idnits 2.17.1 draft-ietf-ippm-2679-bis-03.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 draft header indicates that this document obsoletes RFC2679, but the abstract doesn't seem to directly say this. It does mention RFC2679 though, so this could be OK. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 24, 2015) is 3198 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) ** Downref: Normative reference to an Informational RFC: RFC 2330 ** Obsolete normative reference: RFC 2679 (Obsoleted by RFC 7679) ** Obsolete normative reference: RFC 2680 (Obsoleted by RFC 7680) ** Downref: Normative reference to an Informational RFC: RFC 7312 Summary: 4 errors (**), 0 flaws (~~), 1 warning (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group G. Almes 3 Internet-Draft Texas A&M 4 Obsoletes: 2679 (if approved) S. Kalidindi 5 Intended status: Standards Track Ixia 6 Expires: January 25, 2016 M. Zekauskas 7 Internet2 8 A. Morton, Ed. 9 AT&T Labs 10 July 24, 2015 12 A One-Way Delay Metric for IPPM 13 draft-ietf-ippm-2679-bis-03 15 Abstract 17 This memo (RFC 2679 bis) defines a metric for one-way delay of 18 packets across Internet paths. It builds on notions introduced and 19 discussed in the IPPM Framework document, RFC 2330; the reader is 20 assumed to be familiar with that document. This memo makes RFC 2679 21 obsolete. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at http://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on January 25, 2016. 40 Copyright Notice 42 Copyright (c) 2015 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (http://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Changes from RFC 2679 . . . . . . . . . . . . . . . . . . . . 3 58 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5 59 2.1. Motivation . . . . . . . . . . . . . . . . . . . . . . . 6 60 2.2. General Issues Regarding Time . . . . . . . . . . . . . . 7 61 3. A Singleton Definition for One-way Delay . . . . . . . . . . 8 62 3.1. Metric Name: . . . . . . . . . . . . . . . . . . . . . . 8 63 3.2. Metric Parameters: . . . . . . . . . . . . . . . . . . . 8 64 3.3. Metric Units: . . . . . . . . . . . . . . . . . . . . . . 8 65 3.4. Definition: . . . . . . . . . . . . . . . . . . . . . . . 8 66 3.5. Discussion: . . . . . . . . . . . . . . . . . . . . . . . 9 67 3.6. Methodologies: . . . . . . . . . . . . . . . . . . . . . 10 68 3.7. Errors and Uncertainties: . . . . . . . . . . . . . . . . 11 69 3.7.1. Errors or uncertainties related to Clocks . . . . . . 11 70 3.7.2. Errors or uncertainties related to Wire-time vs Host- 71 time . . . . . . . . . . . . . . . . . . . . . . . . 12 72 3.7.3. Calibration . . . . . . . . . . . . . . . . . . . . . 13 73 3.8. Reporting the metric: . . . . . . . . . . . . . . . . . . 15 74 3.8.1. Type-P . . . . . . . . . . . . . . . . . . . . . . . 15 75 3.8.2. Loss Threshold . . . . . . . . . . . . . . . . . . . 16 76 3.8.3. Calibration Results . . . . . . . . . . . . . . . . . 16 77 3.8.4. Path . . . . . . . . . . . . . . . . . . . . . . . . 16 78 4. A Definition for Samples of One-way Delay . . . . . . . . . . 16 79 4.1. Metric Name: . . . . . . . . . . . . . . . . . . . . . . 17 80 4.2. Metric Parameters: . . . . . . . . . . . . . . . . . . . 17 81 4.3. Metric Units: . . . . . . . . . . . . . . . . . . . . . . 17 82 4.4. Definition: . . . . . . . . . . . . . . . . . . . . . . . 17 83 4.5. Discussion: . . . . . . . . . . . . . . . . . . . . . . . 18 84 4.6. Methodologies: . . . . . . . . . . . . . . . . . . . . . 18 85 4.7. Errors and Uncertainties: . . . . . . . . . . . . . . . . 19 86 4.8. Reporting the metric: . . . . . . . . . . . . . . . . . . 19 87 5. Some Statistics Definitions for One-way Delay . . . . . . . . 19 88 5.1. Type-P-One-way-Delay-Percentile . . . . . . . . . . . . . 19 89 5.2. Type-P-One-way-Delay-Median . . . . . . . . . . . . . . . 20 90 5.3. Type-P-One-way-Delay-Minimum . . . . . . . . . . . . . . 21 91 5.4. Type-P-One-way-Delay-Inverse-Percentile . . . . . . . . . 21 92 6. Security Considerations . . . . . . . . . . . . . . . . . . . 21 93 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22 94 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 22 95 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 96 9.1. Normative References . . . . . . . . . . . . . . . . . . 22 97 9.2. Informative References . . . . . . . . . . . . . . . . . 23 98 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 24 100 1. Changes from RFC 2679 102 Note: This section's placement currently preserves minimal 103 differencer between this memo and RFC 2679. The RFC Editor should 104 place this section in an appropriate place. 106 The following text constitutes RFC 2769 bis proposed for advancement 107 on the IETF Standards Track. This section tracks the changes from 108 [RFC2679]. 110 [RFC6808] provides the test plan and results supporting [RFC2679] 111 advancement along the standards track, according to the process in 112 [RFC6576]. The conclusions of [RFC6808] list four minor 113 modifications: 115 1. Section 6.2.3 of [RFC6808] asserts that the assumption of post- 116 processing to enforce a constant waiting time threshold is 117 compliant, and that the text of the RFC should be revised 118 slightly to include this point (see the last list item of section 119 3.6, below). 121 2. Section 6.5 of [RFC6808] indicates that Type-P-One-way-Delay- 122 Inverse-Percentile statistic has been ignored in both 123 implementations, so it is a candidate for removal or deprecation 124 in RFC2679bis (this small discrepancy does not affect candidacy 125 for advancement) (see section 5.4, below). 127 3. The IETF has reached consensus on guidance for reporting metrics 128 in [RFC6703], and this memo should be referenced in RFC2679bis to 129 incorporate recent experience where appropriate (see the last 130 list item of section 3.6, section 3.8, and section 5 below). 132 4. There is currently one erratum with status "Held for document 133 update" for [RFC2679], and it appears this minor revision and 134 additional text should be incorporated in RFC2679bis (see section 135 5.1). 137 A number of updates to the [RFC2679] text have been implemented in 138 the text below, to reference key IPPM RFCs that were approved after 139 [RFC2679], and to address comments on the IPPM mailing list 140 describing current conditions and experience. 142 1. Near the end of section 2.1, update of a network example using 143 ATM and clarification of TCP's affect on queue occupation and 144 importance of one-way delay measurement. 146 2. Explicit inclusion of the maximum waiting time input parameter 147 in section 3.2 and 4.2, reflecting recognition of this parameter 148 in more recent RFCs and ITU-T Recommendation Y.1540. 150 3. Addition of reference to RFC6703 in the discussion of packet 151 life time and application timeouts in section 3.5. 153 4. Addition of reference to the default requirement (that packets 154 be standard-formed) from RFC2330 as a new list item in section 155 3.5. 157 5. GPS-based NTP experience replaces "to be tested" in section 3.5. 159 6. Replaced "precedence" with updated terminology (DS Field) in 3.6 160 and 3.8.1 (with reference). 162 7. Added parenthetical guidance on minimizing interval between 163 timestamp placement to send time in section 3.6. 165 8. Added text recognizing the impending deployment of transport 166 layer encryption in section 3.6. 168 9. Section 3.7.2 notes that some current systems perform host time 169 stamping on the network interface hardware. 171 10. "instrument" replaced by the defined term "host" in sections 172 3.7.3 and 3.8.3. 174 11. Added reference to RFC 3432 Periodic sampling alongside Poisson 175 sampling in section 4, and also noting that a truncated Poisson 176 distribution may be needed with modern networks as described in 177 the IPPM Framework update, RFC7312. 179 12. Add reference to RFC 4737 Reordering metric in the related 180 discussion of section 4.6, Methodologies. 182 13. Formatting of Example in section 5.1 modified to match the 183 original (issue with conversion to XML in bis version). 185 14. Clarifying the conclusions on two related points on harm to 186 measurements (recognition of measurement traffic for unexpected 187 priority treatment and attacker traffic which emulates 188 measurement) in section 6, Security Considerations. 190 Section 5.4.4 of [RFC6390] suggests a common template for performance 191 metrics partially derived from previous IPPM and BMWG RFCs, but also 192 contains some new items. All of the [RFC6390] Normative points are 193 covered, but not quite in the same section names or orientation. 195 Several of the Informative points are covered. Maintaining the 196 familiar outline of IPPM literature has both value and minimizes 197 unnecessary differences between this revised RFC and current/future 198 IPPM RFCs. 200 The publication of RFC 6921 suggested an area where this memo might 201 need updating. Packet transfer on Faster-Than-Light (FTL) networks 202 could result in negative delays and packet reordering, however both 203 are covered as possibilities in the current text and no revisions are 204 deemed necessary (we also note that this is an April 1st RFC). 206 2. Introduction 208 This memo defines a metric for one-way delay of packets across 209 Internet paths. It builds on notions introduced and discussed in the 210 IPPM Framework document, [RFC2330]; the reader is assumed to be 211 familiar with that document, and its recent update [RFC7312]. 213 This memo is intended to be parallel in structure to a companion 214 document for Packet Loss ("A One-way Packet Loss Metric for IPPM") 215 [RFC2680]. 217 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 218 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 219 document are to be interpreted as described in RFC 2119 [RFC2119]. 220 Although [RFC2119] was written with protocols in mind, the key words 221 are used in this document for similar reasons. They are used to 222 ensure the results of measurements from two different implementations 223 are comparable, and to note instances when an implementation could 224 perturb the network. 226 The structure of the memo is as follows: 228 + A 'singleton' analytic metric, called Type-P-One-way-Delay, will be 229 introduced to measure a single observation of one-way delay. 231 + Using this singleton metric, a 'sample', called Type-P-One-way- 232 Delay-Poisson-Stream, will be introduced to measure a sequence of 233 singleton delays sent at times taken from a Poisson process. 235 + Using this sample, several 'statistics' of the sample will be 236 defined and discussed. This progression from singleton to sample to 237 statistics, with clear separation among them, is important. 239 Whenever a technical term from the IPPM Framework document is first 240 used in this memo, it will be tagged with a trailing asterisk. For 241 example, "term*" indicates that "term" is defined in the Framework. 243 2.1. Motivation 245 One-way delay of a Type-P* packet from a source host* to a 246 destination host is useful for several reasons: 248 + Some applications do not perform well (or at all) if end-to-end 249 delay between hosts is large relative to some threshold value. 251 + Erratic variation in delay makes it difficult (or impossible) to 252 support many real-time applications. 254 + The larger the value of delay, the more difficult it is for 255 transport-layer protocols to sustain high bandwidths. 257 + The minimum value of this metric provides an indication of the 258 delay due only to propagation and transmission delay. 260 + The minimum value of this metric provides an indication of the 261 delay that will likely be experienced when the path* traversed is 262 lightly loaded. 264 + Values of this metric above the minimum provide an indication of 265 the congestion present in the path. 267 The measurement of one-way delay instead of round-trip delay is 268 motivated by the following factors: 270 + In today's Internet, the path from a source to a destination may be 271 different than the path from the destination back to the source 272 ("asymmetric paths"), such that different sequences of routers are 273 used for the forward and reverse paths. Therefore round-trip 274 measurements actually measure the performance of two distinct paths 275 together. Measuring each path independently highlights the 276 performance difference between the two paths which may traverse 277 different Internet service providers, and even radically different 278 types of networks (for example, research versus commodity networks, 279 or networks with asymmetric link capacities, or wireless vs. wireline 280 access). 282 + Even when the two paths are symmetric, they may have radically 283 different performance characteristics due to asymmetric queueing. 285 + Performance of an application may depend mostly on the performance 286 in one direction. For example, a TCP-based communication will 287 experience reduced throughput if congestion occurs in one direction 288 of its communication. Trouble shooting may be simplified if the 289 congested direction of TCP transmission can be identified. 291 + In quality-of-service (QoS) enabled networks, provisioning in one 292 direction may be radically different than provisioning in the reverse 293 direction, and thus the QoS guarantees differ. Measuring the paths 294 independently allows the verification of both guarantees. 296 It is outside the scope of this document to say precisely how delay 297 metrics would be applied to specific problems. 299 2.2. General Issues Regarding Time 301 {Comment: the terminology below differs from that defined by ITU-T 302 documents (e.g., G.810, "Definitions and terminology for 303 synchronization networks" and I.356, "B-ISDN ATM layer cell transfer 304 performance"), but is consistent with the IPPM Framework document. 305 In general, these differences derive from the different backgrounds; 306 the ITU-T documents historically have a telephony origin, while the 307 authors of this document (and the Framework) have a computer systems 308 background. Although the terms defined below have no direct 309 equivalent in the ITU-T definitions, after our definitions we will 310 provide a rough mapping. However, note one potential confusion: our 311 definition of "clock" is the computer operating systems definition 312 denoting a time-of-day clock, while the ITU-T definition of clock 313 denotes a frequency reference.} 315 Whenever a time (i.e., a moment in history) is mentioned here, it is 316 understood to be measured in seconds (and fractions) relative to UTC. 318 As described more fully in the Framework document, there are four 319 distinct, but related notions of clock uncertainty: 321 synchronization* 323 measures the extent to which two clocks agree on what time it is. 324 For example, the clock on one host might be 5.4 msec ahead of the 325 clock on a second host. {Comment: A rough ITU-T equivalent is "time 326 error".} 328 accuracy* 330 measures the extent to which a given clock agrees with UTC. For 331 example, the clock on a host might be 27.1 msec behind UTC. {Comment: 332 A rough ITU-T equivalent is "time error from UTC".} 334 resolution* 336 measures the precision of a given clock. For example, the clock on 337 an old Unix host might tick only once every 10 msec, and thus have a 338 resolution of only 10 msec. {Comment: A very rough ITU-T equivalent 339 is "sampling period".} 341 skew* 343 measures the change of accuracy, or of synchronization, with time. 344 For example, the clock on a given host might gain 1.3 msec per hour 345 and thus be 27.1 msec behind UTC at one time and only 25.8 msec an 346 hour later. In this case, we say that the clock of the given host 347 has a skew of 1.3 msec per hour relative to UTC, which threatens 348 accuracy. We might also speak of the skew of one clock relative to 349 another clock, which threatens synchronization. {Comment: A rough 350 ITU-T equivalent is "time drift".} 352 3. A Singleton Definition for One-way Delay 354 3.1. Metric Name: 356 Type-P-One-way-Delay 358 3.2. Metric Parameters: 360 + Src, the IP address of a host 362 + Dst, the IP address of a host 364 + T, a time 366 + Tmax, a loss threshold waiting time 368 3.3. Metric Units: 370 The value of a Type-P-One-way-Delay is either a real number, or an 371 undefined (informally, infinite) number of seconds. 373 3.4. Definition: 375 For a real number dT, >>the *Type-P-One-way-Delay* from Src to Dst at 376 T is dT<< means that Src sent the first bit of a Type-P packet to Dst 377 at wire-time* T and that Dst received the last bit of that packet at 378 wire-time T+dT. 380 >>The *Type-P-One-way-Delay* from Src to Dst at T is undefined 381 (informally, infinite)<< means that Src sent the first bit of a 382 Type-P packet to Dst at wire-time T and that Dst did not receive that 383 packet (within the loss threshold waiting time, Tmax). 385 Suggestions for what to report along with metric values appear in 386 Section 3.8 after a discussion of the metric, methodologies for 387 measuring the metric, and error analysis. 389 3.5. Discussion: 391 Type-P-One-way-Delay is a relatively simple analytic metric, and one 392 that we believe will afford effective methods of measurement. 394 The following issues are likely to come up in practice: 396 + Real delay values will be positive. Therefore, it does not make 397 sense to report a negative value as a real delay. However, an 398 individual zero or negative delay value might be useful as part of a 399 stream when trying to discover a distribution of a stream of delay 400 values. 402 + Since delay values will often be as low as the 100 usec to 10 msec 403 range, it will be important for Src and Dst to synchronize very 404 closely. GPS systems afford one way to achieve synchronization to 405 within several 10s of usec. Ordinary application of NTP may allow 406 synchronization to within several msec, but this depends on the 407 stability and symmetry of delay properties among those NTP agents 408 used, and this delay is what we are trying to measure. A combination 409 of some GPS-based NTP servers and a conservatively designed and 410 deployed set of other NTP servers should yield good results. This 411 was tested in [RFC6808], where a GPS measurement system's results 412 compared well with a GPS-based NTP synchronized system for the same 413 intercontinental path. 415 + A given methodology will have to include a way to determine whether 416 a delay value is infinite or whether it is merely very large (and the 417 packet is yet to arrive at Dst). As noted by Mahdavi and Paxson 418 [RFC2678], simple upper bounds (such as the 255 seconds theoretical 419 upper bound on the lifetimes of IP packets [RFC0791]) could be used, 420 but good engineering, including an understanding of packet lifetimes, 421 will be needed in practice. {Comment: Note that, for many 422 applications of these metrics, the harm in treating a large delay as 423 infinite might be zero or very small. A TCP data packet, for 424 example, that arrives only after several multiples of the RTT may as 425 well have been lost. See section 4.1.1 of [RFC6703] for examination 426 of unusual packet delays and application performance estimation.} 428 + If the packet is duplicated along the path (or paths) so that 429 multiple non-corrupt copies arrive at the destination, then the 430 packet is counted as received, and the first copy to arrive 431 determines the packet's one-way delay. 433 + If the packet is fragmented and if, for whatever reason, reassembly 434 does not occur, then the packet will be deemed lost. 436 + The packet is standard-formed, the default criteria for all metric 437 definitions defined in Section 15 of [RFC2330], otherwise the packet 438 will be deemed lost. 440 3.6. Methodologies: 442 As with other Type-P-* metrics, the detailed methodology will depend 443 on the Type-P (e.g., protocol number, UDP/TCP port number, size, 444 Differentiated Services (DS) Field [RFC2780]). 446 Generally, for a given Type-P, the methodology would proceed as 447 follows: 449 + Arrange that Src and Dst are synchronized; that is, that they have 450 clocks that are very closely synchronized with each other and each 451 fairly close to the actual time. 453 + At the Src host, select Src and Dst IP addresses, and form a test 454 packet of Type-P with these addresses. Any 'padding' portion of the 455 packet needed only to make the test packet a given size should be 456 filled with randomized bits to avoid a situation in which the 457 measured delay is lower than it would otherwise be due to compression 458 techniques along the path. Note that use of transport layer 459 encryption will counteract the deployment of network-based analysis 460 and may reduce the adoption of network-based payload optimizations 461 like compression. 463 + At the Dst host, arrange to receive the packet. 465 + At the Src host, place a timestamp in the prepared Type-P packet, 466 and send it towards Dst (ideally minimizing time before sending). 468 + If the packet arrives within a reasonable period of time, take a 469 timestamp as soon as possible upon the receipt of the packet. By 470 subtracting the two timestamps, an estimate of one-way delay can be 471 computed. Error analysis of a given implementation of the method 472 must take into account the closeness of synchronization between Src 473 and Dst. If the delay between Src's timestamp and the actual sending 474 of the packet is known, then the estimate could be adjusted by 475 subtracting this amount; uncertainty in this value must be taken into 476 account in error analysis. Similarly, if the delay between the 477 actual receipt of the packet and Dst's timestamp is known, then the 478 estimate could be adjusted by subtracting this amount; uncertainty in 479 this value must be taken into account in error analysis. See the 480 next section, "Errors and Uncertainties", for a more detailed 481 discussion. 483 + If the packet fails to arrive within a reasonable period of time, 484 Tmax, the one-way delay is taken to be undefined (informally, 485 infinite). Note that the threshold of 'reasonable' is a parameter of 486 the metric. These points are examined in detail in [RFC6703], 487 including analysis preferences to assign undefined delay to packets 488 that fail to arrive with the difficulties emerging from the informal 489 "infinite delay" assignment, and an estimation of an upper bound on 490 waiting time for packets in transit. Further, enforcing a specific 491 constant waiting time on stored singletons of one-way delay is 492 compliant with this specification and may allow the results to serve 493 more than one reporting audience. 495 Issues such as the packet format, the means by which Dst knows when 496 to expect the test packet, and the means by which Src and Dst are 497 synchronized are outside the scope of this document. {Comment: We 498 plan to document elsewhere our own work in describing such more 499 detailed implementation techniques and we encourage others to as 500 well.} 502 3.7. Errors and Uncertainties: 504 The description of any specific measurement method should include an 505 accounting and analysis of various sources of error or uncertainty. 506 The Framework document provides general guidance on this point, but 507 we note here the following specifics related to delay metrics: 509 + Errors or uncertainties due to uncertainties in the clocks of the 510 Src and Dst hosts. 512 + Errors or uncertainties due to the difference between 'wire time' 513 and 'host time'. 515 In addition, the loss threshold may affect the results. Each of 516 these are discussed in more detail below, along with a section 517 ("Calibration") on accounting for these errors and uncertainties. 519 3.7.1. Errors or uncertainties related to Clocks 521 The uncertainty in a measurement of one-way delay is related, in 522 part, to uncertainties in the clocks of the Src and Dst hosts. In 523 the following, we refer to the clock used to measure when the packet 524 was sent from Src as the source clock, we refer to the clock used to 525 measure when the packet was received by Dst as the destination clock, 526 we refer to the observed time when the packet was sent by the source 527 clock as Tsource, and the observed time when the packet was received 528 by the destination clock as Tdest. Alluding to the notions of 529 synchronization, accuracy, resolution, and skew mentioned in the 530 Introduction, we note the following: 532 + Any error in the synchronization between the source clock and the 533 destination clock will contribute to error in the delay measurement. 534 We say that the source clock and the destination clock have a 535 synchronization error of Tsynch if the source clock is Tsynch ahead 536 of the destination clock. Thus, if we know the value of Tsynch 537 exactly, we could correct for clock synchronization by adding Tsynch 538 to the uncorrected value of Tdest-Tsource. 540 + The accuracy of a clock is important only in identifying the time 541 at which a given delay was measured. Accuracy, per se, has no 542 importance to the accuracy of the measurement of delay. When 543 computing delays, we are interested only in the differences between 544 clock values, not the values themselves. 546 + The resolution of a clock adds to uncertainty about any time 547 measured with it. Thus, if the source clock has a resolution of 10 548 msec, then this adds 10 msec of uncertainty to any time value 549 measured with it. We will denote the resolution of the source clock 550 and the destination clock as Rsource and Rdest, respectively. 552 + The skew of a clock is not so much an additional issue as it is a 553 realization of the fact that Tsynch is itself a function of time. 554 Thus, if we attempt to measure or to bound Tsynch, this needs to be 555 done periodically. Over some periods of time, this function can be 556 approximated as a linear function plus some higher order terms; in 557 these cases, one option is to use knowledge of the linear component 558 to correct the clock. Using this correction, the residual Tsynch is 559 made smaller, but remains a source of uncertainty that must be 560 accounted for. We use the function Esynch(t) to denote an upper 561 bound on the uncertainty in synchronization. Thus, |Tsynch(t)| <= 562 Esynch(t). 564 Taking these items together, we note that naive computation Tdest- 565 Tsource will be off by Tsynch(t) +/- (Rsource + Rdest). Using the 566 notion of Esynch(t), we note that these clock-related problems 567 introduce a total uncertainty of Esynch(t)+ Rsource + Rdest. This 568 estimate of total clock-related uncertainty should be included in the 569 error/uncertainty analysis of any measurement implementation. 571 3.7.2. Errors or uncertainties related to Wire-time vs Host-time 573 As we have defined one-way delay, we would like to measure the time 574 between when the test packet leaves the network interface of Src and 575 when it (completely) arrives at the network interface of Dst, and we 576 refer to these as "wire times." If the timings are themselves 577 performed by software on Src and Dst, however, then this software can 578 only directly measure the time between when Src grabs a timestamp 579 just prior to sending the test packet and when Dst grabs a timestamp 580 just after having received the test packet, and we refer to these two 581 points as "host times". 583 We note that some systems perform host time stamping on the network 584 interface hardware, in an attempt to minimize the difference from 585 wire times. 587 To the extent that the difference between wire time and host time is 588 accurately known, this knowledge can be used to correct for host time 589 measurements and the corrected value more accurately estimates the 590 desired (wire time) metric. 592 To the extent, however, that the difference between wire time and 593 host time is uncertain, this uncertainty must be accounted for in an 594 analysis of a given measurement method. We denote by Hsource an 595 upper bound on the uncertainty in the difference between wire time 596 and host time on the Src host, and similarly define Hdest for the Dst 597 host. We then note that these problems introduce a total uncertainty 598 of Hsource+Hdest. This estimate of total wire-vs-host uncertainty 599 should be included in the error/uncertainty analysis of any 600 measurement implementation. 602 3.7.3. Calibration 604 Generally, the measured values can be decomposed as follows: 606 measured value = true value + systematic error + random error 608 If the systematic error (the constant bias in measured values) can be 609 determined, it can be compensated for in the reported results. 611 reported value = measured value - systematic error 613 therefore 615 reported value = true value + random error 617 The goal of calibration is to determine the systematic and random 618 error generated by the hosts themselves in as much detail as 619 possible. At a minimum, a bound ("e") should be found such that the 620 reported value is in the range (true value - e) to (true value + e) 621 at least 95 percent of the time. We call "e" the calibration error 622 for the measurements. It represents the degree to which the values 623 produced by the measurement host are repeatable; that is, how closely 624 an actual delay of 30 ms is reported as 30 ms. {Comment: 95 percent 625 was chosen because (1) some confidence level is desirable to be able 626 to remove outliers, which will be found in measuring any physical 627 property; (2) a particular confidence level should be specified so 628 that the results of independent implementations can be compared; and 629 (3) even with a prototype user-level implementation, 95% was loose 630 enough to exclude outliers.} 632 From the discussion in the previous two sections, the error in 633 measurements could be bounded by determining all the individual 634 uncertainties, and adding them together to form 636 Esynch(t) + Rsource + Rdest + Hsource + Hdest. 638 However, reasonable bounds on both the clock-related uncertainty 639 captured by the first three terms and the host-related uncertainty 640 captured by the last two terms should be possible by careful design 641 techniques and calibrating the hosts using a known, isolated, network 642 in a lab. 644 For example, the clock-related uncertainties are greatly reduced 645 through the use of a GPS time source. The sum of Esynch(t) + Rsource 646 + Rdest is small, and is also bounded for the duration of the 647 measurement because of the global time source. 649 The host-related uncertainties, Hsource + Hdest, could be bounded by 650 connecting two hosts back-to-back with a high-speed serial link or 651 isolated LAN segment. In this case, repeated measurements are 652 measuring the same one-way delay. 654 If the test packets are small, such a network connection has a 655 minimal delay that may be approximated by zero. The measured delay 656 therefore contains only systematic and random error in the 657 measurement hosts. The "average value" of repeated measurements is 658 the systematic error, and the variation is the random error. 660 One way to compute the systematic error, and the random error to a 661 95% confidence is to repeat the experiment many times - at least 662 hundreds of tests. The systematic error would then be the median. 663 The random error could then be found by removing the systematic error 664 from the measured values. The 95% confidence interval would be the 665 range from the 2.5th percentile to the 97.5th percentile of these 666 deviations from the true value. The calibration error "e" could then 667 be taken to be the largest absolute value of these two numbers, plus 668 the clock-related uncertainty. {Comment: as described, this bound is 669 relatively loose since the uncertainties are added, and the absolute 670 value of the largest deviation is used. As long as the resulting 671 value is not a significant fraction of the measured values, it is a 672 reasonable bound. If the resulting value is a significant fraction 673 of the measured values, then more exact methods will be needed to 674 compute the calibration error.} 676 Note that random error is a function of measurement load. For 677 example, if many paths will be measured by one host, this might 678 increase interrupts, process scheduling, and disk I/O (for example, 679 recording the measurements), all of which may increase the random 680 error in measured singletons. Therefore, in addition to minimal load 681 measurements to find the systematic error, calibration measurements 682 should be performed with the same measurement load that the hosts 683 will see in the field. 685 We wish to reiterate that this statistical treatment refers to the 686 calibration of the host; it is used to "calibrate the meter stick" 687 and say how well the meter stick reflects reality. 689 In addition to calibrating the hosts for finite one-way delay, two 690 checks should be made to ensure that packets reported as losses were 691 really lost. First, the threshold for loss should be verified. In 692 particular, ensure the "reasonable" threshold is reasonable: that it 693 is very unlikely a packet will arrive after the threshold value, and 694 therefore the number of packets lost over an interval is not 695 sensitive to the error bound on measurements. Second, consider the 696 possibility that a packet arrives at the network interface, but is 697 lost due to congestion on that interface or to other resource 698 exhaustion (e.g. buffers) in the host. 700 3.8. Reporting the metric: 702 The calibration and context in which the metric is measured MUST be 703 carefully considered, and SHOULD always be reported along with metric 704 results. We now present four items to consider: the Type-P of test 705 packets, the threshold of infinite delay (if any), error calibration, 706 and the path traversed by the test packets. This list is not 707 exhaustive; any additional information that could be useful in 708 interpreting applications of the metrics should also be reported (see 709 [RFC6703] for extensive discussion of reporting considerations for 710 different audiences). 712 3.8.1. Type-P 714 As noted in the Framework document [RFC2330], the value of the metric 715 may depend on the type of IP packets used to make the measurement, or 716 "type-P". The value of Type-P-One-way-Delay could change if the 717 protocol (UDP or TCP), port number, size, or arrangement for special 718 treatment (e.g., IP DS Field [RFC2780] or RSVP) changes. The exact 719 Type-P used to make the measurements MUST be accurately reported. 721 3.8.2. Loss Threshold 723 In addition, the threshold (or methodology to distinguish) between a 724 large finite delay and loss MUST be reported. 726 3.8.3. Calibration Results 728 + If the systematic error can be determined, it SHOULD be removed 729 from the measured values. 731 + You SHOULD also report the calibration error, e, such that the true 732 value is the reported value plus or minus e, with 95% confidence (see 733 the last section.) 735 + If possible, the conditions under which a test packet with finite 736 delay is reported as lost due to resource exhaustion on the 737 measurement host SHOULD be reported. 739 3.8.4. Path 741 Finally, the path traversed by the packet SHOULD be reported, if 742 possible. In general it is impractical to know the precise path a 743 given packet takes through the network. The precise path may be 744 known for certain Type-P on short or stable paths. If Type-P 745 includes the record route (or loose-source route) option in the IP 746 header, and the path is short enough, and all routers* on the path 747 support record (or loose-source) route, then the path will be 748 precisely recorded. This is impractical because the route must be 749 short enough, many routers do not support (or are not configured for) 750 record route, and use of this feature would often artificially worsen 751 the performance observed by removing the packet from common-case 752 processing. However, partial information is still valuable context. 753 For example, if a host can choose between two links* (and hence two 754 separate routes from Src to Dst), then the initial link used is 755 valuable context. {Comment: For example, with Merit's NetNow setup, a 756 Src on one NAP can reach a Dst on another NAP by either of several 757 different backbone networks.} 759 4. A Definition for Samples of One-way Delay 761 Given the singleton metric Type-P-One-way-Delay, we now define one 762 particular sample of such singletons. The idea of the sample is to 763 select a particular binding of the parameters Src, Dst, and Type-P, 764 then define a sample of values of parameter T. The means for 765 defining the values of T is to select a beginning time T0, a final 766 time Tf, and an average rate lambda, then define a pseudo-random 767 Poisson process of rate lambda, whose values fall between T0 and Tf. 769 The time interval between successive values of T will then average 1/ 770 lambda. 772 Note that Poisson sampling is only one way of defining a sample. 773 Poisson has the advantage of limiting bias, but other methods of 774 sampling will be appropriate for different situations. For example, 775 a truncated Poisson distribution may be needed to avoid reactive 776 network state changes during intervals of inactivity, see section 4.6 777 of [RFC7312]. Sometimes, the goal is sampling with a known bias, and 778 [RFC3432] describes a method for periodic sampling with random start 779 times. 781 4.1. Metric Name: 783 Type-P-One-way-Delay-Poisson-Stream 785 4.2. Metric Parameters: 787 + Src, the IP address of a host 789 + Dst, the IP address of a host 791 + T0, a time 793 + Tf, a time 795 + Tmax, a loss threshold waiting time 797 + lambda, a rate in reciprocal seconds (or parameters for another 798 distribution) 800 4.3. Metric Units: 802 A sequence of pairs; the elements of each pair are: 804 + T, a time, and 806 + dT, either a real number or an undefined number of seconds. 808 The values of T in the sequence are monotonic increasing. Note that 809 T would be a valid parameter to Type-P-One-way-Delay, and that dT 810 would be a valid value of Type-P-One-way-Delay. 812 4.4. Definition: 814 Given T0, Tf, and lambda, we compute a pseudo-random Poisson process 815 beginning at or before T0, with average arrival rate lambda, and 816 ending at or after Tf. Those time values greater than or equal to T0 817 and less than or equal to Tf are then selected. At each of the times 818 in this process, we obtain the value of Type-P-One-way-Delay at this 819 time. The value of the sample is the sequence made up of the 820 resulting pairs. If there are no such pairs, the 821 sequence is of length zero and the sample is said to be empty. 823 4.5. Discussion: 825 The reader should be familiar with the in-depth discussion of Poisson 826 sampling in the Framework document [RFC2330], which includes methods 827 to compute and verify the pseudo-random Poisson process. 829 We specifically do not constrain the value of lambda, except to note 830 the extremes. If the rate is too large, then the measurement traffic 831 will perturb the network, and itself cause congestion. If the rate 832 is too small, then you might not capture interesting network 833 behavior. {Comment: We expect to document our experiences with, and 834 suggestions for, lambda elsewhere, culminating in a "best current 835 practices" document.} 837 Since a pseudo-random number sequence is employed, the sequence of 838 times, and hence the value of the sample, is not fully specified. 839 Pseudo-random number generators of good quality will be needed to 840 achieve the desired qualities. 842 The sample is defined in terms of a Poisson process both to avoid the 843 effects of self-synchronization and also capture a sample that is 844 statistically as unbiased as possible. {Comment: there is, of course, 845 no claim that real Internet traffic arrives according to a Poisson 846 arrival process.} The Poisson process is used to schedule the delay 847 measurements. The test packets will generally not arrive at Dst 848 according to a Poisson distribution, since they are influenced by the 849 network. 851 All the singleton Type-P-One-way-Delay metrics in the sequence will 852 have the same values of Src, Dst, and Type-P. 854 Note also that, given one sample that runs from T0 to Tf, and given 855 new time values T0' and Tf' such that T0 <= T0' <= Tf' <= Tf, the 856 subsequence of the given sample whose time values fall between T0' 857 and Tf' are also a valid Type-P-One-way-Delay-Poisson-Stream sample. 859 4.6. Methodologies: 861 The methodologies follow directly from: 863 + the selection of specific times, using the specified Poisson 864 arrival process, and 865 + the methodologies discussion already given for the singleton Type- 866 P-One-way-Delay metric. 868 Care must, of course, be given to correctly handle out-of-order 869 arrival of test packets; it is possible that the Src could send one 870 test packet at TS[i], then send a second one (later) at TS[i+1], 871 while the Dst could receive the second test packet at TR[i+1], and 872 then receive the first one (later) at TR[i]. Metrics for reordering 873 may be found in [RFC4737]. 875 4.7. Errors and Uncertainties: 877 In addition to sources of errors and uncertainties associated with 878 methods employed to measure the singleton values that make up the 879 sample, care must be given to analyze the accuracy of the Poisson 880 process with respect to the wire-times of the sending of the test 881 packets. Problems with this process could be caused by several 882 things, including problems with the pseudo-random number techniques 883 used to generate the Poisson arrival process, or with jitter in the 884 value of Hsource (mentioned above as uncertainty in the singleton 885 delay metric). The Framework document shows how to use the Anderson- 886 Darling test to verify the accuracy of a Poisson process over small 887 time frames. {Comment: The goal is to ensure that test packets are 888 sent "close enough" to a Poisson schedule, and avoid periodic 889 behavior.} 891 4.8. Reporting the metric: 893 You MUST report the calibration and context for the underlying 894 singletons along with the stream. (See "Reporting the metric" for 895 Type-P-One-way-Delay.) 897 5. Some Statistics Definitions for One-way Delay 899 Given the sample metric Type-P-One-way-Delay-Poisson-Stream, we now 900 offer several statistics of that sample. These statistics are 901 offered mostly to illustrate what could be done. See [RFC6703] for 902 additional discussion of statistics that are relevant to different 903 audiences. 905 5.1. Type-P-One-way-Delay-Percentile 907 Given a Type-P-One-way-Delay-Poisson-Stream and a percent X between 908 0% and 100%, the Xth percentile of all the dT values in the Stream. 909 In computing this percentile, undefined values are treated as 910 infinitely large. Note that this means that the percentile could 911 thus be undefined (informally, infinite). In addition, the Type-P- 912 One-way-Delay-Percentile is undefined if the sample is empty. 914 Example: suppose we take a sample and the results are: 916 Stream1 = < 918 920 922 924 926 928 > 930 Then the 50th percentile would be 110 msec, since 90 msec and 100 931 msec are smaller and 500 msec and 'undefined' are larger. See 932 Section 11.3 of [RFC2330] for computing percentiles. 934 Note that if the possibility that a packet with finite delay is 935 reported as lost is significant, then a high percentile (90th or 936 95th) might be reported as infinite instead of finite. 938 5.2. Type-P-One-way-Delay-Median 940 Given a Type-P-One-way-Delay-Poisson-Stream, the median of all the dT 941 values in the Stream. In computing the median, undefined values are 942 treated as infinitely large. As with Type-P-One-way-Delay- 943 Percentile, Type-P-One-way-Delay-Median is undefined if the sample is 944 empty. 946 As noted in the Framework document, the median differs from the 50th 947 percentile only when the sample contains an even number of values, in 948 which case the mean of the two central values is used. 950 Example: suppose we take a sample and the results are: 952 Stream2 = < 954 956 958 960 961 > 963 Then the median would be 105 msec, the mean of 100 msec and 110 msec, 964 the two central values. 966 5.3. Type-P-One-way-Delay-Minimum 968 Given a Type-P-One-way-Delay-Poisson-Stream, the minimum of all the 969 dT values in the Stream. In computing this, undefined values are 970 treated as infinitely large. Note that this means that the minimum 971 could thus be undefined (informally, infinite) if all the dT values 972 are undefined. In addition, the Type-P-One-way-Delay-Minimum is 973 undefined if the sample is empty. 975 In the above example, the minimum would be 90 msec. 977 5.4. Type-P-One-way-Delay-Inverse-Percentile 979 Note: This statistic is deprecated in this version of the memo 980 because of lack of use. 982 Given a Type-P-One-way-Delay-Poisson-Stream and a time duration 983 threshold, the fraction of all the dT values in the Stream less than 984 or equal to the threshold. The result could be as low as 0% (if all 985 the dT values exceed threshold) or as high as 100%. Type-P-One-way- 986 Delay-Inverse-Percentile is undefined if the sample is empty. 988 In the above example, the Inverse-Percentile of 103 msec would be 989 50%. 991 6. Security Considerations 993 Conducting Internet measurements raises both security and privacy 994 concerns. This memo does not specify an implementation of the 995 metrics, so it does not directly affect the security of the Internet 996 nor of applications which run on the Internet. However, 997 implementations of these metrics must be mindful of security and 998 privacy concerns. 1000 There are two types of security concerns: potential harm caused by 1001 the measurements, and potential harm to the measurements. The 1002 measurements could cause harm because they are active, and inject 1003 packets into the network. The measurement parameters MUST be 1004 carefully selected so that the measurements inject trivial amounts of 1005 additional traffic into the networks they measure. If they inject 1006 "too much" traffic, they can skew the results of the measurement, and 1007 in extreme cases cause congestion and denial of service. 1009 The measurements themselves could be harmed by routers giving 1010 measurement traffic a different priority than "normal" traffic, or by 1011 an attacker injecting artificial measurement traffic. If routers can 1012 recognize measurement traffic and treat it separately, the 1013 measurements will not reflect actual user traffic. Therefore, the 1014 measurement methodologies SHOULD include appropriate techniques to 1015 reduce the probability measurement traffic can be distinguished from 1016 "normal" traffic. 1018 If an attacker injects packets emulating traffic that are accepted as 1019 legitimate, the loss ratio or other measured values could be 1020 corrupted. Authentication techniques, such as digital signatures, 1021 may be used where appropriate to guard against injected traffic 1022 attacks. 1024 The privacy concerns of network measurement are limited by the active 1025 measurements described in this memo. Unlike passive measurements, 1026 there can be no release of existing user data. 1028 7. IANA Considerations 1030 This memo makes no requests of IANA. 1032 8. Acknowledgements 1034 For [RFC2679], special thanks are due to Vern Paxson of Lawrence 1035 Berkeley Labs for his helpful comments on issues of clock uncertainty 1036 and statistics. Thanks also to Garry Couch, Will Leland, Andy 1037 Scherrer, Sean Shapira, and Roland Wittig for several useful 1038 suggestions. 1040 For RFC 2679 bis, thanks to Joachim Fabini, Ruediger Geib, Nalini 1041 Elkins, and Barry Constantine for sharing their measurement 1042 experience as part of their careful reviews. 1044 9. References 1046 9.1. Normative References 1048 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 1049 DOI 10.17487/RFC0791, September 1981, 1050 . 1052 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1053 Requirement Levels", BCP 14, RFC 2119, 1054 DOI 10.17487/RFC2119, March 1997, 1055 . 1057 [RFC2330] Paxson, V., Almes, G., Mahdavi, J., and M. Mathis, 1058 "Framework for IP Performance Metrics", RFC 2330, 1059 DOI 10.17487/RFC2330, May 1998, 1060 . 1062 [RFC2678] Mahdavi, J. and V. Paxson, "IPPM Metrics for Measuring 1063 Connectivity", RFC 2678, DOI 10.17487/RFC2678, September 1064 1999, . 1066 [RFC2679] Almes, G., Kalidindi, S., and M. Zekauskas, "A One-way 1067 Delay Metric for IPPM", RFC 2679, DOI 10.17487/RFC2679, 1068 September 1999, . 1070 [RFC2680] Almes, G., Kalidindi, S., and M. Zekauskas, "A One-way 1071 Packet Loss Metric for IPPM", RFC 2680, 1072 DOI 10.17487/RFC2680, September 1999, 1073 . 1075 [RFC2780] Bradner, S. and V. Paxson, "IANA Allocation Guidelines For 1076 Values In the Internet Protocol and Related Headers", 1077 BCP 37, RFC 2780, DOI 10.17487/RFC2780, March 2000, 1078 . 1080 [RFC3432] Raisanen, V., Grotefeld, G., and A. Morton, "Network 1081 performance measurement with periodic streams", RFC 3432, 1082 DOI 10.17487/RFC3432, November 2002, 1083 . 1085 [RFC6576] Geib, R., Ed., Morton, A., Fardid, R., and A. Steinmitz, 1086 "IP Performance Metrics (IPPM) Standard Advancement 1087 Testing", BCP 176, RFC 6576, DOI 10.17487/RFC6576, March 1088 2012, . 1090 [RFC7312] Fabini, J. and A. Morton, "Advanced Stream and Sampling 1091 Framework for IP Performance Metrics (IPPM)", RFC 7312, 1092 DOI 10.17487/RFC7312, August 2014, 1093 . 1095 9.2. Informative References 1097 [RFC4737] Morton, A., Ciavattone, L., Ramachandran, G., Shalunov, 1098 S., and J. Perser, "Packet Reordering Metrics", RFC 4737, 1099 DOI 10.17487/RFC4737, November 2006, 1100 . 1102 [RFC6390] Clark, A. and B. Claise, "Guidelines for Considering New 1103 Performance Metric Development", BCP 170, RFC 6390, 1104 DOI 10.17487/RFC6390, October 2011, 1105 . 1107 [RFC6703] Morton, A., Ramachandran, G., and G. Maguluri, "Reporting 1108 IP Network Performance Metrics: Different Points of View", 1109 RFC 6703, DOI 10.17487/RFC6703, August 2012, 1110 . 1112 [RFC6808] Ciavattone, L., Geib, R., Morton, A., and M. Wieser, "Test 1113 Plan and Results Supporting Advancement of RFC 2679 on the 1114 Standards Track", RFC 6808, DOI 10.17487/RFC6808, December 1115 2012, . 1117 Authors' Addresses 1119 Guy Almes 1120 Texas A&M 1122 Email: almes@acm.org 1124 Sunil Kalidindi 1125 Ixia 1127 Email: skalidindi@ixiacom.com 1129 Matt Zekauskas 1130 Internet2 1132 Email: matt@internet2.edu 1134 Al Morton (editor) 1135 AT&T Labs 1136 200 Laurel Avenue South 1137 Middletown, NJ 07748 1138 USA 1140 Phone: +1 732 420 1571 1141 Fax: +1 732 368 1192 1142 Email: acmorton@att.com 1143 URI: http://home.comcast.net/~acmacm/