performance Measurement", 1213 draft-mizrahi-ippm-marking-00 (work in progress), October 1214 2021. 1216 [I-D.zhou-ippm-enhanced-alternate-marking] 1217 Zhou, T., Fioccola, G., Liu, Y., Lee, S., Cociglio, M., 1218 and W. Li, "Enhanced Alternate Marking Method", draft- 1219 zhou-ippm-enhanced-alternate-marking-07 (work in 1220 progress), July 2021. 1222 [IEEE-Network-PNPM] 1223 IEEE Network, "AM-PM: Efficient Network Telemetry using 1224 Alternate Marking", DOI 10.1109/MNET.2019.1800152, 2019. 1226 [RFC3393] Demichelis, C. and P. Chimento, "IP Packet Delay Variation 1227 Metric for IP Performance Metrics (IPPM)", RFC 3393, 1228 DOI 10.17487/RFC3393, November 2002, 1229 . 1231 [RFC4656] Shalunov, S., Teitelbaum, B., Karp, A., Boote, J., and M. 1232 Zekauskas, "A One-way Active Measurement Protocol 1233 (OWAMP)", RFC 4656, DOI 10.17487/RFC4656, September 2006, 1234 . 1236 [RFC5357] Hedayat, K., Krzanowski, R., Morton, A., Yum, K., and J. 1237 Babiarz, "A Two-Way Active Measurement Protocol (TWAMP)", 1238 RFC 5357, DOI 10.17487/RFC5357, October 2008, 1239 . 1241 [RFC5481] Morton, A. and B. Claise, "Packet Delay Variation 1242 Applicability Statement", RFC 5481, DOI 10.17487/RFC5481, 1243 March 2009, . 1245 [RFC6374] Frost, D. and S. Bryant, "Packet Loss and Delay 1246 Measurement for MPLS Networks", RFC 6374, 1247 DOI 10.17487/RFC6374, September 2011, 1248 . 1250 [RFC6390] Clark, A. and B. Claise, "Guidelines for Considering New 1251 Performance Metric Development", BCP 170, RFC 6390, 1252 DOI 10.17487/RFC6390, October 2011, 1253 . 1255 [RFC6703] Morton, A., Ramachandran, G., and G. Maguluri, "Reporting 1256 IP Network Performance Metrics: Different Points of View", 1257 RFC 6703, DOI 10.17487/RFC6703, August 2012, 1258 . 1260 [RFC7276] Mizrahi, T., Sprecher, N., Bellagamba, E., and Y. 1261 Weingarten, "An Overview of Operations, Administration, 1262 and Maintenance (OAM) Tools", RFC 7276, 1263 DOI 10.17487/RFC7276, June 2014, 1264 . 1266 [RFC7384] Mizrahi, T., "Security Requirements of Time Protocols in 1267 Packet Switched Networks", RFC 7384, DOI 10.17487/RFC7384, 1268 October 2014, . 1270 [RFC7679] Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, 1271 Ed., "A One-Way Delay Metric for IP Performance Metrics 1272 (IPPM)", STD 81, RFC 7679, DOI 10.17487/RFC7679, January 1273 2016, . 1275 [RFC7680] Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, 1276 Ed., "A One-Way Loss Metric for IP Performance Metrics 1277 (IPPM)", STD 82, RFC 7680, DOI 10.17487/RFC7680, January 1278 2016, . 1280 [RFC7799] Morton, A., "Active and Passive Metrics and Methods (with 1281 Hybrid Types In-Between)", RFC 7799, DOI 10.17487/RFC7799, 1282 May 2016, . 1284 [RFC8321] Fioccola, G., Ed., Capello, A., Cociglio, M., Castaldelli, 1285 L., Chen, M., Zheng, L., Mirsky, G., and T. Mizrahi, 1286 "Alternate-Marking Method for Passive and Hybrid 1287 Performance Monitoring", RFC 8321, DOI 10.17487/RFC8321, 1288 January 2018, . 1290 [RFC8799] Carpenter, B. and B. Liu, "Limited Domains and Internet 1291 Protocols", RFC 8799, DOI 10.17487/RFC8799, July 2020, 1292 . 1294 Appendix A. Changes Log 1296 Changes from RFC 8321 include: 1298 o Minor editorial changes 1300 o Replacement of the section on "Applications, Implementation, and 1301 Deployment" with "Finding of the Alternate Marking Implementations 1302 and Deployments" 1304 o Moved advantages and benefits of the method from "Introduction" to 1305 the new section on "Finding of the Alternate Marking 1306 Implementations and Deployments" 1308 o Removed section on "Hybrid Measurement" 1310 Changes in v-(01) include: 1312 o Considerations on the reference: [IEEE-Network-PNPM] 1314 o Clarified that the method based on a fixed timer is specified in 1315 this document while the method based on a fixed number of packets 1316 is only mentioned but not detailed. 1318 o Explanation of the the intrinsic error in section 3.3.1 on 1319 "Single-Marking Methodology" 1321 o Deleted some parts in section 4 "Considerations" that no longer 1322 apply 1324 o New section on "Packet Fragmentation" 1326 Authors' Addresses 1328 Giuseppe Fioccola (editor) 1329 Huawei Technologies 1330 Riesstrasse, 25 1331 Munich 80992 1332 Germany 1334 Email: giuseppe.fioccola@huawei.com 1336 Mauro Cociglio 1337 Telecom Italia 1338 Via Reiss Romoli, 274 1339 Torino 10148 1340 Italy 1342 Email: mauro.cociglio@telecomitalia.it 1344 Greg Mirsky 1345 Ericsson 1347 Email: gregimirsky@gmail.com 1349 Tal Mizrahi 1350 Huawei Technologies 1352 Email: tal.mizrahi.phd@gmail.com 1353 Tianran Zhou 1354 Huawei Technologies 1355 156 Beiqing Rd. 1356 Beijing 100095 1357 China 1359 Email: zhoutianran@huawei.com