idnits 2.17.1 draft-ietf-bmwg-b2b-frame-04.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 RFC2544, but the abstract doesn't seem to directly say this. It does mention RFC2544 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 seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords -- however, there's a paragraph with a matching beginning. Boilerplate error? (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). (Using the creation date from RFC2544, updated by this document, for RFC5378 checks: 1999-03-01) -- 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 (December 18, 2020) is 1215 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- -- Obsolete informational reference (is this intentional?): RFC 1944 (Obsoleted by RFC 2544) Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group A. Morton 3 Internet-Draft AT&T Labs 4 Updates: 2544 (if approved) December 18, 2020 5 Intended status: Informational 6 Expires: June 21, 2021 8 Updates for the Back-to-back Frame Benchmark in RFC 2544 9 draft-ietf-bmwg-b2b-frame-04 11 Abstract 13 Fundamental Benchmarking Methodologies for Network Interconnect 14 Devices of interest to the IETF are defined in RFC 2544. This memo 15 updates the procedures of the test to measure the Back-to-back frames 16 Benchmark of RFC 2544, based on further experience. 18 This memo updates Section 26.4 of RFC 2544. 20 Requirements Language 22 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 23 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 24 "OPTIONAL" in this document are to be interpreted as described in BCP 25 14[RFC2119] [RFC8174] when, and only when, they appear in all 26 capitals, as shown here. 28 Status of This Memo 30 This Internet-Draft is submitted in full conformance with the 31 provisions of BCP 78 and BCP 79. 33 Internet-Drafts are working documents of the Internet Engineering 34 Task Force (IETF). Note that other groups may also distribute 35 working documents as Internet-Drafts. The list of current Internet- 36 Drafts is at https://datatracker.ietf.org/drafts/current/. 38 Internet-Drafts are draft documents valid for a maximum of six months 39 and may be updated, replaced, or obsoleted by other documents at any 40 time. It is inappropriate to use Internet-Drafts as reference 41 material or to cite them other than as "work in progress." 43 This Internet-Draft will expire on June 21, 2021. 45 Copyright Notice 47 Copyright (c) 2020 IETF Trust and the persons identified as the 48 document authors. All rights reserved. 50 This document is subject to BCP 78 and the IETF Trust's Legal 51 Provisions Relating to IETF Documents 52 (https://trustee.ietf.org/license-info) in effect on the date of 53 publication of this document. Please review these documents 54 carefully, as they describe your rights and restrictions with respect 55 to this document. Code Components extracted from this document must 56 include Simplified BSD License text as described in Section 4.e of 57 the Trust Legal Provisions and are provided without warranty as 58 described in the Simplified BSD License. 60 Table of Contents 62 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 63 2. Scope and Goals . . . . . . . . . . . . . . . . . . . . . . . 3 64 3. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 4 65 4. Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . 6 66 5. Back-to-back Frames . . . . . . . . . . . . . . . . . . . . . 7 67 5.1. Preparing the list of Frame sizes . . . . . . . . . . . . 7 68 5.2. Test for a Single Frame Size . . . . . . . . . . . . . . 8 69 5.3. Test Repetition and Benchmark . . . . . . . . . . . . . . 9 70 5.4. Benchmark Calculations . . . . . . . . . . . . . . . . . 9 71 6. Reporting . . . . . . . . . . . . . . . . . . . . . . . . . . 11 72 7. Security Considerations . . . . . . . . . . . . . . . . . . . 12 73 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12 74 9. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . 12 75 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 13 76 10.1. Normative References . . . . . . . . . . . . . . . . . . 13 77 10.2. Informative References . . . . . . . . . . . . . . . . . 13 78 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 15 80 1. Introduction 82 The IETF's fundamental Benchmarking Methodologies are defined in 83 [RFC2544], supported by the terms and definitions in [RFC1242], and 84 [RFC2544] actually obsoletes an earlier specification, [RFC1944]. 85 Over time, the benchmarking community has updated [RFC2544] several 86 times, including the Device Reset Benchmark [RFC6201], and the 87 important Applicability Statement [RFC6815] concerning use outside 88 the Isolated Test Environment (ITE) required for accurate 89 benchmarking. Other specifications implicitly update [RFC2544], such 90 as the IPv6 Benchmarking Methodologies in [RFC5180]. 92 Recent testing experience with the Back-to-back Frame test and 93 Benchmark in Section 26.4 of [RFC2544] indicates that an update is 94 warranted [OPNFV-2017] [VSPERF-b2b]. In particular, analysis of the 95 results indicates that buffer size matters when compensating for 96 interruptions of software packet processing, and this finding 97 increases the importance of the Back-to-back frame characterization 98 described here. This memo describes additional rationale and 99 provides the updated method. 101 [RFC2544] (which obsoletes [RFC1944]) provides its own Requirements 102 Language consistent with [RFC2119], since [RFC1944] pre-dates 103 [RFC2119] and all three memos share common authorship. 104 Today,[RFC8174] clarifies the usage of Requirements Language, so the 105 requirements presented in this memo are expressed in [RFC8174] terms, 106 and intended for those performing/reporting laboratory tests to 107 improve clarity and repeatability, and for those designing devices 108 that facilitate these tests. 110 2. Scope and Goals 112 The scope of this memo is to define an updated method to 113 unambiguously perform tests, measure the benchmark(s), and report the 114 results for Back-to-back Frames (presently described in Section 26.4 115 of [RFC2544]). 117 The goal is to provide more efficient test procedures where possible, 118 and to expand reporting with additional interpretation of the 119 results. The tests described in this memo address the cases in which 120 the maximum frame rate of a single ingress port cannot be transferred 121 loss-free to an egress port (for some frame sizes of interest). 123 [RFC2544] Benchmarks rely on test conditions with constant frame 124 sizes, with the goal of understanding what network device capability 125 has been tested. Tests with the smallest size stress the header 126 processing capacity, and tests with the largest size stress the 127 overall bit processing capacity. Tests with sizes in-between may 128 determine the transition between these two capacities. However, 129 conditions simultaneously sending a mixture of Internet frame sizes 130 (IMIX), such as those described in [RFC6985], MUST NOT be used in 131 Back-to-back Frame testing. 133 Section 3 of [RFC8239] describes buffer size testing for physical 134 networking devices in a data center. The [RFC8239] methods measure 135 buffer latency directly with traffic on multiple ingress ports that 136 overload an egress port on the Device Under Test (DUT) and are not 137 subject to the revised calculations presented in this memo. 138 Likewise, the methods of [RFC8239] SHOULD be used for test cases 139 where the egress port buffer is the known point of overload. 141 3. Motivation 143 Section 3.1 of [RFC1242] describes the rationale for the Back-to-back 144 Frames Benchmark. To summarize, there are several reasons that 145 devices on a network produce bursts of frames at the minimum allowed 146 spacing; and it is, therefore, worthwhile to understand the Device 147 Under Test (DUT) limit on the length of such bursts in practice. 148 Also, [RFC1242] states: 150 "Tests of this parameter are intended to determine the extent 151 of data buffering in the device." 153 After this test was defined, there have been occasional discussions 154 of the stability and repeatability of the results, both over time and 155 across labs. Fortunately, the Open Platform for Network Function 156 Virtualization (OPNFV) VSPERF project's Continuous Integration (CI) 157 [VSPERF-CI] testing routinely repeats Back-to-back Frame tests to 158 verify that test functionality has been maintained through 159 development of the test control programs. These tests were used as a 160 basis to evaluate stability and repeatability, even across lab set- 161 ups when the test platform was migrated to new DUT hardware at the 162 end of 2016. 164 When the VSPERF CI results were examined [VSPERF-b2b], several 165 aspects of the results were considered notable: 167 1. Back-to-back Frame Benchmark was very consistent for some fixed 168 frame sizes, and somewhat variable for other frame sizes. 170 2. The number of Back-to-back Frames with zero loss reported for 171 large frame sizes was unexpectedly long (translating to 30 172 seconds of buffer time), and no explanation or measurement limit 173 condition was indicated. It was important that the buffering 174 time calculations were part of the referenced testing and 175 analysis[VSPERF-b2b], because the calculated buffer times of 30 176 seconds for some frame sizes were clearly wrong or highly 177 suspect. On the other hand, a result expressed only as a large 178 number of Back-to-back Frames does not permit such an easy 179 comparison with reality. 181 3. Calculation of the extent of buffer time in the DUT helped to 182 explain the results observed with all frame sizes (for example, 183 tests with some frame sizes cannot exceed the frame header 184 processing rate of the DUT and thus no buffering occurs; 185 therefore, the results depended on the test equipment and not the 186 DUT). 188 4. It was found that a better estimate of the DUT buffer time could 189 be calculated using measurements of both the longest burst in 190 frames without loss and results from the Throughput tests 191 conducted according to Section 26.1 of [RFC2544]. It is apparent 192 that the DUT's frame processing rate empties the buffer during a 193 trial and tends to increase the "implied" buffer size estimate 194 (measured according to Section 26.4 of [RFC2544] because many 195 frames have departed the buffer when the burst of frames ends). 196 A calculation using the Throughput measurement can reveal a 197 "corrected" buffer size estimate. 199 Further, if the Throughput tests of Section 26.1 of [RFC2544] are 200 conducted as a prerequisite test, the number of frame sizes required 201 for Back-to-back Frame Benchmarking can be reduced to one or more of 202 the small frame sizes, or the results for large frame sizes can be 203 noted as invalid in the results if tested anyway (these are the 204 larger frame sizes for which the back-to-back frame rate cannot 205 exceed the frame header processing rate of the DUT and little or no 206 buffering occurs). 208 The material below provides the details of the calculation to 209 estimate the actual buffer storage available in the DUT, using 210 results from the Throughput tests for each frame size, and the 211 maximum theoretical frame rate for the DUT links (which constrain the 212 minimum frame spacing). 214 In reality, there are many buffers and packet header processing steps 215 in a typical DUT. The simplified model used in these calculations 216 for the DUT includes a packet header processing function with limited 217 rate of operation, as shown below: 219 |------------ DUT --------| 220 Generator -> Ingress -> Buffer -> HeaderProc -> Egress -> Receiver 222 So, in the Back-to-back Frame testing: 224 1. The ingress burst arrives at Max Theoretical Frame Rate, and 225 initially the frames are buffered. 227 2. The packet header processing function (HeaderProc) operates at 228 the "Measured Throughput" (Section 26.1 of [RFC2544]), removing 229 frames from the buffer (this is the best approximation we have). 231 3. Frames that have been processed are clearly not in the buffer, so 232 the Corrected DUT buffer time equation (Section 5.4) estimates 233 and removes the frames that the DUT forwarded on egress during 234 the burst. We define buffer time as the number of frames 235 occupying the buffer divided by the Maximum Theoretical Frame 236 Rate (on ingress) for the frame size under test. 238 4. A helpful concept is the buffer filling rate, which is the 239 difference between the Max Theoretical Frame Rate (ingress) and 240 the Measured Throughput (HeaderProc on egress). If the actual 241 buffer size in frames was known, the time to fill the buffer 242 during a measurement can be calculated using the filling rate as 243 a check on measurements. However, the buffer in the model 244 represents many buffers of different sizes in the DUT data path. 246 Knowledge of approximate buffer storage size (in time or bytes) may 247 be useful to estimate whether frame losses will occur if DUT 248 forwarding is temporarily suspended in a production deployment, due 249 to an unexpected interruption of frame processing (an interruption of 250 duration greater than the estimated buffer would certainly cause lost 251 frames). In Section 5, the calculations for the correct buffer time 252 use the combination of offered load at Max Theoretical Frame Rate and 253 header processing speed at 100% of Measured Throughput. Other 254 combinations are possible, such as changing the percent of measured 255 Throughput to account for other processes reducing the header 256 processing rate. 258 The presentation of OPNFV VSPERF evaluation and development of 259 enhanced search algorithms [VSPERF-BSLV] was discussed at IETF-102. 260 The enhancements are intended to compensate for transient interrupts 261 that may cause loss at near-Throughput levels of offered load. 262 Subsequent analysis of the results indicates that buffers within the 263 DUT can compensate for some interrupts, and this finding increases 264 the importance of the Back-to-back frame characterization described 265 here. 267 4. Prerequisites 269 The Test Setup MUST be consistent with Figure 1 of [RFC2544], or 270 Figure 2 when the tester's sender and receiver are different devices. 271 Other mandatory testing aspects described in [RFC2544] MUST be 272 included, unless explicitly modified in the next section. 274 The ingress and egress link speeds and link layer protocols MUST be 275 specified and used to compute the maximum theoretical frame rate when 276 respecting the minimum inter-frame gap. 278 The test results for the Throughput Benchmark conducted according to 279 Section 26.1 of [RFC2544] for all [RFC2544]-RECOMMENDED frame sizes 280 MUST be available to reduce the tested frame size list, or to note 281 invalid results for individual frame sizes (because the burst length 282 may be essentially infinite for large frame sizes). 284 Note that: 286 o the Throughput and the Back-to-back Frame measurement 287 configuration traffic characteristics (unidirectional or bi- 288 directional, and number of flows generated) MUST match. 290 o the Throughput measurement MUST be under zero-loss conditions, 291 according to Section 26.1 of [RFC2544]. 293 The Back-to-back Benchmark described in Section 3.1 of [RFC1242] MUST 294 be measured directly by the tester, where buffer size is inferred 295 from Back-to-back Frame bursts and associated packet loss 296 measurements. Therefore, sources of packet loss that are unrelated 297 to consistent evaluation of buffer size SHOULD be identified and 298 removed or mitigated. Example sources include: 300 o On-path active components that are external to the DUT 302 o Operating system environment interrupting DUT operation 304 o Shared resource contention between the DUT and other off-path 305 component(s) impacting DUT's behaviour, sometimes called the 306 "noisy neighbour" problem with virtualized network functions. 308 Mitigations applicable to some of the sources above are discussed in 309 Section 5.2, with the other measurement requirements described below 310 in Section 5. 312 5. Back-to-back Frames 314 Objective: To characterize the ability of a DUT to process back-to- 315 back frames as defined in [RFC1242]. 317 The Procedure follows. 319 5.1. Preparing the list of Frame sizes 321 From the list of RECOMMENDED frame sizes (Section 9 of [RFC2544]), 322 select the subset of frame sizes whose measured Throughput (during 323 prerequisite testing) was less than the Maximum Theoretical Frame 324 Rate of the DUT/test-set-up. These are the only frame sizes where it 325 is possible to produce a burst of frames that cause the DUT buffers 326 to fill and eventually overflow, producing one or more discarded 327 frames. 329 5.2. Test for a Single Frame Size 331 Each trial in the test requires the tester to send a burst of frames 332 (after idle time) with the minimum inter-frame gap, and to count the 333 corresponding frames forwarded by the DUT. 335 The duration of the trial includes three REQUIRED components: 337 1. The time to send the burst of frames (at the back-to-back rate), 338 determined by the search algorithm. 340 2. The time to receive the transferred burst of frames (at the 341 [RFC2544] Throughput rate), possibly truncated by buffer 342 overflow, and certainly including the latency of the DUT. 344 3. At least 2 seconds not overlapping the time to receive the burst 345 (2.), to ensure that DUT buffers have depleted. Longer times 346 MUST be used when conditions warrant, such as when buffer times 347 >2 seconds are measured or when burst sending times are >2 348 seconds, but care is needed since this time component directly 349 increases trial duration and many trials and tests comprise a 350 complete benchmarking study. 352 The upper search limit for the time to send each burst MUST be 353 configurable, to values as high as 30 seconds (buffer time results 354 reported at or near the configured upper limit are likely invalid, 355 and the test MUST be repeated with a higher search limit). 357 If all frames have been received, the tester increases the length of 358 the burst according to the search algorithm and performs another 359 trial. 361 If the received frame count is less than the number of frames in the 362 burst, then the limit of DUT processing and buffering may have been 363 exceeded, and the burst length is determined by the search algorithm 364 for the next trial (the burst length is typically reduced, but see 365 below). 367 Classic search algorithms have been adapted for use in benchmarking, 368 where the search requires discovery of a pair of outcomes, one with 369 no loss and another with loss, at load conditions within the 370 acceptable tolerance or accuracy. Conditions encountered when 371 benchmarking the Infrastructure for Network Function Virtualization 372 require algorithm enhancement. Fortunately, the adaptation of Binary 373 Search, and an enhanced Binary Search with Loss Verification have 374 been specified in clause 12.3 of [TST009]. These algorithms can 375 easily be used for Back-to-back Frame benchmarking by replacing the 376 Offered Load level with burst length in frames. [TST009] Annex B 377 describes the theory behind the enhanced Binary Search with Loss 378 Verification algorithm. 380 There is also promising work-in-progress that may prove useful in 381 Back-to-back Frame benchmarking. 382 [I-D.vpolak-mkonstan-bmwg-mlrsearch] and [I-D.vpolak-bmwg-plrsearch] 383 are two such examples. 385 Either the [TST009] Binary Search or Binary Search with Loss 386 Verification algorithms MUST be used, and input parameters to the 387 algorithm(s) MUST be reported. 389 The tester usually imposes a (configurable) minimum step size for 390 burst length, and the step size MUST be reported with the results (as 391 this influences the accuracy and variation of test results). 393 The original Section 26.4 of [RFC2544] definition is stated below: 395 The Back-to-back Frame value is the longest burst of frames that 396 the DUT can successfully process and buffer without frame loss, as 397 determined from the series of trials. 399 5.3. Test Repetition and Benchmark 401 On this topic, Section 26.4 of [RFC2544] requires: 403 The trial length MUST be at least 2 seconds and SHOULD be repeated 404 at least 50 times with the average of the recorded values being 405 reported. 407 Therefore, the Back-to-back Frame Benchmark is the average of burst 408 length values over repeated tests to determine the longest burst of 409 frames that the DUT can successfully process and buffer without frame 410 loss. Each of the repeated tests completes an independent search 411 process. 413 In this update, the test MUST be repeated N times (the number of 414 repetitions is now a variable that must be reported),for each frame 415 size in the subset list, and each Back-to-back Frame value made 416 available for further processing (below). 418 5.4. Benchmark Calculations 420 For each frame size, calculate the following summary statistics for 421 longest Back-to-back Frame values over the N tests: 423 o Average (Benchmark) 424 o Minimum 426 o Maximum 428 o Standard Deviation 430 Further, calculate the Implied DUT Buffer Time and the Corrected DUT 431 Buffer Time in seconds, as follows: 433 Implied DUT Buffer Time = 435 Average num of Back-to-back Frames / Max Theoretical Frame Rate 437 The formula above is simply expressing the burst of frames in units 438 of time. 440 The next step is to apply a correction factor that accounts for the 441 DUT's frame forwarding operation during the test (assuming the simple 442 model of the DUT composed of a buffer and a forwarding function, 443 described in Section 3). 445 Corrected DUT Buffer Time = 446 / \ 447 Implied DUT |Implied DUT Measured Throughput | 448 = Buffer Time - |Buffer Time * -------------------------- | 449 | Max Theoretical Frame Rate | 450 \ / 452 where: 454 1. The "Measured Throughput" is the [RFC2544] Throughput Benchmark 455 for the frame size tested, as augmented by methods including the 456 Binary Search with Loss Verification algorithm in [TST009] where 457 applicable, and MUST be expressed in frames per second in this 458 equation. 460 2. The "Max Theoretical Frame Rate" is a calculated value for the 461 interface speed and link layer technology used, and MUST be 462 expressed in frames per second in this equation. 464 The term on the far right in the formula for Corrected DUT Buffer 465 Time accounts for all the frames in the Burst that were transmitted 466 by the DUT *while the Burst of frames were sent in*. So, these frames 467 are not in the buffer and the buffer size is more accurately 468 estimated by excluding them. 470 6. Reporting 472 The back-to-back frame results SHOULD be reported in the format of a 473 table with a row for each of the tested frame sizes. There SHOULD be 474 columns for the frame size and for the resultant average frame count 475 for each type of data stream tested. 477 The number of tests Averaged for the Benchmark, N, MUST be reported. 479 The Minimum, Maximum, and Standard Deviation across all complete 480 tests SHOULD also be reported (they are referred to as 481 "Min,Max,StdDev" in the table below). 483 The Corrected DUT Buffer Time SHOULD also be reported. 485 If the tester operates using a limited maximum burst length in 486 frames, then this maximum length SHOULD be reported. 488 +--------------+----------------+----------------+------------------+ 489 | Frame Size, | Ave B2B | Min,Max,StdDev | Corrected Buff | 490 | octets | Length, frames | | Time, Sec | 491 +--------------+----------------+----------------+------------------+ 492 | 64 | 26000 | 25500,27000,20 | 0.00004 | 493 +--------------+----------------+----------------+------------------+ 495 Back-to-Back Frame Results 497 Static and configuration parameters (reported with the table above): 499 Number of test repetitions, N 501 Minimum Step Size (during searches), in frames. 503 If the tester has a specific (actual) frame rate of interest (less 504 than the Throughput rate), it is useful to estimate the buffer time 505 at that actual frame rate: 507 Actual Buffer Time = 508 Max Theoretical Frame Rate 509 = Corrected DUT Buffer Time * -------------------------- 510 Actual Frame Rate 512 and report this value, properly labeled. 514 7. Security Considerations 516 Benchmarking activities as described in this memo are limited to 517 technology characterization using controlled stimuli in a laboratory 518 environment, with dedicated address space and the other constraints 519 of[RFC2544]. 521 The benchmarking network topology will be an independent test setup 522 and MUST NOT be connected to devices that may forward the test 523 traffic into a production network, or misroute traffic to the test 524 management network. See [RFC6815]. 526 Further, benchmarking is performed on an "opaque-box" (a.k.a. 527 "black-box") basis, relying solely on measurements observable 528 external to the DUT/SUT. 530 The DUT developers are commonly independent from the personnel and 531 institutions conducting benchmarking studies. DUT developers might 532 have incentives to alter the performance of the DUT if the test 533 conditions can be detected. Special capabilities SHOULD NOT exist in 534 the DUT/SUT specifically for benchmarking purposes. Procedures 535 described in this document are not designed to detect such activity. 536 Additional testing outside of the scope of this document would be 537 needed and has been used successfully in the past to discover such 538 malpractices. 540 Any implications for network security arising from the DUT/SUT SHOULD 541 be identical in the lab and in production networks. 543 8. IANA Considerations 545 This memo makes no requests of IANA. 547 9. Acknowledgements 549 Thanks to Trevor Cooper, Sridhar Rao, and Martin Klozik of the VSPERF 550 project for many contributions to the early testing [VSPERF-b2b]. 551 Yoshiaki Itou has also investigated the topic, and made useful 552 suggestions. Maciek Konstantyowicz and Vratko Polak also provided 553 many comments and suggestions based on extensive integration testing 554 and resulting search algorithm proposals - the most up-to-date 555 feedback possible. Tim Carlin also provided comments and support for 556 the draft. Warren Kumari's review improved readability in several 557 key passages. David Black, Martin Duke, and Scott Bradner's comments 558 improved the clarity and configuration advice on trial duration. 559 Malisa Vucinic suggested additional text on DUT design cautions in 560 the Security Considerations section. 562 10. References 564 10.1. Normative References 566 [RFC1242] Bradner, S., "Benchmarking Terminology for Network 567 Interconnection Devices", RFC 1242, DOI 10.17487/RFC1242, 568 July 1991, . 570 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 571 Requirement Levels", BCP 14, RFC 2119, 572 DOI 10.17487/RFC2119, March 1997, 573 . 575 [RFC2544] Bradner, S. and J. McQuaid, "Benchmarking Methodology for 576 Network Interconnect Devices", RFC 2544, 577 DOI 10.17487/RFC2544, March 1999, 578 . 580 [RFC6985] Morton, A., "IMIX Genome: Specification of Variable Packet 581 Sizes for Additional Testing", RFC 6985, 582 DOI 10.17487/RFC6985, July 2013, 583 . 585 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 586 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 587 May 2017, . 589 [RFC8239] Avramov, L. and J. Rapp, "Data Center Benchmarking 590 Methodology", RFC 8239, DOI 10.17487/RFC8239, August 2017, 591 . 593 [TST009] Morton, A., "ETSI GS NFV-TST 009 V3.4.1 (2020-12), 594 "Network Functions Virtualisation (NFV) Release 3; 595 Testing; Specification of Networking Benchmarks and 596 Measurement Methods for NFVI"", December 2020, 597 . 600 10.2. Informative References 602 [I-D.vpolak-bmwg-plrsearch] 603 Konstantynowicz, M. and V. Polak, "Probabilistic Loss 604 Ratio Search for Packet Throughput (PLRsearch)", draft- 605 vpolak-bmwg-plrsearch-03 (work in progress), March 2020. 607 [I-D.vpolak-mkonstan-bmwg-mlrsearch] 608 Konstantynowicz, M. and V. Polak, "Multiple Loss Ratio 609 Search for Packet Throughput (MLRsearch)", draft-vpolak- 610 mkonstan-bmwg-mlrsearch-03 (work in progress), March 2020. 612 [OPNFV-2017] 613 Cooper, T., Morton, A., and S. Rao, "Dataplane 614 Performance, Capacity, and Benchmarking in OPNFV", June 615 2017, 616 . 619 [RFC1944] Bradner, S. and J. McQuaid, "Benchmarking Methodology for 620 Network Interconnect Devices", RFC 1944, 621 DOI 10.17487/RFC1944, May 1996, 622 . 624 [RFC5180] Popoviciu, C., Hamza, A., Van de Velde, G., and D. 625 Dugatkin, "IPv6 Benchmarking Methodology for Network 626 Interconnect Devices", RFC 5180, DOI 10.17487/RFC5180, May 627 2008, . 629 [RFC6201] Asati, R., Pignataro, C., Calabria, F., and C. Olvera, 630 "Device Reset Characterization", RFC 6201, 631 DOI 10.17487/RFC6201, March 2011, 632 . 634 [RFC6815] Bradner, S., Dubray, K., McQuaid, J., and A. Morton, 635 "Applicability Statement for RFC 2544: Use on Production 636 Networks Considered Harmful", RFC 6815, 637 DOI 10.17487/RFC6815, November 2012, 638 . 640 [VSPERF-b2b] 641 Morton, A., "Back2Back Testing Time Series (from CI)", 642 June 2017, . 646 [VSPERF-BSLV] 647 Morton, A. and S. Rao, "Evolution of Repeatability in 648 Benchmarking: Fraser Plugfest (Summary for IETF BMWG)", 649 July 2018, 650 . 654 [VSPERF-CI] 655 Tahhan, M., "OPNFV VSPERF CI", June 2019, 656 . 658 Author's Address 660 Al Morton 661 AT&T Labs 662 200 Laurel Avenue South 663 Middletown,, NJ 07748 664 USA 666 Phone: +1 732 420 1571 667 Fax: +1 732 368 1192 668 Email: acmorton@att.com