idnits 2.17.1 draft-ietf-ippm-2330-update-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 updates RFC2330, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC2330, updated by this document, for RFC5378 checks: 1998-05-01) -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (March 23, 2014) is 3687 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'Mat98' is mentioned on line 497, but not defined == Unused Reference: 'RFC2026' is defined on line 616, but no explicit reference was found in the text == Unused Reference: 'RFC2680' is defined on line 629, but no explicit reference was found in the text == Unused Reference: 'RFC5657' is defined on line 644, but no explicit reference was found in the text ** Obsolete normative reference: RFC 2679 (Obsoleted by RFC 7679) ** Obsolete normative reference: RFC 2680 (Obsoleted by RFC 7680) Summary: 2 errors (**), 0 flaws (~~), 5 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group J. Fabini 3 Internet-Draft Vienna University of Technology 4 Updates: 2330 (if approved) A. Morton 5 Intended status: Informational AT&T Labs 6 Expires: September 24, 2014 March 23, 2014 8 Advanced Stream and Sampling Framework for IPPM 9 draft-ietf-ippm-2330-update-03 11 Abstract 13 To obtain repeatable results in modern networks, test descriptions 14 need an expanded stream parameter framework that also augments 15 aspects specified as Type-P for test packets. This memo proposes to 16 update the IP Performance Metrics (IPPM) Framework with advanced 17 considerations for measurement methodology and testing. The existing 18 framework mostly assumes deterministic connectivity, and that a 19 single test stream will represent the characteristics of the path 20 when it is aggregated with other flows. Networks have evolved and 21 test stream descriptions must evolve with them, otherwise unexpected 22 network features may dominate the measured performance. This memo 23 describes new stream parameters for both network characterization and 24 support of application design using IPPM metrics. 26 Requirements Language 28 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 29 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 30 document are to be interpreted as described in RFC 2119 [RFC2119]. 32 Status of This Memo 34 This Internet-Draft is submitted in full conformance with the 35 provisions of BCP 78 and BCP 79. 37 Internet-Drafts are working documents of the Internet Engineering 38 Task Force (IETF). Note that other groups may also distribute 39 working documents as Internet-Drafts. The list of current Internet- 40 Drafts is at http://datatracker.ietf.org/drafts/current/. 42 Internet-Drafts are draft documents valid for a maximum of six months 43 and may be updated, replaced, or obsoleted by other documents at any 44 time. It is inappropriate to use Internet-Drafts as reference 45 material or to cite them other than as "work in progress." 47 This Internet-Draft will expire on September 24, 2014. 49 Copyright Notice 51 Copyright (c) 2014 IETF Trust and the persons identified as the 52 document authors. All rights reserved. 54 This document is subject to BCP 78 and the IETF Trust's Legal 55 Provisions Relating to IETF Documents 56 (http://trustee.ietf.org/license-info) in effect on the date of 57 publication of this document. Please review these documents 58 carefully, as they describe your rights and restrictions with respect 59 to this document. Code Components extracted from this document must 60 include Simplified BSD License text as described in Section 4.e of 61 the Trust Legal Provisions and are provided without warranty as 62 described in the Simplified BSD License. 64 Table of Contents 66 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 67 1.1. Definition: Reactive Path Behavior . . . . . . . . . . . 3 68 2. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 69 3. New or Revised Stream Parameters . . . . . . . . . . . . . . 5 70 3.1. Test Packet Type-P . . . . . . . . . . . . . . . . . . . 6 71 3.1.1. Multiple Test Packet Lengths . . . . . . . . . . . . 6 72 3.1.2. Test Packet Payload Content Optimization . . . . . . 7 73 3.2. Packet History . . . . . . . . . . . . . . . . . . . . . 7 74 3.3. Access Technology Change . . . . . . . . . . . . . . . . 8 75 3.4. Time-Slotted Randomness Cancellation . . . . . . . . . . 8 76 4. Quality of Metrics and Methodologies . . . . . . . . . . . . 9 77 4.1. Repeatability . . . . . . . . . . . . . . . . . . . . . . 9 78 4.2. Continuity . . . . . . . . . . . . . . . . . . . . . . . 10 79 4.3. Actionable . . . . . . . . . . . . . . . . . . . . . . . 11 80 4.4. Conservative . . . . . . . . . . . . . . . . . . . . . . 12 81 4.5. Spatial and Temporal Composition . . . . . . . . . . . . 12 82 4.6. Poisson Sampling . . . . . . . . . . . . . . . . . . . . 12 83 5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 13 84 6. Security Considerations . . . . . . . . . . . . . . . . . . . 13 85 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 13 86 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 13 87 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 13 88 9.1. Normative References . . . . . . . . . . . . . . . . . . 13 89 9.2. Informative References . . . . . . . . . . . . . . . . . 14 90 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 15 92 1. Introduction 94 The IETF IP Performance Metrics (IPPM) working group first created a 95 framework for metric development in [RFC2330]. This framework has 96 stood the test of time and enabled development of many fundamental 97 metrics, while only being updated once in a specific area [RFC5835]. 99 The IPPM framework [RFC2330] generally relies on several assumptions, 100 one of which is not explicitly stated but assumed: lightly loaded 101 paths conform to the linear "delay = packet size / capacity" 102 equation, being state/history-less (with some exceptions, firewalls 103 are mentioned). However, this does not hold true for many modern 104 network technologies, such as reactive paths (those with demand- 105 driven resource allocation) and links with time-slotted operation. 106 Per-flow state can be observed on test packet streams, and such 107 treatment will influence network characterization if it is not taken 108 into account. Flow history will also affect the performance of 109 applications and be perceived by their users. 111 Moreover, Sections 4 and 6.2 of [RFC2330] explicitly recommend 112 repeatable measurement metrics and methodologies. Measurements in 113 today's access networks illustrate that methodological guidelines of 114 [RFC2330] must be extended to capture the reactive nature of these 115 networks. Although the proposed extensions can support methodologies 116 to fulfill the continuity requirement stated in section 6.2 of 117 [RFC2330], there is no guarantee. Practical measurements confirm 118 that some link types exhibit distinct responses to repeated 119 measurements with identical stimulus, i.e., identical traffic 120 patterns. If feasible, appropriate fine-tuning of measurement 121 traffic patterns can improve measurement continuity and repeatability 122 for these link types as shown in [IBD]. 124 We stress that this update of [RFC2330] does not invalidate or 125 require changes to the analytic metric definitions prepared in the 126 IPPM working group to date. Rather, it adds considerations for 127 active measurement methodologies and expands the importance of 128 existing conventions and notions in [RFC2330], such as "packets of 129 Type-P". 131 Among the evolutionary networking changes is a phenomenon we call 132 "reactive behavior", defined below. 134 1.1. Definition: Reactive Path Behavior 136 Reactive path behavior will be observable by the test packet stream 137 as a repeatable phenomenon where packet transfer performance 138 characteristics *change* according to prior observations of the 139 packet flow of interest (at the reactive host or link). Therefore, 140 reactive path behavior is nominally deterministic with respect to the 141 flow of interest. Other flows or traffic load conditions may result 142 in additional performance-affecting reactions, but these are external 143 to the characteristics of the flow of interest. 145 In practice, a sender may not have absolute control of the ingress 146 packet stream characteristics at a reactive host or link, but this 147 does not change the deterministic reactions present there. If we 148 measure a path, the arrival characteristics at the reactive host/link 149 are determined by the sending characteristics and the transfer 150 characteristics of intervening hosts and links. Identical traffic 151 patterns at the sending host might generate distinct patterns at the 152 reactive host's/link's input due to impairments in the intermediate 153 subpath. The reactive host/link is expected to provide deterministic 154 response on identical input patterns. 156 Other than the size of the payload at the layer of interest and the 157 header itself, packet content does not influence the measurement. 158 Reactive behavior at the IP layer is not influenced by the TCP ports 159 in use, for example. Therefore, the indication of reactive behavior 160 must include the layer at which measurements are instituted. 162 Examples include links with Active/In-active state detectors, and 163 hosts or links that revise their traffic serving and forwarding rates 164 (up or down) based on packet arrival history. 166 Although difficult to handle from a measurement point of view, 167 reactive paths entities are usually designed to improve overall 168 network performance and user experience, for example by making 169 capacity available to an active user. Reactive behavior may be an 170 artifact of solutions to allocate scarce resources according to the 171 demands of users, thus it is an important problem to solve for 172 measurement and other disciplines, such as application design. 174 2. Scope 176 The purpose of this memo is to foster repeatable measurement results 177 in modern networks by highlighting the key aspects of test streams 178 and packets and make them part of the IPPM performance metric 179 framework. 181 The scope is to update key sections of [RFC2330], adding 182 considerations that will aid the development of new measurement 183 methodologies intended for today's IP networks. Specifically, this 184 memo describes useful stream parameters in addition to the 185 information in Section 11.1 of [RFC2330] and described in [RFC3432] 186 for periodic streams. 188 The memo also provides new considerations to update the criteria for 189 metrics in section 4 of [RFC2330], the measurement methodology in 190 section 6.2 of [RFC2330], and other topics related to the quality of 191 metrics and methods (see section 4). 193 Other topics in [RFC2330] which might be updated or augmented are 194 deferred to future work. This includes the topics of passive and 195 various forms of of hybrid active/passive measurements. 197 3. New or Revised Stream Parameters 199 There are several areas where measurement methodology definition and 200 test result interpretation will benefit from an increased 201 understanding of the stream characteristics and the (possibly 202 unknown) network condition that influence the measured metrics. 204 1. Network treatment depends on the fullest extent on the "packet of 205 Type-P" definition in [RFC2330], and has for some time. 207 * State is often maintained on the per-flow basis at various 208 points in the path, where "flows" are determined by IP and 209 other layers. Significant treatment differences occur with 210 the simplest of Type-P parameters: packet length. Use of 211 multiple lengths is RECOMMENDED. 213 * Payload content optimization (compression or format 214 conversion) in intermediate segments. This breaks the 215 convention of payload correspondence when correlating 216 measurements made at different points in a path. 218 2. Packet history (instantaneous or recent test rate or inactivity, 219 also for non-test traffic) profoundly influences measured 220 performance, in addition to all the Type-P parameters described 221 in [RFC2330]. 223 3. Access technology may change during testing. A range of transfer 224 capacities and access methods may be encountered during a test 225 session. When different interfaces are used, the host seeking 226 access will be aware of the technology change which 227 differentiates this form of path change from other changes in 228 network state. Section 14 of [RFC2330] treats the possibility 229 that a host may have more than one attachment to the network, and 230 also that assessment of the measurement path (route) is valid for 231 some length of time (in Section 5 and Section 7 of [RFC2330]). 232 Here we combine these two considerations under the assumption 233 that changes may be more frequent and possibly have greater 234 consequences on performance metrics. 236 4. Paths including links or nodes with time-slotted service 237 opportunities represent several challenges to measurement (when 238 service time period is appreciable): 240 * Random/unbiased sampling is not possible beyond one such link 241 in the path. 243 * The above encourages a segmented approach to end to end 244 measurement, as described in [RFC6049] for Network 245 Characterization (as defined in [RFC6703]) to understand the 246 full range of delay and delay variation on the path. 247 Alternatively, if application performance estimation is the 248 goal (also defined in [RFC6703]), then a stream with un-biased 249 or known-bias properties [RFC3432] may be sufficient. 251 * Multi-modal delay variation makes central statistics 252 unimportant, others must be used instead. 254 Each of these topics is treated in detail below. 256 3.1. Test Packet Type-P 258 We recommend two Type-P parameters to be added to the factors which 259 have impact on path performance measurements, namely packet length 260 and payload type. Carefully choosing these parameters can improve 261 measurement methodologies in their continuity and repeatability when 262 deployed in reactive paths. 264 3.1.1. Multiple Test Packet Lengths 266 Many instances of network characterization using IPPM metrics have 267 relied on a single test packet length. When testing to assess 268 application performance or an aggregate of traffic, benchmarking 269 methods have used a range of fixed lengths and frequently augmented 270 fixed size tests with a mixture of sizes, or IMIX as described in 271 [RFC6985]. 273 Test packet length influences delay measurements, in that the IPPM 274 one-way delay metric [RFC2679] includes serialization time in its 275 first-bit to last bit time stamping requirements. However, different 276 sizes can have a larger influence on link delay and link delay 277 variation than serialization would explain alone. This effect can be 278 non-linear and change the instantaneous network performance when a 279 different size is used, or the performance of packets following the 280 size change. 282 Repeatability is a main measurement methodology goal as stated in 283 section 6.2 of [RFC2330]. To eliminate packet length as a potential 284 measurement uncertainty factor, successive measurements must use 285 identical traffic patterns. In practice a combination of random 286 payload and random start time can yield representative results as 287 illustrated in [IRR]. 289 3.1.2. Test Packet Payload Content Optimization 291 The aim for efficient network resource use has resulted in deployment 292 of server-only or client-server lossless or lossy payload compression 293 techniques on some links or paths. These optimizers attempt to 294 compress high-volume traffic in order to reduce network load. Files 295 are analyzed by application-layer parsers, and parts (like comments) 296 might be dropped. Although typically acting on HTTP or JPEG files, 297 compression might affect measurement packets, too. In particular, 298 measurement packets are qualified for efficient compression when they 299 use standard plain-text payload. 301 IPPM-conforming measurements should add packet payload content as a 302 Type-P parameter which can help to improve measurement determinism. 303 Some packet payloads are more susceptible to compression than others, 304 but optimizers in the measurement path can be out ruled by using 305 incompressible packet payload. This payload content could be either 306 generated by a random device or by using part of a compressed file 307 (e.g., a part of a ZIP compressed archive). 309 Optimization can go beyond the scope of one single data- or 310 measurement stream. Many more client- or network-centric 311 optimization technologies have been proposed or standardized so far, 312 including Robust Header Compression (ROHC) and Voice over IP 313 aggregation as presented for instance in [EEAW]. The trend towards 314 optimization being ubiquitous, many more of these technologies will 315 follow. As general observation, the more concurrent flows an 316 intermediate host treats and the longer the paths shared by flows 317 are, the higher becomes the incentive of hosts to aggregate flows 318 belonging to distinct sources. Measurements should consider this 319 potential additional source of uncertainty with respect to 320 repeatability. Aggregation of flows in networking devices can, for 321 instance, result in reciprocal timing and performance influence of 322 these flows which may exceed typical reciprocical queueing effects by 323 orders of magnitude. 325 3.2. Packet History 327 Recent packet history and instantaneous data rate influence 328 measurement results for reactive links supporting on-demand capacity 329 allocation. Measurement uncertainty may be reduced by knowledge of 330 measurement packet history and total host load. Additionally, small 331 changes in history, e.g., because of lost packets along the path, can 332 be the cause of large performance variations. 334 For instance, delay in reactive 3G networks like High Speed Packet 335 Access (HSPA) depends to a large extent on the test traffic data 336 rate. The reactive resource allocation strategy in these networks 337 affects the uplink direction in particular. Small changes in data 338 rate can be the reason of more than 200% increase in delay, depending 339 on the specific packet size. A detailed theoretical and practical 340 analysis of RRC link transitions, which can cause such behavior in 341 Universal Mobile Terrestrial System (UMTS) networks, is presented, 342 e.g., in [RRC]. 344 3.3. Access Technology Change 346 [RFC2330] discussed the scenario of multi-homed hosts. If hosts 347 become aware of access technology changes (e.g., because of IP 348 address changes or lower layer information) and make this information 349 available, measurement methodologies can use this information to 350 improve measurement representativeness and relevance. 352 However, today's various access network technologies can present the 353 same physical interface to the host. A host may or may not become 354 aware when its access technology changes on such an interface. 355 Measurements for paths which support on-demand capacity allocation 356 are therefore challenging, in that it is difficult to differentiate 357 between access technology changes (e.g., because of mobility) and 358 reactive path behavior (e.g., because of data rate change). 360 3.4. Time-Slotted Randomness Cancellation 362 Time-Slotted operation of path entities - interfaces, routers or 363 links - in a network path is a particular challenge for measurements, 364 especially if the time slot period is substantial. The central 365 observation as an extension to Poisson stream sampling in [RFC2330] 366 is that the first such time-slotted component cancels unbiased 367 measurement stream sampling. In the worst case, time-slotted 368 operation converts an unbiased, random measurement packet stream into 369 a periodic packet stream. Being heavily biased, these packets may 370 interact with periodic behavior of subsequent time-slotted network 371 entities[TSRC]. 373 Time-slotted randomness cancellation (TSRC) sources can be found in 374 virtually any system, network component or path, their impact on 375 measurements being a matter of the order of magnitude when compared 376 to the metric under observation. Examples of TSRC sources include 377 but are not limited to system clock resolution, operating system 378 ticks, time-slotted component or network operation, etc. The amount 379 of measurement bias is determined by the particular measurement 380 stream, relative offset between allocated time-slots in subsequent 381 path entities, delay variation in these paths, and other sources of 382 variation. Measurement results might change over time, depending on 383 how accurately the sending host, receiving host, and time-slotted 384 components in the measurement path are synchronized to each other and 385 to global time. If path segments maintain flow state, flow parameter 386 change or flow re-allocations can cause substantial variation in 387 measurement results. 389 Practical measurements confirm that such interference limits delay 390 measurement variation to a sub-set of theoretical value range. 391 Measurement samples for such cases can aggregate on artificial 392 limits, generating multi-modal distributions as demonstrated in 393 [IRR]. In this context, the desirable measurement sample statistics 394 differentiate between multi-modal delay distributions caused by 395 reactive path behavior and the ones due to time-slotted interference. 397 Measurement methodology selection for time-slotted paths depends to a 398 large extent on the respective viewpoint. End-to-end metrics can 399 provide accurate measurement results for short-term sessions and low 400 likelihood of flow state modifications. Applications or services 401 which aim at approximating path performance for a short time interval 402 (in the order of minutes) and expect stable path conditions should 403 therefore prefer end-to-end metrics. Here stable path conditions 404 refer to any kind of global knowledge concerning measurement path 405 flow state and flow parameters. 407 However, if long-term forecast of time-slotted path performance is 408 the main measurement goal, a segmented approach relying on 409 measurement of sub-path metrics is preferred. Re-generating unbiased 410 measurement traffic at any hop can help to reveal the true range of 411 path performance for all path segments. 413 4. Quality of Metrics and Methodologies 415 [RFC6808] proposes repeatability and continuity as one of the metric 416 and methodology properties to infer on measurement quality. 417 Depending mainly on the set of controlled measurement parameters, 418 measurements repeated for a specific network path using a specific 419 methodology may or may not yield repeatable results. Challenging 420 measurement scenarios for adequate parameter control include 421 wireless, reactive, or time-slotted networks as discussed earlier in 422 this document. This section presents an expanded definition of 423 "repeatability" beyond the definition in [RFC2330] and an expanded 424 examination of the [RFC2330] concept of "continuity" and its limited 425 applicability. 427 4.1. Repeatability 429 [RFC2330] defines repeatability in a general way: 431 "A methodology for a metric should have the property that it is 432 repeatable: if the methodology is used multiple times under identical 433 conditions, the same measurements should result in the same 434 measurements." 436 The challenge is to develop this definition further, such that it 437 becomes an objective measurable criterion (and does not depend on the 438 concept of continuity discussed below). Fortunately, this topic has 439 been treated in other IPPM work. In BCP 176 [RFC6576], the criteria 440 of equivalent results was agreed as the surrogate for 441 interoperability when assessing metric RFCs for standards track 442 advancement. The criteria of equivalence were expressed as objective 443 statistical requirements for comparison across same implementations 444 and independent implementations in the test plans specific to each 445 RFC evaluated ([RFC2679] in the test plan of [RFC6808]). 447 The tests of [RFC6808] rely on nearly identical conditions to be 448 present for analysis, but accept that these conditions cannot be 449 exactly identical in the production network paths used. The test 450 plans allow some correction factors to be applied (some statistical 451 tests are hyper-sensitive to differences in the mean of 452 distributions), and recognize the original findings of [RFC2330] 453 regarding excess sample sizes. 455 One way to view the reliance on identical conditions is to view it as 456 a challenge: how few parameters and path conditions need to be 457 controlled and still produce repeatable methods/measurements? 459 Although the [RFC6808] test plan documented numerical criteria for 460 equivalence, we cannot specify the exact numerical criteria for 461 repeatability *in general*. The process in the BCP [RFC6576] and 462 statistics in [RFC6808] have been used successfully, and the 463 numerical criteria to declare a metric repeatable should be agreed by 464 all interested parties prior to measurement. 466 We revise the definition slightly, as follows: 468 "A methodology for a metric should have the property that it is 469 repeatable: if the methodology is used multiple times under identical 470 conditions, the methods should produce equivalent measurement 471 results." 473 4.2. Continuity 475 In the original framework [RFC2330], the concept of continuity was 476 introduced to provide a relaxed criteria for judging repeatability, 477 and was described in section 6.2 of [RFC2330] as follows: 479 "...a methodology for a given metric exhibits continuity if, for 480 small variations in conditions, it results in small variations in the 481 resulting measurements." 483 Although there are conditions where metrics may exhibit continuity, 484 there are others where this criteria would fail for both user traffic 485 and active measurement traffic. Consider link fragmentation, and the 486 non-linear increase in delay when we increase packet size just beyond 487 the limit of a single fragment. An active measurement packet would 488 see the same delay increase when exceeding the fragment size. 490 The Bulk Transfer Capacity (BTC) [RFC3148] gives another example at 491 bottom of page 2: 493 "There is also evidence that most TCP implementations exhibit non- 494 linear performance over some portion of their operating region. It 495 is possible to construct simple simulation examples where incremental 496 improvements to a path (such as raising the link data rate) results 497 in lower overall TCP throughput (or BTC) [Mat98]." 499 Clearly, the time-slotted network elements described in section 3.4 500 above also qualifies as a new exception to the ideal of continuity. 501 Therefore, we deprecate continuity as an alternate criterion on 502 metrics, and prefer the more exact evaluation of repeatability 503 instead. 505 4.3. Actionable 507 The IP Performance Metrics Framework [RFC2330] includes usefulness as 508 a metric criterion: 510 "...The metrics must be useful to users and providers in 511 understanding the performance they experience or provide...". 513 When considering measurements as part of a maintenance process, 514 evaluation of measurement results for a path under observation can 515 draw attention to potential performance problems "somewhere" on the 516 path. Anomaly detection is therefore an important phase and first 517 step which already satisfies the usefulness criterion for many 518 metrics. 520 This concept of usefulness can be extended, becoming a sub-set of 521 what we refer to as "actionable" criterion in the following. Central 522 to maintenance is the isolation of the root cause of reported 523 anomalies down to a specific sub-path, link or host, and metrics 524 should support this second step as well. While detection of path 525 anomaly may be the result of an on-going monitoring process, the 526 second step of cause isolation consists of specific, directed on- 527 demand measurements on components and sub-paths. Metrics must 528 support users in this directed search, becoming actionable: 530 Metrics must enable users and operators to understand path 531 performance and SHOULD help to direct corrective actions when 532 warranted, based on the measurement results. 534 Besides characterizing metrics, usefulness and actionable properties 535 are also applicable to methodologies and measurements. 537 4.4. Conservative 539 [RFC2330] adopts the term "conservative" for measurement 540 methodologies for which: 542 "... the act of measurement does not modify, or only slightly 543 modifies, the value of the performance metric the methodology 544 attempts to measure." 546 It should be noted that this definition of "conservative" in the 547 sense of [RFC2330] depends to a large extent on the measurement 548 path's technology and characteristics. In particular, when deployed 549 on reactive paths, sub-paths, links or hosts conforming to the 550 definition in Section 1.1 of this document, measurement packets can 551 originate capacity (re)allocations. In addition, small measurement 552 flow variations can result in other users on the same path perceiving 553 significant variations in measurement results. 555 4.5. Spatial and Temporal Composition 557 Concepts related to temporal and spatial composition of metrics in 558 Section 9 of [RFC2330] have been extended in [RFC5835]. [RFC5835] 559 defines multiple new types of metrics, including Spatial Composition, 560 Temporal Aggregation, and Spatial Aggregation. So far, only the 561 metrics for Spatial Composition have been standardized [RFC6049], 562 providing the ability to estimate the performance of a complete path 563 from subpath metrics. Spatial Composition aligns with the finding of 564 [TSRC], that unbiased sampling is not possible beyond the first time- 565 slotted link within a measurement path. In cases where measurement 566 of subpaths is not feasible, restoring randomness of measurement 567 samples when necessary is recommended as presented in [TSRC]. 569 4.6. Poisson Sampling 571 Section 11.1.1 of [RFC2330] describes Poisson sampling, where the 572 inter-packet send times have a Poisson distribution. A path element 573 with reactive behavior sensitive to flow inactivity could change 574 state if the random inter-packet time is too long. It is recommended 575 to truncate the tail of Poisson distribution to avoid reactive 576 element state changes. Truncation has been used without issue to 577 ensure that minimum sample sizes can be attained in a fixed test 578 interval. 580 5. Conclusions 582 Safeguarding repeatability as a key property of measurement 583 methodologies is highly challenging and sometimes impossible in 584 reactive paths. Measurements in paths with demand-driven allocation 585 strategies must use a prototypical application packet stream to infer 586 a specific application's performance. Measurement repetition with 587 unbiased network and flow states (e.g., by rebooting measurement 588 hosts) can help to avoid interference with periodic network behavior, 589 randomness being a mandatory feature for avoiding correlation with 590 network timing. Inferring the path performance between one 591 measurement session or packet stream and other streams with alternate 592 characteristics is generally discouraged with reactive paths because 593 of the huge set of global parameters which have influence on 594 instantaneous path performance. 596 6. Security Considerations 598 The security considerations that apply to any active measurement of 599 live paths are relevant here as well. See [RFC4656] and [RFC5357]. 601 7. IANA Considerations 603 This memo makes no requests of IANA. 605 8. Acknowledgements 607 The authors thank Rudiger Geib, Matt Mathis and Konstantinos 608 Pentikousis for their helpful comments on this memo, and Ann Cerveny 609 for her editorial review and comments that helped to improve 610 readability overall. 612 9. References 614 9.1. Normative References 616 [RFC2026] Bradner, S., "The Internet Standards Process -- Revision 617 3", BCP 9, RFC 2026, October 1996. 619 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 620 Requirement Levels", BCP 14, RFC 2119, March 1997. 622 [RFC2330] Paxson, V., Almes, G., Mahdavi, J., and M. Mathis, 623 "Framework for IP Performance Metrics", RFC 2330, May 624 1998. 626 [RFC2679] Almes, G., Kalidindi, S., and M. Zekauskas, "A One-way 627 Delay Metric for IPPM", RFC 2679, September 1999. 629 [RFC2680] Almes, G., Kalidindi, S., and M. Zekauskas, "A One-way 630 Packet Loss Metric for IPPM", RFC 2680, September 1999. 632 [RFC3432] Raisanen, V., Grotefeld, G., and A. Morton, "Network 633 performance measurement with periodic streams", RFC 3432, 634 November 2002. 636 [RFC4656] Shalunov, S., Teitelbaum, B., Karp, A., Boote, J., and M. 637 Zekauskas, "A One-way Active Measurement Protocol 638 (OWAMP)", RFC 4656, September 2006. 640 [RFC5357] Hedayat, K., Krzanowski, R., Morton, A., Yum, K., and J. 641 Babiarz, "A Two-Way Active Measurement Protocol (TWAMP)", 642 RFC 5357, October 2008. 644 [RFC5657] Dusseault, L. and R. Sparks, "Guidance on Interoperation 645 and Implementation Reports for Advancement to Draft 646 Standard", BCP 9, RFC 5657, September 2009. 648 [RFC5835] Morton, A. and S. Van den Berghe, "Framework for Metric 649 Composition", RFC 5835, April 2010. 651 [RFC6049] Morton, A. and E. Stephan, "Spatial Composition of 652 Metrics", RFC 6049, January 2011. 654 [RFC6576] Geib, R., Morton, A., Fardid, R., and A. Steinmitz, "IP 655 Performance Metrics (IPPM) Standard Advancement Testing", 656 BCP 176, RFC 6576, March 2012. 658 [RFC6703] Morton, A., Ramachandran, G., and G. Maguluri, "Reporting 659 IP Network Performance Metrics: Different Points of View", 660 RFC 6703, August 2012. 662 9.2. Informative References 664 [EEAW] Pentikousis, K., Piri, E., Pinola, J., Fitzek, F., 665 Nissilae, T., and I. Harjula, "Empirical Evaluation of 666 VoIP Aggregation over a Fixed WiMAX Testbed", Proceedings 667 of the 4th International Conference on Testbeds and 668 research infrastructures for the development of networks 669 and communities (TridentCom '08) 670 http://dl.acm.org/citation.cfm?id=1390599, March 2008. 672 [IBD] Fabini, J., Karner, W., Wallentin, L., and T. Baumgartner, 673 "The Illusion of Being Deterministic - Application-Level 674 Considerations on Delay in 3G HSPA Networks", Lecture 675 Notes in Computer Science, Springer, Volume 5550, 2009, pp 676 301-312 , May 2009. 678 [IRR] Fabini, J., Wallentin, L., and P. Reichl, "The Importance 679 of Being Really Random: Methodological Aspects of IP-Layer 680 2G and 3G Network Delay Assessment", ICC'09 Proceedings of 681 the 2009 IEEE International Conference on Communications, 682 doi: 10.1109/ICC.2009.5199514, June 2009. 684 [RFC3148] Mathis, M. and M. Allman, "A Framework for Defining 685 Empirical Bulk Transfer Capacity Metrics", RFC 3148, July 686 2001. 688 [RFC6808] Ciavattone, L., Geib, R., Morton, A., and M. Wieser, "Test 689 Plan and Results Supporting Advancement of RFC 2679 on the 690 Standards Track", RFC 6808, December 2012. 692 [RFC6985] Morton, A., "IMIX Genome: Specification of Variable Packet 693 Sizes for Additional Testing", RFC 6985, July 2013. 695 [RRC] Peraelae, P., Barbuzzi, A., Boggia, G., and K. 696 Pentikousis, "Theory and Practice of RRC State Transitions 697 in UMTS Networks", IEEE Globecom 2009 Workshops doi: 698 10.1109/GLOCOMW.2009.5360763, November 2009. 700 [TSRC] Fabini, J. and M. Abmayer, "Delay Measurement Methodology 701 Revisited: Time-slotted Randomness Cancellation", IEEE 702 Transactions on Instrumentation and Measurement 703 doi:10.1109/TIM.2013.2263914, October 2013. 705 Authors' Addresses 706 Joachim Fabini 707 Vienna University of Technology 708 Gusshausstrasse 25/E389 709 Vienna 1040 710 Austria 712 Phone: +43 1 58801 38813 713 Fax: +43 1 58801 38898 714 Email: Joachim.Fabini@tuwien.ac.at 715 URI: http://www.tc.tuwien.ac.at/about-us/staff/joachim-fabini/ 717 Al Morton 718 AT&T Labs 719 200 Laurel Avenue South 720 Middletown, NJ 07748 721 USA 723 Phone: +1 732 420 1571 724 Fax: +1 732 368 1192 725 Email: acmorton@att.com 726 URI: http://home.comcast.net/~acmacm/