idnits 2.17.1 draft-ietf-tcpm-cubic-00.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 18, 2015) is 3235 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 4960 (Obsoleted by RFC 9260) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TCP Maintenance and Minor Extensions (TCPM) WG I. Rhee 3 Internet-Draft NCSU 4 Intended status: Informational L. Xu 5 Expires: December 20, 2015 UNL 6 S. Ha 7 NCSU 8 A. Zimmermann 9 L. Eggert 10 R. Scheffenegger 11 NetApp 12 June 18, 2015 14 CUBIC for Fast Long-Distance Networks 15 draft-ietf-tcpm-cubic-00 17 Abstract 19 CUBIC is an extension to the current TCP standards. The protocol 20 differs from the current TCP standards only in the congestion window 21 adjustment function in the sender side. In particular, it uses a 22 cubic function instead of a linear window increase of the current TCP 23 standards to improve scalability and stability under fast and long 24 distance networks. BIC-TCP, a predecessor of CUBIC, has been a 25 default TCP adopted by Linux since year 2005 and has already been 26 deployed globally and in use for several years by the Internet 27 community at large. CUBIC is using a similar window growth function 28 as BIC-TCP and is designed to be less aggressive and fairer to TCP in 29 bandwidth usage than BIC-TCP while maintaining the strengths of BIC- 30 TCP such as stability, window scalability and RTT fairness. Through 31 extensive testing in various Internet scenarios, we believe that 32 CUBIC is safe for deployment and testing in the global Internet. The 33 intent of this document is to provide the protocol specification of 34 CUBIC for a third party implementation and solicit the community 35 feedback through experimentation on the performance of CUBIC. 37 Status of This Memo 39 This Internet-Draft is submitted in full conformance with the 40 provisions of BCP 78 and BCP 79. 42 Internet-Drafts are working documents of the Internet Engineering 43 Task Force (IETF). Note that other groups may also distribute 44 working documents as Internet-Drafts. The list of current Internet- 45 Drafts is at http://datatracker.ietf.org/drafts/current/. 47 Internet-Drafts are draft documents valid for a maximum of six months 48 and may be updated, replaced, or obsoleted by other documents at any 49 time. It is inappropriate to use Internet-Drafts as reference 50 material or to cite them other than as "work in progress." 52 This Internet-Draft will expire on December 20, 2015. 54 Copyright Notice 56 Copyright (c) 2015 IETF Trust and the persons identified as the 57 document authors. All rights reserved. 59 This document is subject to BCP 78 and the IETF Trust's Legal 60 Provisions Relating to IETF Documents 61 (http://trustee.ietf.org/license-info) in effect on the date of 62 publication of this document. Please review these documents 63 carefully, as they describe your rights and restrictions with respect 64 to this document. Code Components extracted from this document must 65 include Simplified BSD License text as described in Section 4.e of 66 the Trust Legal Provisions and are provided without warranty as 67 described in the Simplified BSD License. 69 Table of Contents 71 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 72 2. Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 5 73 3. CUBIC Congestion Control . . . . . . . . . . . . . . . . . . 5 74 3.1. Window growth function . . . . . . . . . . . . . . . . . 5 75 3.2. TCP-friendly region . . . . . . . . . . . . . . . . . . . 6 76 3.3. Concave region . . . . . . . . . . . . . . . . . . . . . 6 77 3.4. Convex region . . . . . . . . . . . . . . . . . . . . . . 7 78 3.5. Multiplicative decrease . . . . . . . . . . . . . . . . . 7 79 3.6. Fast convergence . . . . . . . . . . . . . . . . . . . . 7 80 4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 8 81 4.1. Fairness to standard TCP . . . . . . . . . . . . . . . . 8 82 4.2. Using Spare Capacity . . . . . . . . . . . . . . . . . . 10 83 4.3. Difficult Environments . . . . . . . . . . . . . . . . . 10 84 4.4. Investigating a Range of Environments . . . . . . . . . . 11 85 4.5. Protection against Congestion Collapse . . . . . . . . . 11 86 4.6. Fairness within the Alternative Congestion Control 87 Algorithm. . . . . . . . . . . . . . . . . . . . . . . . 11 88 4.7. Performance with Misbehaving Nodes and Outside Attackers 11 89 4.8. Responses to Sudden or Transient Events . . . . . . . . . 11 90 4.9. Incremental Deployment . . . . . . . . . . . . . . . . . 11 91 5. Security Considerations . . . . . . . . . . . . . . . . . . . 11 92 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11 93 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 12 94 8. References . . . . . . . . . . . . . . . . . . . . . . . . . 12 95 8.1. Normative References . . . . . . . . . . . . . . . . . . 12 96 8.2. Informative References . . . . . . . . . . . . . . . . . 12 98 Appendix A. ToDo List . . . . . . . . . . . . . . . . . . . . . 13 99 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 13 101 1. Introduction 103 The low utilization problem of TCP in fast long-distance networks is 104 well documented in [K03][RFC3649]. This problem arises from a slow 105 increase of congestion window following a congestion event in a 106 network with a large bandwidth delay product (BDP). Our experience 107 [HKLRX06] indicates that this problem is frequently observed even in 108 the range of congestion window sizes over several hundreds of packets 109 (each packet is sized around 1000 bytes) especially under a network 110 path with over 100ms round-trip times (RTTs). This problem is 111 equally applicable to all Reno style TCP standards and their 112 variants, including TCP-RENO [RFC5681], TCP-NewReno [RFC6582], TCP- 113 SACK [RFC2018], SCTP [RFC4960], TFRC [RFC5348] that use the same 114 linear increase function for window growth, which we refer to 115 collectively as Standard TCP below. 117 CUBIC [HRX08] is a modification to the congestion control mechanism 118 of Standard TCP, in particular, to the window increase function of 119 Standard TCP senders, to remedy this problem. It uses a cubic 120 increase function in terms of the elapsed time from the last 121 congestion event. While most alternative algorithms to Standard TCP 122 uses a convex increase function where after a loss event, the window 123 increment is always increasing, CUBIC uses both the concave and 124 convex profiles of a cubic function for window increase. After a 125 window reduction following a loss event, it registers the window size 126 where it got the loss event as W_max and performs a multiplicative 127 decrease of congestion window and the regular fast recovery and 128 retransmit of Standard TCP. After it enters into congestion 129 avoidance from fast recovery, it starts to increase the window using 130 the concave profile of the cubic function. The cubic function is set 131 to have its plateau at W_max so the concave growth continues until 132 the window size becomes W_max. After that, the cubic function turns 133 into a convex profile and the convex window growth begins. This 134 style of window adjustment (concave and then convex) improves 135 protocol and network stability while maintaining high network 136 utilization [CEHRX07]. This is because the window size remains 137 almost constant, forming a plateau around W_max where network 138 utilization is deemed highest and under steady state, most window 139 size samples of CUBIC are close to W_max, thus promoting high network 140 utilization and protocol stability. Note that protocols with convex 141 increase functions have the maximum increments around W_max and 142 introduces a large number of packet bursts around the saturation 143 point of the network, likely causing frequent global loss 144 synchronizations. 146 Another notable feature of CUBIC is that its window increase rate is 147 mostly independent of RTT, and follows a (cubic) function of the 148 elapsed time since the last loss event. This feature promotes per- 149 flow fairness to Standard TCP as well as RTT-fairness. Note that 150 Standard TCP performs well under short RTT and small bandwidth (or 151 small BDP) networks. Only in a large long RTT and large bandwidth 152 (or large BDP) networks, it has the scalability problem. An 153 alternative protocol to Standard TCP designed to be friendly to 154 Standard TCP at a per-flow basis must operate must increase its 155 window much less aggressively in small BDP networks than in large BDP 156 networks. In CUBIC, its window growth rate is slowest around the 157 inflection point of the cubic function and this function does not 158 depend on RTT. In a smaller BDP network where Standard TCP flows are 159 working well, the absolute amount of the window decrease at a loss 160 event is always smaller because of the multiplicative decrease. 161 Therefore, in CUBIC, the starting window size after a loss event from 162 which the window starts to increase, is smaller in a smaller BDP 163 network, thus falling nearer to the plateau of the cubic function 164 where the growth rate is slowest. By setting appropriate values of 165 the cubic function parameters, CUBIC sets its growth rate always no 166 faster than Standard TCP around its inflection point. When the cubic 167 function grows slower than the window of Standard TCP, CUBIC simply 168 follows the window size of Standard TCP to ensure fairness to 169 Standard TCP in a small BDP network. We call this region where CUBIC 170 behaves like Standard TCP, the TCP-friendly region. 172 CUBIC maintains the same window growth rate independent of RTTs 173 outside of the TCP-friendly region, and flows with different RTTs 174 have the similar window sizes under steady state when they operate 175 outside the TCP-friendly region. This ensures CUBIC flows with 176 different RTTs to have their bandwidth shares linearly proportional 177 to the inverse of their RTT ratio (the longer RTT, the smaller the 178 share). This behavior is the same as that of Standard TCP under high 179 statistical multiplexing environments where packet losses are 180 independent of individual flow rates. However, under low statistical 181 multiplexing environments, the bandwidth share ratio of Standard TCP 182 flows with different RTTs is squarely proportional to the inverse of 183 their RTT ratio [XHR04]. CUBIC always ensures the linear ratio 184 independent of the levels of statistical multiplexing. This is an 185 improvement over Standard TCP. While there is no consensus on a 186 particular bandwidth share ratios of different RTT flows, we believe 187 that under wired Internet, use of the linear share notion seems more 188 reasonable than equal share or a higher order shares. HTCP [LS08] 189 currently uses the equal share. 191 CUBIC sets the multiplicative window decrease factor to 0.2 while 192 Standard TCP uses 0.5. While this improves the scalability of the 193 protocol, a side effect of this decision is slower convergence 194 especially under low statistical multiplexing environments. This 195 design choice is following the observation that the author of HSTCP 196 [RFC3649] has made along with other researchers (e.g., [GV02]): the 197 current Internet becomes more asynchronous with less frequent loss 198 synchronizations with high statistical multiplexing. Under this 199 environment, even strict MIMD can converge. CUBIC flows with the 200 same RTT always converge to the same share of bandwidth independent 201 of statistical multiplexing, thus achieving intra-protocol fairness. 202 We also find that under the environments with sufficient statistical 203 multiplexing, the convergence speed of CUBIC flows is reasonable. 205 In the ensuing sections, we provide the exact specification of CUBIC 206 and discuss the safety features of CUBIC following the guidelines 207 specified in [RFC5033]. 209 2. Conventions 211 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 212 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 213 document are to be interpreted as described in [RFC2119]. 215 3. CUBIC Congestion Control 217 3.1. Window growth function 219 CUBIC maintains the acknowledgment (ACK) clocking of Standard TCP by 220 increasing congestion window only at the reception of ACK. The 221 protocol does not make any change to the fast recovery and retransmit 222 of TCP-NewReno [RFC6582] and TCP-SACK [RFC2018]. During congestion 223 avoidance after fast recovery, CUBIC changes the window update 224 algorithm of Standard TCP. Suppose that W_max is the window size 225 before the window is reduced in the last fast retransmit and 226 recovery. 228 The window growth function of CUBIC uses the following function: 230 W(t) = C*(t-K)^3 + W_max (Eq. 1) 232 where C is a constant fixed to determine the aggressiveness of window 233 growth in high BDP networks, t is the elapsed time from the last 234 window reduction,and K is the time period that the above function 235 takes to increase W to W_max when there is no further loss event and 236 is calculated by using the following equation: 238 K = cubic_root(W_max*beta/C) (Eq. 2) 240 where beta is the multiplication decrease factor. We discuss how we 241 set C in the next Section in more details. 243 Upon receiving an ACK during congestion avoidance, CUBIC computes the 244 window growth rate during the next RTT period using Eq. 1. It sets 245 W(t+RTT) as the candidate target value of congestion window. Suppose 246 that the current window size is cwnd. Depending on the value of 247 cwnd, CUBIC runs in three different modes. First, if cwnd is less 248 than the window size that Standard TCP would reach at time t after 249 the last loss event, then CUBIC is in the TCP friendly region (we 250 describe below how to determine this window size of Standard TCP in 251 term of time t). Otherwise, if cwnd is less than W_max, then CUBIC 252 is the concave region, and if cwnd is larger than W_max, CUBIC is in 253 the convex region. Below, we describe the exact actions taken by 254 CUBIC in each region. 256 3.2. TCP-friendly region 258 When receiving an ACK in congestion avoidance, we first check whether 259 the protocol is in the TCP region or not. This is done as follows. 260 We can analyze the window size of Standard TCP in terms of the 261 elapsed time t. Using a simple analysis in [FHP00], we can analyze 262 the average window size of additive increase and multiplicative 263 decrease (AIMD) with an additive factor alpha and a multiplicative 264 factor beta to be the following function: 266 (alpha/2 * (2-beta)/beta * 1/p)^0.5 (Eq. 3) 268 By the same analysis, the average window size of Standard TCP with 269 alpha 1 and beta 0.5 is (3/2 *1/p)^0.5. Thus, for Eq. 3 to be the 270 same as that of Standard TCP, alpha must be equal to 3*beta/(2-beta). 271 As Standard TCP increases its window by alpha per RTT, we can get the 272 window size of Standard TCP in terms of the elapsed time t as 273 follows: 275 W_tcp(t) = W_max*(1-beta) + 3*beta/(2-beta)* t/RTT (Eq. 4) 277 If cwnd is less than W_tcp(t), then the protocol is in the TCP 278 friendly region and cwnd SHOULD be set to W_tcp(t) at each reception 279 of ACK. 281 3.3. Concave region 283 When receiving an ACK in congestion avoidance, if the protocol is not 284 in the TCP-friendly region and cwnd is less than W_max, then the 285 protocol is in the concave region. In this region, cwnd MUST be 286 incremented by (W(t+RTT) - cwnd)/cwnd. 288 3.4. Convex region 290 When the window size of CUBIC is larger than W_max, it passes the 291 plateau of the cubic function after which CUBIC follows the convex 292 profile of the cubic function. Since cwnd is larger than the 293 previous saturation point W_max, this indicates that the network 294 conditions might have been perturbed since the last loss event, 295 possibly implying more available bandwidth after some flow 296 departures. Since the Internet is highly asynchronous, some amount 297 of perturbation is always possible without causing a major change in 298 available bandwidth. In this phase, CUBIC is being very careful by 299 very slowly increasing its window size. The convex profile ensures 300 that the window increases very slowly at the beginning and gradually 301 increases its growth rate. We also call this phase as the maximum 302 probing phase since CUBIC is searching for a new W_max. In this 303 region, cwnd MUST be incremented by (W(t+RTT) - cwnd)/cwnd for each 304 received ACK. 306 3.5. Multiplicative decrease 308 When a packet loss occurs, CUBIC reduces its window size by a factor 309 of beta. Parameter beta SHOULD be set to 0.2. 311 W_max = cwnd; // save window size before reduction 312 cwnd = cwnd * (1-beta); // window reduction 314 A side effect of setting beta to a smaller value than 0.5 is slower 315 convergence. We believe that while a more adaptive setting of beta 316 could result in faster convergence, it will make the analysis of the 317 protocol much harder. This adaptive adjustment of beta is an item 318 for the next version of CUBIC. 320 3.6. Fast convergence 322 To improve the convergence speed of CUBIC, we add a heuristic in the 323 protocol. When a new flow joins the network, existing flows in the 324 network need to give up their bandwidth shares to allow the flow some 325 room for growth if the existing flows have been using all the 326 bandwidth of the network. To increase this release of bandwidth by 327 existing flows, the following mechanism called fast convergence 328 SHOULD be implemented. 330 With fast convergence, when a loss event occurs, before a window 331 reduction of congestion window, a flow remembers the last value of 332 W_max before it updates W_max for the current loss event. Let us 333 call the last value of W_max to be W_last_max. 335 if (W_max < W_last_max){ // check downward trend 336 W_last_max = W_max; // remember the last W_max 337 W_max = W_max*(2-beta)/2; // further reduce W_max 338 } else { // check upward trend 339 W_last_max = W_max // remember the last W_max 340 } 342 This allows W_max to be slightly less than the original W_max. Since 343 flows spend most of time around their W_max, flows with larger 344 bandwidth shares tend to spend more time around the plateau allowing 345 more time for flows with smaller shares to increase their windows. 347 4. Discussion 349 With a deterministic loss model where the number of packets between 350 two successive lost events is always 1/p, CUBIC always operates with 351 the concave window profile which greatly simplifies the performance 352 analysis of CUBIC. The average window size of CUBIC can be obtained 353 by the following function: 355 (C*(4-beta)/4/beta)^0.25 * RTT^0.75 / p^0.75 (Eq. 5) 357 With beta set to 0.2, the above formula is reduced to: 359 (C*3.8/0.8)^0.25 * RTT^0.75 / p^0.75 (Eq. 6) 361 We will determine the value of C in the following subsection using 362 Eq. 6. 364 4.1. Fairness to standard TCP 366 In environments where standard TCP is able to make reasonable use of 367 the available bandwidth, CUBIC does not significantly change this 368 state. 370 Standard TCP performs well in the following two types of networks: 372 1. networks with a small bandwidth-delay product (BDP) 374 2. networks with a short RTT, but not necessarily a small BDP 376 CUBIC is designed to behave very similarly to standard TCP in the 377 above two types of networks. The following two tables show the 378 average window size of standard TCP, HSTCP, and CUBIC. The average 379 window size of standard TCP and HSTCP is from [RFC3649]. The average 380 window size of CUBIC is calculated by using Eq. 6 and CUBIC TCP 381 friendly mode for three different values of C. 383 +----------+-------+--------+-------------+-------------+-----------+ 384 | Loss | TCP | HSTCP | CUBIC | CUBIC | CUBIC | 385 | Rate P | | | (C=0.04) | (C=0.4) | (C=4) | 386 +----------+-------+--------+-------------+-------------+-----------+ 387 | 10^-2 | 12 | 12 | 12 | 12 | 12 | 388 | 10^-3 | 38 | 38 | 38 | 38 | 66 | 389 | 10^-4 | 120 | 263 | 120 | 209 | 371 | 390 | 10^-5 | 379 | 1795 | 660 | 1174 | 2087 | 391 | 10^-6 | 1200 | 12279 | 3713 | 6602 | 11740 | 392 | 10^-7 | 3795 | 83981 | 20878 | 37126 | 66022 | 393 | 10^-8 | 12000 | 574356 | 117405 | 208780 | 371269 | 394 +----------+-------+--------+-------------+-------------+-----------+ 396 Response function of standard TCP, HSTCP, and CUBIC in networks with 397 RTT = 100ms. The average window size W is in MSS-sized segments. 399 Table 1 401 +--------+-----------+-----------+------------+-----------+---------+ 402 | Loss | Average | Average | CUBIC | CUBIC | CUBIC | 403 | Rate P | TCP W | HSTCP W | (C=0.04) | (C=0.4) | (C=4) | 404 +--------+-----------+-----------+------------+-----------+---------+ 405 | 10^-2 | 12 | 12 | 12 | 12 | 12 | 406 | 10^-3 | 38 | 38 | 38 | 38 | 38 | 407 | 10^-4 | 120 | 263 | 120 | 120 | 120 | 408 | 10^-5 | 379 | 1795 | 379 | 379 | 379 | 409 | 10^-6 | 1200 | 12279 | 1200 | 1200 | 2087 | 410 | 10^-7 | 3795 | 83981 | 3795 | 6603 | 11740 | 411 | 10^-8 | 12000 | 574356 | 20878 | 37126 | 66022 | 412 +--------+-----------+-----------+------------+-----------+---------+ 414 Response function of standard TCP, HSTCP, and CUBIC in networks with 415 RTT = 10ms. The average window size W is in MSS-sized segments. 417 Table 2 419 Both tables show that CUBIC with any of these three C values is more 420 friendly to TCP than HSTCP, especially in networks with a short RTT 421 where TCP performs reasonably well. For example, in a network with 422 RTT = 10ms and p=10^-6, TCP has an average window of 1200 packets. 423 If the packet size is 1500 bytes, then TCP can achieve an average 424 rate of 1.44 Gbps. In this case, CUBIC with C=0.04 or C=0.4 achieves 425 exactly the same rate as Standard TCP, whereas HSTCP is about ten 426 times more aggressive than Standard TCP. 428 We can see that C determines the aggressiveness of CUBIC in competing 429 with other protocols for the bandwidth. CUBIC is more friendly to 430 the Standard TCP, if the value of C is lower. However, we do not 431 recommend to set C to a very low value like 0.04, since CUBIC with a 432 low C cannot efficiently use the bandwidth in long RTT and high 433 bandwidth networks. Based on these observations, we find C=0.4 gives 434 a good balance between TCP-friendliness and aggressiveness of window 435 growth. Therefore, C SHOULD be set to 0.4. With C set to 0.4, Eq. 6 436 is reduced to: 438 1.17 * RTT^0.75 / p^0.75 (Eq. 7) 440 Eq. 7 is then used in the next subsection to show the scalability of 441 CUBIC. 443 4.2. Using Spare Capacity 445 CUBIC uses a more aggressive window growth function than Standard TCP 446 under long RTT and high bandwidth networks. 448 The following table shows that to achieve 10Gbps rate, standard TCP 449 requires a packet loss rate of 2.0e-10, while CUBIC requires a packet 450 loss rate of 3.4e-8. 452 +------------------+-----------+---------+---------+---------+ 453 | Throughput(Mbps) | Average W | TCP P | HSTCP P | CUBIC P | 454 +------------------+-----------+---------+---------+---------+ 455 | 1 | 8.3 | 2.0e-2 | 2.0e-2 | 2.0e-2 | 456 | 10 | 83.3 | 2.0e-4 | 3.9e-4 | 3.3e-4 | 457 | 100 | 833.3 | 2.0e-6 | 2.5e-5 | 1.6e-5 | 458 | 1000 | 8333.3 | 2.0e-8 | 1.5e-6 | 7.3e-7 | 459 | 10000 | 83333.3 | 2.0e-10 | 1.0e-7 | 3.4e-8 | 460 +------------------+-----------+---------+---------+---------+ 462 Required packet loss rate for Standard TCP, HSTCP, and CUBIC to 463 achieve a certain throughput. We use 1500-byte packets and an RTT of 464 0.1 seconds. 466 Table 3 468 Our test results in [HKLRX06] indicate that CUBIC uses the spare 469 bandwidth left unused by existing Standard TCP flows in the same 470 bottleneck link without taking away much bandwidth from the existing 471 flows. 473 4.3. Difficult Environments 475 CUBIC is designed to remedy the poor performance of TCP in fast long- 476 distance networks. It is not designed for wireless networks. 478 4.4. Investigating a Range of Environments 480 CUBIC has been extensively studied by using both NS-2 simulation and 481 test-bed experiments covering a wide range of network environments. 482 More information can be found in [HKLRX06]. 484 4.5. Protection against Congestion Collapse 486 In case that there is congestion collapse, CUBIC behaves likely 487 standard TCP since CUBIC modifies only the window adjustment 488 algorithm of TCP. Thus, it does not modify the ACK clocking and 489 Timeout behaviors of Standard TCP. 491 4.6. Fairness within the Alternative Congestion Control Algorithm. 493 CUBIC ensures convergence of competing CUBIC flows with the same RTT 494 in the same bottleneck links to an equal bandwidth share. When 495 competing flows have different RTTs, their bandwidth shares are 496 linearly proportional to the inverse of their RTT ratios. This is 497 true independent of the level of statistical multiplexing in the 498 link. 500 4.7. Performance with Misbehaving Nodes and Outside Attackers 502 This is not considered in the current CUBIC. 504 4.8. Responses to Sudden or Transient Events 506 In case that there is a sudden congestion, a routing change, or a 507 mobility event, CUBIC behaves the same as Standard TCP. 509 4.9. Incremental Deployment 511 CUBIC requires only the change of TCP senders, and does not require 512 any assistant of routers. 514 5. Security Considerations 516 This proposal makes no changes to the underlying security of TCP. 518 6. IANA Considerations 520 There are no IANA considerations regarding this document. 522 7. Acknowledgements 524 Alexander Zimmermann and Lars Eggert have received funding from the 525 European Union's Horizon 2020 research and innovation program 526 2014-2018 under grant agreement No. 644866 (SSICLOPS). This document 527 reflects only the authors' views and the European Commission is not 528 responsible for any use that may be made of the information it 529 contains. 531 8. References 533 8.1. Normative References 535 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP 536 Selective Acknowledgment Options", RFC 2018, October 1996. 538 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 539 Requirement Levels", BCP 14, RFC 2119, March 1997. 541 [RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows", 542 RFC 3649, December 2003. 544 [RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC 545 4960, September 2007. 547 [RFC5033] Floyd, S. and M. Allman, "Specifying New Congestion 548 Control Algorithms", BCP 133, RFC 5033, August 2007. 550 [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP 551 Friendly Rate Control (TFRC): Protocol Specification", RFC 552 5348, September 2008. 554 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 555 Control", RFC 5681, September 2009. 557 [RFC6582] Henderson, T., Floyd, S., Gurtov, A., and Y. Nishida, "The 558 NewReno Modification to TCP's Fast Recovery Algorithm", 559 RFC 6582, April 2012. 561 8.2. Informative References 563 [CEHRX07] Cai, H., Eun, D., Ha, S., Rhee, I., and L. Xu, "Stochastic 564 Ordering for Internet Congestion Control and its 565 Applications", In Proceedings of IEEE INFOCOM , May 2007. 567 [FHP00] Floyd, S., Handley, M., and J. Padhye, "A Comparison of 568 Equation-Based and AIMD Congestion Control", May 2000. 570 [GV02] Gorinsky, S. and H. Vin, "Extended Analysis of Binary 571 Adjustment Algorithms", Technical Report TR2002-29, 572 Department of Computer Sciences , The University of Texas 573 at Austin , August 2002. 575 [HKLRX06] Ha, S., Kim, Y., Le, L., Rhee, I., and L. Xu, "A Step 576 toward Realistic Performance Evaluation of High-Speed TCP 577 Variants", International Workshop on Protocols for Fast 578 Long-Distance Networks , February 2006. 580 [HRX08] Ha, S., Rhee, I., and L. Xu, "CUBIC: A New TCP-Friendly 581 High-Speed TCP Variant", ACM SIGOPS Operating System 582 Review , 2008. 584 [K03] Kelly, T., "Scalable TCP: Improving Performance in 585 HighSpeed Wide Area Networks", ACM SIGCOMM Computer 586 Communication Review , April 2003. 588 [LS08] Leith, D. and R. Shorten, "H-TCP: TCP Congestion Control 589 for High Bandwidth-Delay Product Paths", Internet-draft 590 draft-leith-tcp-htcp-06 , April 2008. 592 [XHR04] Xu, L., Harfoush, K., and I. Rhee, "Binary Increase 593 Congestion Control for Fast, Long Distance Networks", In 594 Proceedings of IEEE INFOCOM , March 2004. 596 Appendix A. ToDo List 598 o Incorporate ICCRG's feedback (see 599 http://trac.tools.ietf.org/group/irtf/trac/wiki/ICCRG_cubic) 601 o ncorporate feedback from Neal Cardwell (see http://www.ietf.org/ 602 mail-archive/web/tcpm/current/msg09508.html) 604 Authors' Addresses 606 Injong Rhee 607 North Carolina State University 608 Department of Computer Science 609 Raleigh, NC 27695-7534 610 US 612 Email: rhee@ncsu.edu 613 Lisong Xu 614 University of Nebraska-Lincoln 615 Department of Computer Science and Engineering 616 Lincoln, NE 68588-01150 617 US 619 Email: xu@unl.edu 621 Sangtae Ha 622 University of Colorado at Boulder 623 Department of Computer Science 624 Boulder, CO 80309-0430 625 US 627 Email: sangtae.ha@colorado.edu 629 Alexander Zimmermann 630 NetApp 631 Sonnenallee 1 632 Kirchheim 85551 633 Germany 635 Phone: +49 89 900594712 636 Email: alexander.zimmermann@netapp.com 638 Lars Eggert 639 NetApp 640 Sonnenallee 1 641 Kirchheim 85551 642 Germany 644 Phone: +49 151 12055791 645 Email: lars@netapp.com 647 Richard Scheffenegger 648 NetApp 649 Am Euro Platz 2 650 Vienna 1120 651 Austria 653 Phone: +43 1 3676811 3146 654 Email: rs@netapp.com