idnits 2.17.1 draft-ietf-idr-route-damp-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** Cannot find the required boilerplate sections (Copyright, IPR, etc.) in this document. Expected boilerplate is as follows today (2024-04-27) according to https://trustee.ietf.org/license-info : IETF Trust Legal Provisions of 28-dec-2009, Section 6.a: This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. IETF Trust Legal Provisions of 28-dec-2009, Section 6.b(i), paragraph 2: Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved. IETF Trust Legal Provisions of 28-dec-2009, Section 6.b(i), paragraph 3: This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- ** Missing expiration date. The document expiration date should appear on the first and last page. ** The document seems to lack a 1id_guidelines paragraph about Internet-Drafts being working documents. ** The document seems to lack a 1id_guidelines paragraph about 6 months document validity -- however, there's a paragraph with a matching beginning. Boilerplate error? ** The document seems to lack a 1id_guidelines paragraph about the list of current Internet-Drafts. ** The document seems to lack a 1id_guidelines paragraph about the list of Shadow Directories. ** The document is more than 15 pages and seems to lack a Table of Contents. == No 'Intended status' indicated for this document; assuming Proposed Standard Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) ** The document seems to lack separate sections for Informative/Normative References. All references will be assumed normative when checking for downward references. ** There are 14 instances of too long lines in the document, the longest one being 5 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == Line 617 has weird spacing: '...in path exc...' -- 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 (January 8, 1998) is 9606 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Unused Reference: '2' is defined on line 1300, but no explicit reference was found in the text == Unused Reference: '3' is defined on line 1307, but no explicit reference was found in the text == Unused Reference: '4' is defined on line 1311, but no explicit reference was found in the text == Unused Reference: '6' is defined on line 1321, but no explicit reference was found in the text == Unused Reference: '7' is defined on line 1326, but no explicit reference was found in the text == Unused Reference: '8' is defined on line 1330, but no explicit reference was found in the text ** Obsolete normative reference: RFC 1268 (ref. '1') (Obsoleted by RFC 1655) -- Possible downref: Non-RFC (?) normative reference: ref. '2' ** Downref: Normative reference to an Historic RFC: RFC 1267 (ref. '3') ** Obsolete normative reference: RFC 1771 (ref. '5') (Obsoleted by RFC 4271) ** Downref: Normative reference to an Historic RFC: RFC 1520 (ref. '6') ** Downref: Normative reference to an Informational RFC: RFC 1774 (ref. '7') ** Downref: Normative reference to an Informational RFC: RFC 1773 (ref. '8') Summary: 16 errors (**), 0 flaws (~~), 8 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Internet Engineering Task Force Curtis Villamizar 3 INTERNET-DRAFT ANS 4 draft-ietf-idr-route-damp-01 Ravi Chandra 5 Cisco 6 Ramesh Govindan 7 ISI 8 January 8, 1998 10 BGP Route Flap Damping 12 Status of this Memo 14 This document is an Internet-Draft. Internet-Drafts are working 15 documents of the Internet Engineering Task Force (IETF), its areas, 16 and its working groups. Note that other groups may also distribute 17 working documents as Internet-Drafts. 19 Internet-Drafts are draft documents valid for a maximum of six months 20 and may be updated, replaced, or obsoleted by other documents at any 21 time. It is inappropriate to use Internet- Drafts as reference 22 material or to cite them other than as ``work in progress.'' 24 To view the entire list of current Internet-Drafts, please check the 25 ``1id-abstracts.txt'' listing contained in the Internet-Drafts Shadow 26 Directories on ftp.is.co.za (Africa), ftp.nordu.net (Europe), 27 munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or 28 ftp.isi.edu (US West Coast). 30 Abstract 32 A usage of the BGP routing protocol is described which is capable of 33 reducing the routing traffic passed on to routing peers and therefore 34 the load on these peers without adversely affecting route convergence 35 time for relatively stable routes. This technique has been 36 implemented in commercial products supporting BGP. The technique is 37 also applicable to IDRP. 39 The overall goals are: 41 o to provide a mechanism capable of reducing router processing load 42 caused by instability 44 o in doing so prevent sustained routing oscillations 46 o to do so without sacrificing route convergence time for generally 47 well behaved routes. 49 This must be accomplished keeping other goals of BGP in mind: 51 o pack changes into a small number of updates 53 o preserve consistent routing 55 o minimal addition space and computational overhead 57 An excessive rate of update to the advertised reachability of a subset 58 of Internet prefixes has been widespread in the Internet. This 59 observation was made in the early 1990s by many people involved in 60 Internet operations and remains the case. These excessive updates are 61 not necessarily periodic so route oscillation would be a misleading 62 term. The informal term used to describe this effect is ``route 63 flap''. The techniques described here are now widely deployed and are 64 commonly referred to as ``route flap damping''. 66 1 Overview 68 To maintain scalability of a routed internet, it is necessary to 69 reduce the amount of change in routing state propagated by BGP in 70 order to limit processing requirements. The primary contributors of 71 processing load resulting from BGP updates are the BGP decision 72 process and adding and removing forwarding entries. 74 Consider the following example. A widely deployed BGP implementation 75 may tend to fail due to high routing update volume. For example, it 76 may be unable to maintain it's BGP or IGP sessions if sufficiently 77 loaded. The failure of one router can further contribute to the load 78 on other routers. This additional load may cause failures in other 79 instances of the same implementation or other implementations with a 80 similar weakness. In the worst case, a stable oscillation could 81 result. Such worse cases have already been observed in practice. 83 A BGP implementation must be prepared for a large volume of routing 84 traffic. A BGP implementation cannot rely upon the sender to 85 sufficiently shield it from route instabilities. The guidelines here 86 are designed to prevent sustained oscillations, but do not eliminate 87 the need for robust and efficient implementations. The mechanisms 88 described here allow routing instability to be contained at an AS 89 border router bordering the instability. 91 Even where BGP implementations are highly robust, the performance of 92 the routing process is limited. Limiting the propagation of 93 unnecessary change then becomes an issue of maintaining reasonable 94 route change convergence time as a routing topology grows. 96 2 Methods of Limiting Route Advertisement 98 Two methods of controlling the frequency of route advertisement are 99 described here. The first involves fixed timers. The fixed timer 100 technique has no space overhead per route but has the disadvantage of 101 slowing route convergence for the normal case where a route does not 102 have a history of instability. The second method overcomes this 103 limitation at the expense of maintaining some additional space 104 overhead. The additional overhead includes a small amount of state 105 per route and a very small processing overhead. 107 It is possible and desirable to combine both techniques. In practice, 108 fixed timers have been set to very short time intervals and have 109 proven useful to pack routes (NLRI) into a smaller number of updates 110 when routes arrive in separate updates. 112 Seldom are fixed timers set to the tens of minutes to hours that would 113 be necessary to actually damp route flap. To do so would produce the 114 undesirable effect of severely limiting routing convergence. 116 2.1 Existing Fixed Timer Recommendations 118 BGP-3 does not make specific recommendations in this area [1]. The 119 short section entitled ``Frequency of Route Selection'' simply 120 recommends that something be done and makes broad statements regarding 121 certain properties that are desirable or undesirable. 123 BGP4 retains the ``Frequency of Route Advertisement'' section and adds 124 a ``Frequency of Route Origination'' section. BGP-4 describes a 125 method of limiting route advertisement involving a fixed 126 (configurable) MinRouteAdvertisementInterval timer and fixed 127 MinASOriginationInterval timer [5]. The recommended timer values of 128 MinRouteAdvertisementInterval is 30 seconds and 129 MinASOriginationInterval is 15 seconds. 131 2.2 Desirable Properties of Damping Algorithms 133 Before describing damping algorithms the objectives need to be clearly 134 defined. Some key properties are examined to clarify the design 135 rationale. 137 The overall objective is to reduce the route update load without 138 limiting convergence time for well behaved routes. To accomplish 139 this, criteria must be defined for well behaved and poorly behaved 140 routes. An algorithm must be defined which allows poorly behaved 141 routes to be identified. Ideally, this measure would be a prediction 142 of the future stability of a route. 144 Any delay in propagation of well behaved routes should be minimal. 145 Some delay is tolerable to support better packing of updates. Delay 146 of poorly behave routes should, if possible, be proportional to a 147 measure of the expected future instability of the route. Delay in 148 propagating an unstable route should cause the unstable route to be 149 suppressed until there is some degree of confidence that the route has 150 stabilized. 152 If a large number of route changes are received in separate updates 153 over some very short period of time and these updates have the 154 potential to be combined into a single update then these should be 155 packed as efficiently as possible before propagating further. Some 156 small delay in propagating well behaved routes is tolerable and is 157 necessary to allow better packing of updates. 159 Where routes are unstable, use and announcement of the routes should 160 be suppressed rather than suppressing their removal. Where one route 161 to a destination is stable, and another route to the same destination 162 is somewhat unstable, if possible, the unstable route should be 163 suppressed more aggressively than if there were no alternate path. 165 Routing consistency within an AS is very important. Only very minimal 166 delay of internal BGP (IBGP) should be done. Routing consistency 167 across AS boundaries is also very important. It is highly undesirable 168 to advertise a route that is different from the route that is being 169 used, except for a very minimal time. It is more desirable to 170 suppress the acceptance of a route (and therefore the use of that 171 route in the IGP) rather than suppress only the redistribution. 173 It is clearly not possible to accurately predict the future stability 174 of a route. The recent history of stability is generally regarded as 175 a good basis for estimating the likelihood of future stability. The 176 criteria that is used to distinguish well behaved from poorly behaved 177 routes is therefore based on the recent history of stability of the 178 route. There is no simple quantitative expression of recent stability 179 so a figure of merit must be defined. Some desirable characteristics 180 of this figure of merit would be that the farther in the past that 181 instability occurred, the less it's affect on the figure of merit and 182 that the instability measure would be cumulative rather than 183 reflecting only the most recent event. 185 The algorithms should behave such that for routes which have a history 186 of stability but make a few transitions, those transitions should be 187 made quickly. If transitions continue, advertisement of the route 188 should be suppressed. There should be some memory of prior instabil- 189 ity. The degree to which prior instability is considered should be 190 gradually reduced as long as the route remains announced and stable. 192 2.3 Design Choices 194 After routes have been accepted their readvertisement will be briefly 195 suppressed to improve packing of updates. There may be a lengthy 196 suppression of the acceptance of an external route. How long a route 197 will be suppressed is based on a figure of merit that is expected to 198 be correlated to the probability of future instability of a route. 199 Routes with high figure of merit values will be suppressed. An 200 exponential decay algorithm was chosen as the basis for reducing the 201 figure of merit over time. These choices should be viewed as 202 suggestions for implementation. 204 An exponential decay function has the property that previous 205 instability can be remembered for a fairly long time. The rate at 206 which the instability figure of merit decays slows as time goes on. 207 Exponential decay has the following property. 209 f(f(figure-of-merit, t1), t2) = f(figure-of-merit, t1+t2) 211 This property allows the decay for a long period to be computed in a 212 single operation regardless of the current value (figure-of-merit). 213 As a performance optimization, the decay can be applied in fixed time 214 increments. Given a desired decay half life, the decay for a single 215 time increment can be computed ahead of time. The decay for multiple 216 time increments is expressed below. 218 f(figure-of-merit, n*t0) = f(figure-of-merit, t0)**n = K**n 220 The values of K ** n can be precomputed for a reasonable number of 221 ``n'' and stored in an array. The value of ``K'' is always less than 222 one. The array size can be bounded since the value quickly approaches 223 zero. This makes the decay easy to compute using an array bound 224 check, an array lookup and a single multiply regardless as to how much 225 time has elapsed. 227 3 Limiting Route Advertisements using Fixed Timers 229 This method of limiting route advertisements involves the use of fixed 230 timers applied to the process of sending routes. It's primary purpose 231 is to improve the packing of routes in BGP update messages. The delay 232 in advertising a stable route should be bounded and minimal. The 233 delay in advertising an unreachable need not be zero, but should also 234 be bounded and should probably have a separate bound set less than or 235 equal to the bound for a reachable advertisement. 237 Routes that need to be readvertised can be marked in the RIB or an 238 external set of structures maintained, which references the RIB. 239 Periodically, a subset of the marked routes can be flushed. This is 240 fairly straightforward and accomplishes the objectives. Computation 241 for too simple an implementation may be order N squared. To avoid N 242 squared performance, some form of data structure is needed to group 243 routes with common attributes. 245 An implementation should pack updates efficiently, provide a minimum 246 readvertisement delay, provide a bounds on the maximum readvertisement 247 delay that would be experienced solely as a result of the algorithm 248 used to provide a minimum delay, and must be computationally efficient 249 in the presence of a very large number of candidates for 250 readvertisement. 252 4 Stability Sensitive Suppression of Route Advertisement 254 This method of limiting route advertisements uses a measure of route 255 stability applied on a per route basis. This technique is applied 256 when receiving updates from external peers only (EBGP). Applying this 257 technique to IBGP learned routes or to advertisement to IBGP or EBGP 258 peers after making a route selection can result in routing loops. 260 A figure of merit based on a measure of instability is maintained on a 261 per route basis. This figure of merit is used in the decision to 262 suppress the use of the route. Routes with high figure of merit are 263 suppressed. Each time a route is withdrawn, the figure of merit is 264 incremented. While the route is not changing the figure of merit 265 value is decayed exponentially with separate decay rates depending on 266 whether the route is stable and reachable or has been stable and 267 unreachable. The decay rate may be slower when the route is unreach- 268 able, or the stability figure of merit could remain fixed (not decay 269 at all) while the route remains unreachable. Whether to decay un- 270 reachable routes at the same rate, a slower rate, or not at all is an im- 271 plementation choice. Decaying at a slower rate is recommended. 273 A very efficient implementation is suggested in the following 274 sections. The implementation only requires computation for the routes 275 contained in an update, when an update is received or withdrawn (as 276 opposed to the simplistic approach of periodically decaying each 277 route). The suggested implementation involves only a small number of 278 simple operations, and can be implemented using scaled integers. 280 The behavior of unstable routes is fairly predictable. Severely 281 flapping routes will often be advertised and withdrawn at regular time 282 intervals corresponding to the timers of a particular protocol (the 283 IGP or exterior protocol in use where the problem exists). Marginal 284 circuits or mild congestion can result in a long term pattern of 285 occasional brief route withdrawal or occasional brief connectivity. 287 4.1 Single vs. Multiple Configuration Parameter Sets 289 The behavior of the algorithm is modified by a number of configurable 290 parameters. It is possible to configure separate sets of parameters 291 designed to handle short term severe route flap and chronic milder 292 route flap (a pattern of occasional drops over a long time period). 293 The former would require a fast decay and low threshold (allowing a 294 small number of consecutive flaps to cause a route to be suppressed, 295 but allowing it to be reused after a relatively short period of 296 stability). The latter would require a very slow decay and a higher 297 threshold and might be appropriate for routes for which there was an 298 alternate path of similar bandwidth. 300 It may also be desirable to configure different thresholds for routes 301 with roughly equivalent alternate paths than for routes where the 302 alternate paths have a lower bandwidth or tend to be congested. This 303 can be solved by associating a different set of parameters with 304 different ranges of preference values. Parameter selection could be 305 based on BGP LOCAL_PREF. 307 Parameter selection could also be based on whether an alternate route 308 was known. A route would be considered if, for any applicable 309 parameter set, an alternate route with the specified preference value 310 existed and the figure of merit associated with the parameter set did 311 not indicate a need to suppress the route. A less aggressive 312 suppression would be applied to the case where no alternate route at 313 all existed. In the simplest case, a more aggressive suppression 314 would be applied if any alternate route existed. Only the highest 315 preference (most preferred) value needs to be specified, since the 316 ranges may overlap. 318 It might also be desirable to configure a different set of thresholds 319 for routes which rely on switched services and may disconnect at times 320 to reduce connect charges. Such routes might be expected to change 321 state somewhat more often, but should be suppressed if continuous 322 state changes indicate instability. 324 While not essential, it might be desirable to be able to configure 325 multiple sets of configuration parameters per route. It may also be 326 desirable to be able to configure sets of parameters that only 327 correspond to a set of routes (identified by AS path, peer router, 328 specific destinations or other means). Experience may dictate how 329 much flexibility is needed and how to best to set the parameters. 330 Whether to allow different damping parameter sets for different 331 routes, and whether to allow multiple figures of merit per route is an 332 implementation choice. 334 Parameter selection can also be based on prefix length. The rationale 335 is that longer prefixes tend to reach less end systems and are less 336 important and these less important prefixes can be damped more 337 aggressively. This technique is in fairly widespread use. Small 338 sites or those with dense address allocation who are multihomed are 339 often reachable by long prefixes which are not easily aggregated. 340 These sites tend to dispute the choice of prefix length for parameter 341 selection. Advocates of the technique point out that it encourages 342 better aggregation. 344 4.2 Configuration Parameters 346 At configuration time, a number of parameters may be specified by the 347 user. The configuration parameters are expressed in units meaningful 348 to the user. These differ from the parameters used at run time which 349 are in unit convenient for computation. The run time parameters are 350 derived from the configuration parameters. Suggested configuration 351 parameters are listed below. 353 cutoff threshold (cut) 355 This value is expressed as a number of route withdrawals. It is 356 the value above which a route advertisement will be suppressed. 358 reuse threshold (reuse) 360 This value is expressed as a number of route withdrawals. It is 361 the value below which a suppressed route will now be used again. 363 maximum hold down time (T-hold) 365 This value is the maximum time a route can be suppressed no matter 366 how unstable it has been prior to this period of stability. 368 decay half life while reachable (decay-ok) 370 This value is the time duration in minutes or seconds during which 371 the accumulated stability figure of merit will be reduced by half 372 if the route if considered reachable (whether suppressed or not). 374 decay half life while unreachable (decay-ng) 376 This value is the time duration in minutes or seconds during which 377 the accumulated stability figure of merit will be reduced by half 378 if the route if considered unreachable. If not specified or set to 379 zero, no decay will occur while a route remains unreachable. 381 decay memory limit (Tmax-ok or Tmax-ng) 383 This is the maximum time that any memory of previous instability 384 will be retained given that the route's state remains unchanged, 385 whether reachable or unreachable. This parameter is generally used 386 to determine array sizes. 388 There may be multiple sets of the parameters above as described in 389 Section 4.1. The configuration parameters listed below would be 390 applied system wide. These include the time granularity of all 391 computations, and the parameters used to control reevaluation of 392 routes that have previously been suppressed. 394 time granularity (delta-t) 396 This is the time granularity in seconds used to perform all decay 397 computations. 399 reuse list time granularity (delta-reuse) 401 This is the time interval between evaluations of the reuse lists. 402 Each reuse lists corresponds to an additional time increment. 404 reuse list memory reuse-list-max 406 This is the time value corresponding to the last reuse list. This 407 may be the maximum value of T-hold for all parameter sets of may be 408 configured. 410 number of reuse lists (reuse-list-size) 412 This is the number of reuse lists. It may be determined from 413 reuse-list-max or set explicitly. 415 FIGURES ARE FOUND IN THE POSTSCRIPT AND HTML VERSIONS ONLY 417 Figure 1: Instability figure of merit for flap at a constant rate 419 A necessary optimization is described in Section 4.8.6 that involves 420 an array referred to as the ``reuse index array''. A reuse index 421 array is needed for each decay rate in use. The reuse index array is 422 used to estimate which reuse list to place a route when it is 423 suppressed. Proper placement avoids the need to periodically evaluate 424 decay to determine if a route can be reused. Using the reuse index 425 array avoids the need to compute a logarithm to determine placement. 426 One additional system wide parameter can be introduced. 428 reuse index array size (reuse-index-array-size) 430 This is the size of reuse index arrays. This size determines the 431 accuracy with which suppressed routes can be placed within the set 432 of reuse lists when suppressed for a long time. 434 4.3 Guidelines for Setting Parameters 436 The decay half life should be set to a time considerably longer than 437 the period of the route flap it is intended to address. For example, 438 if the decay is set to ten minutes and a route is withdrawn and 439 readvertised exactly every ten minutes, the route would continue to 440 flap if the cutoff was set to a value of 2 or above. 442 The stability figure of merit itself is an accumulated time decayed 443 total. This must be kept in mind in setting the decay time, cutoff 444 values and reuse values. For example, if a route flaps at four times 445 the decay rate, it will reach 3 in 4 cycles, 4 in 6 cycles, 5 in 10 446 cycles, and will converge at about 6.3. At twice the decay time, it 447 will reach 3 in 7 cycles, and converge at a value of less than 3.5. 449 Figure 1 shows the stability figure of merit for route flap at a 450 constant rate. The time axis is labeled in multiples of the decay 451 half life. The plots represent route flap with a period of 1/2, 1/3, 452 1/4, and 1/8 times the decay half life. A ceiling of 4.5 was set, 453 which can be seen to affect three of the plots, effectively limiting 454 the time it takes to readvertise the route regardless of the prior 455 history. With the cutoff and reuse thresholds suggested by the dotted 456 lines, routes would be suppressed after being declared unreachable 2-3 457 times and be used again after approximately 2 decay half life periods 458 of stability. 460 FIGURES ARE FOUND IN THE POSTSCRIPT AND HTML VERSIONS ONLY 462 Figure 2: Separate decay constants when unreachable 464 From either maximum hold time value (Tmax-ok or Tmax-ng), a ratio of 465 the cutoff to a ceiling can be determined. An integer value for the 466 ceiling can then be chosen such that overflow will not be a problem 467 and all other values can be scaled accordingly. If both cutoffs are 468 specified or if multiple parameter sets are used the highest ceiling 469 will be used. 471 Figure 2 show the effect of configuring separate decay rates to be 472 used when the route is reachable or unreachable. The decay rate is 473 5 times slower when the route is unreachable. In the three case 474 shown, the period of the route flap is equal to the decay half life 475 but the route is reachable 1/8 of the time in one, reachable 1/2 the 476 time in one, and reachable 7/8 of the time in the other. In the last 477 case the route is not suppressed until after the third unreachable 478 (when it is above the top threshold after becoming reachable again). 480 In both Figure 1 and Figure 2, routes would be suppressed. Routes 481 flapping at the decay half life or less would be withdrawn two or 482 three times and then remain withdrawn until they had remained stably 483 announced and stable for on the order of 1 1/2 to 2 1/2 times the 484 decay half life (given the ceiling in the example). 486 A larger time granularity will keep table storage down. The time 487 granularity should be less than a minimal reasonable time between 488 expected worse case route flaps. It might be reasonable to fix this 489 parameter at compile time or set a default and strongly recommend that 490 the user leave it alone. With an exponential decay, array size can be 491 greatly reduced by setting a period of complete stability after which 492 the decayed total will be considered zero rather than retaining a tiny 493 quantity. Alternately, very long decays can be implemented by 494 multiplying more than once if array bounds are exceeded. 496 The reuse lists hold suppressed routes grouped according to how long 497 it will be before the routes are eligible for reuse. Periodically 498 each list will be advanced by one position and one list removed as de- 499 scribed in Section 4.8.7. All of the suppressed routes in the removed 500 list will be reevaluated and either used or placed in another list 501 according to how much additional time must elapse before the route can 502 be reused. The last list will always contain all the routes which 503 will not be advertised for more time than is appropriate for the re- 504 maining list heads. When the last list advances to the front, some of 505 the routes will not be ready to be used and will have to be requeued. 506 The time interval for reconsidering suppressed routes and number of list 507 heads should be configurable. Reasonable defaults might be 30 seconds and 508 64 list heads. A route suppressed for a long time would need to be reeval- 509 uated every 32 minutes. 511 4.4 Run Time Data Structures 513 A fixed small amount of per system storage will be required. Where 514 sets of multiple configuration parameters are used, storage will be 515 required per set of parameters. A small amount of per route storage 516 is required. A set of list heads is needed. These list heads are 517 used to arrange suppressed routes according to the time remaining 518 until they can be reused. 520 If multiple sets of configuration parameters are allowed per route, 521 there is a need for some means of associating more than one figure of 522 merit and set of parameters with each route. Building a linked list 523 of these objects seems like one of a number of reasonable 524 implementations. Similarly, a means of associating a route to a reuse 525 list is required. A small overhead will be required for the pointers 526 needed to implement whatever data structure is chosen for the reuse 527 lists. The suggested implementation uses a double linked lists and so 528 requires two pointers per figure of merit. 530 Each set of configuration parameters can reference decay arrays and 531 reuse arrays. These arrays should be shared among multiple sets of 532 parameters since their storage requirement is not negligible. There 533 will be only one set of reuse list heads for the entire router. 535 4.4.1 Data Structures for Configuration Parameter Sets 537 Based on the configuration parameters described in the previous 538 section, the following values can be computed as scaled integers 539 directly from the corresponding configuration parameters. 541 o decay array scale factor (decay-array-scale-factor) 543 o cutoff value (cut) 545 o reuse value (reuse) 547 o figure of merit ceiling (ceiling) 549 Each configuration parameter set will reference one or two decay 550 arrays and one or two reuse arrays. Only one array will be needed if 551 the decay rate is the same while a route is unreachable as while it is 552 reachable, or if the stability figure of merit does not decay while a 553 route is unreachable. 555 4.4.2 Data Structures per Decay Array and Reuse Index Array 557 The following are also computed from the configuration parameters 558 though not as directly. 560 o decay rate per tick (decay-delta-t) 562 o decay array size (decay-array-size) 564 o decay array (decay) 566 o reuse index array size (reuse-index-array-size) 568 o reuse index array (reuse-index-array) 570 For each decay rate specified, an array will be used to store the 571 value of a computed parameter raised to the power of the index of each 572 array element. This is to speed computations. The decay rate per 573 tick is an intermediate value expressed as a real number and used to 574 compute the values stored in the decay arrays. The array size is 575 computed from the decay memory limit configuration parameter expressed 576 as an array size or as a maximum hold time. 578 The decay array size must be of sufficient size to accommodate the 579 specified decay memory given the time granularity, or sufficient to 580 hold the number of array elements until integer rounding produces a 581 zero result if that value is smaller, or a implementation imposed 582 reasonable size to prevent configurations which use excessive memory. 583 Implementations may chose to make the array size shorter and multiply 584 more than once when decaying a long time interval to reduce storage. 586 The reuse index arrays serve a similar purpose to the decay arrays. 587 The amount of time until a route can be reused can be determined using 588 a array lookup. The array can be built given the decay rate. The 589 array is indexed using a scaled integer proportional to the ratio 590 between a current stability figure of merit value and the value needed 591 for the route to be reused. 593 4.4.3 Per Route State 595 Information must be maintained per some tuple representing a route. 596 At the very minimum, the NLRI (BGP prefix and length) must be 597 contained in the tuple. Different BGP attributes may be included or 598 excluded depending on the specific situation. The AS path should also 599 be contained in the tuple be default. The tuple may also optionally 600 contain other BGP attributes such as MULTI_EXIT_DISCRIMINATOR (MED). 602 The tuple representing a route for the purpose of route flap damping 603 is: 605 tuple entry default options 607 ------------------------------------------- 609 NLRI 611 prefix required 613 length required 615 AS path included option to exclude 617 last AS set in path excluded option to include 619 next hop excluded option to include 621 MED excluded option to include 623 in comparisons only 625 The AS path is generally included in order to identify downstream 626 instability which is not being damped or not being sufficiently damped 627 and is alternating between a stable and an unstable path. Under rare 628 circumstances it may be desirable to exclude AS path for all or a 629 subset of prefixes. If an AS path ends in an AS set, in practice the 630 path is always for an aggregate. Changes to the trailing AS set 631 should be ignored. Ideally the AS path comparison should insure that 632 at least one AS has remained constant in the old and new AS set, but 633 completely ignoring the contents of a trailing AS set is also 634 acceptable. 636 Including next hop and MED changes can help suppress the use of an AS 637 which is internally unstable or avoid a next hop which is closer to an 638 unstable IGP path in the adjacent AS. If a large number of MED values 639 are used, the increase in the amount of state may become a problem. 640 For this reason MED is disabled by default and enabled only as part of 641 the tuple comparison, using a single state entry regardless of MED 642 value. Including MED will suppress the use of the adjacent AS even 643 though the change need not be propagated further. Using MED is only a 644 safe practice if a path is known to exist through another AS or where 645 there are enough peering sites with the adjacent AS such that routes 646 heard at only a subset of the peering sites will be suppressed. 648 4.4.4 Data Structures per Route 650 The following information must be maintained per route. A route here 651 is considered to be a tuple usually containing NLRI, next hop, and AS 652 path as defined in Section 4.4.3. 654 stability figure of merit (figure-of-merit) 656 Each route must have a stability figure of merit per applicable 657 parameter set. 659 last time updated (time-update) 661 The exact last time updated must be maintained to allow exponential 662 decay of the accumulated figure of merit to be deferred until the 663 route might reasonable be considered eligible for a change in 664 status (having gone from unreachable to reachable or advancing 665 within the reuse lists). 667 config block pointer 669 Any implementation that supports multiple parameter sets must 670 provide a means of quickly identifying which set of parameters 671 corresponds to the route currently being considered. For 672 implementations supporting only parameter sets where all routes 673 must be treated the same, this pointer is not required. 675 reuse list traversal pointers 677 If doubly linked lists are used to implement reuse lists, then two 678 pointers will be needed, previous and next. Generally there is a 679 double linked list which is unused when a route is suppressed from 680 use that can be used for reuse list traversal eliminating the need 681 for additional pointer storage. 683 4.5 Processing Configuration Parameters 685 From the configuration parameters, it is possible to precompute a 686 number of values that will be used repeatedly and retain these to 687 speed later computations that will be required frequently. 689 The methods of scaled integer arithmetic are not described in detail 690 here. The methods of determining the real values are given. 691 Translation into scaled integer values and the details of scaled 692 integer arithmetic are left up to the individual implementations. 694 figure of merit scale factor ( scale-figure-of-merit ) 696 The ceiling value can be set to be the largest integer that can fit 697 in half the bits available for an unsigned integer. This will 698 allow the scaled integers to be multiplied by the scaled decay 699 value and then shifted down. Implementations may prefer to use 700 real numbers or may use any integer scaling deemed appropriate for 701 their architecture. 703 penalty value and thresholds (as proportional scaled integers) 705 The figure of merit penalty for one route withdrawal and the cutoff 706 values must be scaled according to the above scaling factor. 708 decay rate per tick (decay[1]) 710 The decay value per increment of time as defined by the time 711 granularity must be determined (at least initially as a floating 712 point number). The per tick decay is a number slightly less than 713 one. It is the Nth root of the one half where N is the half life 714 divided by the time granularity. 716 decay[1] = exp ((1 / (decay-rate/delta-t)) * log (1/2)) 718 decay array size (decay-array-size) 720 The decay array size is the decay memory divided by the time 721 granularity. If integer truncation brings the value of an array 722 element to zero, the array can be made smaller. An implementation 723 should also impose a maximum reasonable array size or allow more 724 than one multiplication. 726 decay-array-size = (Tmax/delta-t) 728 decay array (decay[]) 729 Each i-th element of the decay array is the per tick delay raised 730 to the i-th power. This might be best done by successive floating 731 point multiplies followed by scaling and integer rounding or 732 truncation. The array itself need only be computed at startup. 734 decay[i] = decay[1] ** i 736 4.6 Building the Reuse Index Arrays 738 The reuse lists may be accessed quite frequently if a lot of routes 739 are flapping sufficiently to be suppressed. A method of speeding the 740 determination of which reuse list to use for a given route is 741 suggested. This method is introduced in Section 4.2, its 742 configuration described in Section 4.4.2 and the algorithms described 743 in Section 4.8.6 and Section 4.8.7. This section describes building 744 the reuse list index arrays. 746 A ratio of the figure of merit of the route under consideration to the 747 cutoff value is used as the basis for an array lookup. The ratio is 748 scaled and truncated to an integer and used to index the array. The 749 array entry is an integer used to determine which reuse list to use. 751 reuse array maximum ratio (max-ratio) 753 This is the maximum ratio between the current value of the 754 stability figure of merit and the target reuse value that can be 755 indexed by the reuse array. It may be limited by the ceiling 756 imposed by the maximum hold time or by the amount of time that the 757 reuse lists cover. 759 max-ratio = min(ceiling/reuse, exp((1 / 760 (half-life/reuse-array-time)) * log(1/2))) 762 reuse array scale factor ( scale-factor ) 764 Since the reuse array is an estimator, the reuse array scale factor 765 has to be computed such that the full size of the reuse array is 766 used. 768 scale-factor = (max-ratio - 1) / reuse-array-size 770 reuse index array (reuse) 771 Each reuse index array entry should contain an index into the reuse 772 list array pointing to one of the list heads. This index should 773 corresponding to the reuse list that will be evaluated just after a 774 route would be eligible for reuse given the ratio of current value 775 of the stability figure of merit to target reuse value 776 corresponding the the reuse array entry. 778 reuse-array[j] = integer(log(1 / (1 + ((j+1) * 779 (max-ratio-1)))) / reuse-time-granularity) 781 To determine which reuse queue to place a route which is being 782 suppressed, the following procedure is used. Divide the current 783 figure of merit by the cutoff. Subtract one. Multiply by the scale 784 factor. This is the array index. If it is off the end of the array 785 use the last queue otherwise look in the array and pick the number of 786 the queue from the array at that index. This is quite fast and well 787 worth the setup and storage required. 789 4.7 A Sample Configuration 791 A simple example is presented here in which the space overhead is 792 estimated for a set of configuration parameters. The design here 793 assumes: 795 1. there is a single parameter set used for all routes, 797 2. decay time for unreachable routes is slower than for reachable 798 routes 800 3. the arrays must be full size, rather than allow more than one 801 multiply per decay operation to reduce the array size. 803 This example is used in later sections. The use of multiple parameter 804 sets complicates the examples somewhat. Where multiple parameter sets 805 are allowed for a single route, the decay portion of the algorithm is 806 repeated for each parameter set. If different routes are allowed to 807 have different parameter sets, the routes must have pointers to the 808 parameter sets to keep the time to locate to a minimum, but the 809 algorithms are otherwise unchanged. 811 A sample set of configuration parameters and a sample set of 812 implementation parameters are provided in in the two following lists. 814 1. Configuration Parameters 816 o cut = 1.25 818 o reuse = 0.5 820 o T-hold = 15 mins 822 o decay-ok = 5 min 824 o decay-ng = 15 min 826 o Tmax-ok, Tmax-ng = 15, 30 mins 828 2. Implementation Parameters 830 o delta-t = 1 sec 832 o delta-reuse 834 o reuse-list-size = 256 836 o reuse-index-array-size = 1,024 838 Using these configuration and implementation parameters and the 839 equations in Section 4.5, the space overhead can be computed. There 840 is a fixed space overhead that is independent of the number of routes. 841 There is a space requirement associated with a stable route. There is 842 a larger space requirement associated with an unstable route. The 843 space requirements for the parameters above are provide in the lists 844 below. 846 1. fixed overhead (using parameters from previous example) 848 o 900 * integer - decay array 850 o 1,800 * integer - decay array 852 o 120 * pointer - reuse list-heads 854 o 2,048 * integer - reuse index arrays 856 2. overhead per stable route 858 o pointer - containing null entry 859 FIGURES ARE FOUND IN THE POSTSCRIPT AND HTML VERSIONS ONLY 861 Figure 3: Some fairly long route flap cycles, repeated for 12 862 minutes, followed by a period of stability. 864 3. overhead per unstable route 866 o pointer - to a damping structure containing the following 868 o integer - figure of merit + bit for state 870 o integer - last time updated 872 o pointer (optional) to configuration parameter block 874 o 2 * pointer - reuse list pointers (prev, next) 876 Figure 3 shows the behavior of the algorithm with the parameters given 877 above. Four cases are given in this example. In all four, there is a 878 twelve minute period of route oscillations. Two periods of oscilla- 879 tion are used, 2 minutes and 4 minutes. Two duty cycles are used, one 880 in which the route is reachable during 20% of the cycle and the other 881 where the route is reachable during 80% of the cycle. In all four 882 cases, the route becomes suppressed after it becomes unreachable the 883 second time. Once suppressed, it remains suppressed until some period 884 after becoming stable. The routes which oscillate over a 4 minute pe- 885 riod are no longer suppressed within 9-11 minutes after becoming sta- 886 ble. The routes with a 2 minute period of oscillation are suppressed for 887 nearly the maximum 15 minute period after becoming stable. 889 4.8 Processing Routing Protocol Activity 891 The prior sections concentrate on configuration parameters and their 892 relationship to the parameters and arrays used at run time and provide 893 the algorithms for initializing run time storage. This section 894 provides the steps taken in processing routing events and timer events 895 when running. 897 The routing events are: 899 1. A BGP peer or new route comes up for the first time (or after an 900 extended down time) (Section 4.8.1) 902 2. A route becomes unreachable (Section 4.8.2) 904 3. A route becomes reachable again (Section 4.8.3) 906 4. A route changes (Section 4.8.4) 908 5. A peer goes down (Section 4.8.5) 910 The reuse list is used to provide a means of fast evaluation of route 911 that had been suppressed, but had been stable long enough to be reused 912 again or had been suppressed long enough that it can be treated as a 913 new route. The following two operations are described. 915 1. Inserting into a reuse list (Section 4.8.6) 917 2. Reuse list processing every delta-t seconds (Section 4.8.7) 919 4.8.1 Processing a New Peer or New Routes 921 When a peer comes up, no action is required if the routes had no 922 previous history of instability, for example if this is the first time 923 the peer is coming up and announcing these routes. For each route, 924 the pointer to the damping structure would be zeroed and route used. 925 The same action is taken for a new route or a route that has been down 926 long enough that the figure of merit reached zero and the damping 927 structure was deleted. 929 4.8.2 Processing Unreachable Messages 931 When a route is withdrawn or changed (Section 4.8.4 describes how a 932 change is handled), the following procedure is used. 934 If there is no previous stability history (the damping structure 935 pointer is zero), then: 937 1. allocate a damping structure 939 2. set figure-of-merit = 1 941 3. withdraw the route 943 Otherwise, if there is an existing damping structure, then: 945 1. set t-diff = t-now - t-updated 947 2. if (t-diff puts you off the end of the array) { 949 set figure-of-merit =1 951 }else { 953 set figure-of-merit =figure-of-merit *decay-array-ok [t-diff ]+ 1 955 if(figure-of-merit >ceiling) { 957 set figure-of-merit =ceiling 959 } 961 } 963 3. remove the route from a reuse list if it is on one 965 4. withdraw the route unless it is already suppressed 967 In either case then: 969 1. set t-updated = t-now 971 2. insert into a reuse list (see Section 4.8.6) 973 If there was a stability history, the previous value of the stability 974 figure of merit is decayed. This is done using the decay array 975 (decay-array). The index is determined by subtracting the current 976 time and the last time updated, then dividing by the time granularity. 977 If the index is zero, the figure of merit is unchanged (no decay). If 978 it is greater than the array size, it is zeroed. Otherwise use the 979 index to fetch a decay array element and multiply the figure of merit 980 by the array element. If using the suggested scaled integer method, 981 shift down half an integer. Add the scaled penalty for one more un- 982 reachable (shown above as 1). If the result is above the ceiling re- 983 place it with the ceiling value. Now update the last time updated field 984 (preferably taking into account how much time was truncated before doing 985 the decay calculation). 987 When a route becomes unreachable, alternate paths must be considered. 988 This process is complicated slightly if different configuration param- 989 eters are used in the presence or absence of viable alternate paths. 990 If all of these alternate paths have been suppressed because there had 991 previously been an alternate route and the new route withdrawal 992 changes that condition, the suppressed alternate paths must be reeval- 993 uated. They should be reevaluated in order of normal route prefer- 994 ence. When one of these alternate routes is encountered that had been 995 suppressed but is now usable since there is no alternate route, no 996 further routes need to be reevaluated. This only applies if routes 997 are given two different reuse thresholds, one for use when there is an al- 998 ternate path and a higher threshold to use when suppressing the route would 999 result in making the destination completely unreachable. 1001 4.8.3 Processing Route Advertisements 1003 When a route is readvertised if there is no damping structure, then 1004 the procedure is the same as in Section 4.8.1. 1006 1. don't create a new damping structure 1008 2. use the route 1010 If an damping structure exists, the figure of merit is decayed and the 1011 figure of merit and last time updated fields are updated. A decision 1012 is now made as to whether the route can be used immediately or needs 1013 to be suppressed for some period of time. 1015 1. set t-diff = t-now - t-updated 1017 2. if (t-diff puts you off the end of the array) { 1019 set figure-of-merit =0 1021 }else { 1023 set figure-of-merit= figure-of-merit* decay-array-ng[t-diff] 1025 } 1027 3. if ( not suppressed and figure-of-merit < cut ) { 1029 use the route 1031 }else if( suppressed and figure-of-merit< reuse) { 1033 set state to not suppressed 1035 remove the route from a reuse list 1036 use the route 1038 }else { 1040 set state to suppressed 1042 don't use the route 1044 insert into a reuse list(see Section4.8.6) 1046 } 1048 4. if ( figure-of-merit > 0 ) { 1050 set t-updated= t-now 1052 }else { 1054 recover memory for damping struct 1056 zero pointer to damping struct 1058 } 1060 If the route is deemed usable, a search for the current best route 1061 must be made. The newly reachable route is then evaluated according 1062 to the BGP protocol rules for route selection. 1064 If the new route is usable, the previous best route is examined. 1065 Prior to route comparisons, the current best route may have to be 1066 reevaluated if separate parameter sets are used depending on the 1067 presence or absence of an alternate route. If there had been no 1068 alternate the previous best route may be suppressed. 1070 If the new route is to be suppressed it is placed on a reuse list only 1071 if it would have been preferred to the current best route had the new 1072 route been accepted as stable. There is no reason to queue a route on 1073 a reuse list if after the route becomes usable it would not be used 1074 anyway due to the existence of a more preferred route. Such a route 1075 would not have to be reevaluated unless the preferred route became 1076 unreachable. As specified here, the less preferred route would be 1077 reevaluated and potentially used or potentially added to a reuse list 1078 when processing the withdrawal of a more preferred best route. 1080 4.8.4 Processing Route Changes 1082 If a route is replaced by a peer router by supplying a new path, the 1083 route that is being replaced should be treated as if an unreachable 1084 were received (see Section 4.8.2). This will occur when a peer 1085 somewhere back in the AS path is continuously switching between two AS 1086 paths and that peer is not damping route flap (or applying less 1087 damping). There is no way to determine if one AS path is stable and 1088 the other is flapping, or if they are both flapping. If the cycle is 1089 sufficiently short compared to convergence times neither route through 1090 that peer will deliver packets very reliably. Since there is no way 1091 to affect the peer such that it chooses the stable of the two AS 1092 paths, the only viable option is to penalize both routes by considering 1093 each change as an unreachable followed by a route advertisement. 1095 4.8.5 Processing A Peer Router Loss 1097 When a peer routing session is broken, either all individual routes 1098 advertised by that peer may be marked as unstable, or the peering 1099 session itself may be marked as unstable. Marking the peer will save 1100 considerable memory. Since the individual routes are advertised as 1101 unreachable to routers beyond the immediate problem, per route state 1102 will be incurred beyond the peer immediately adjacent to the BGP 1103 session that went down. If the instability continues, the immediately 1104 adjacent router need only keep track of the peer stability history. 1105 The routers beyond that point will receive no further advertisements 1106 or withdrawal of routes and will dispose of the damping structure over 1107 time. 1109 BGP notification through an optional transitive attribute that damping 1110 will already be applied may be considered in the future to reduce the 1111 number of routers that incur damping structure storage overhead. 1113 4.8.6 Inserting into the Reuse Timer List 1115 The reuse lists are used to provide a means of fast evaluation of 1116 route that had been suppressed, but had been stable long enough to be 1117 reused again. The data structure consists of a series of list heads. 1118 Each list contains a set of routes that are scheduled for reevaluation 1119 at approximately the same time. The set of reuse list heads are 1120 treated as a circular array. 1122 A simple implementation of the circular array of list heads would be 1123 an array containing the list heads with an offset. The offset would 1124 identify the first list. The Nth list would be at the index 1125 corresponding to N plus the offset modulo the number of list heads. 1126 This design will be assumed in the examples that follow. 1128 A key requirement is to be able to insert an entry in the most 1129 appropriate queue with a minimum of computation. The computation is 1130 given only the current value of figure-of-merit. The array, scale, 1131 and bounds are precomputed to map figure-of-merit to the nearest list 1132 head without requiring a logarithm to be computed (see Section 4.5). 1134 1. scale figure-of-merit for the index array lookup producing index 1136 2. check index against the array bound 1138 3. if (within the array bound) { 1140 set index =reuse-array [index ] 1142 }else { 1144 set index =reuse-list-size -1 1146 } 1148 4. insert into the list 1150 reuse-list[ modulo reuse-list-size (index +offset )] 1152 Choosing the correct reuse list involves only a multiply and shift to 1153 do the scaling, an integer truncation, then an array lookup. The most 1154 common method of implementing a circular array is to use an array and 1155 apply an offset and modulo operation to pick the correct array entry. 1156 The offset is incremented to rotate the the circular array. 1158 4.8.7 Handling Reuse Timer Events 1160 The granularity of the reuse timer should be more course that that of 1161 the decay timer. As a result, when the reuse timer fires, suppressed 1162 routes should be decayed by multiple increments of decay time. Some 1163 computation can be avoided by always inserting into the reuse list 1164 corresponding to one time increment past reuse eligibility. In cases 1165 where the reuse lists have a longer ``memory'' than the ``decay 1166 memory'' (described above), all of the routes in the first queue will 1167 be available for immediate reuse if reachable or the history entry 1168 could be disposed of if unreachable. 1170 When it is time to advance the lists, the first queue on the reuse 1171 list must be processed and the circular queue must be rotated. Using 1172 an array and an offset as a circular array (as described in 1173 Section 4.8.6), the algorithm below is repeated every t-reuse seconds. 1175 1. save a pointer to the current zeroth queue head and zero the list 1176 head entry 1178 2. set offset = modulo reuse-list-size ( offset + 1 ), thereby 1179 rotating the circular queue of list-heads 1181 3. if ( the saved list head pointer is non-empty ) 1183 foreach entry { 1185 sett-diff =t-now -t-updated 1187 set figure-of-merit =figure-of-merit *decay-array-ok [t-diff ] 1189 sett-updated =t-now 1191 if( figure-of-merit< reuse) 1193 reuse the route 1195 else 1197 re-insert into another list (see Section 4.8.6) 1199 } 1201 The value of the zeroth list head would be saved and the array entry 1202 itself zeroed. The list heads would then be advanced by incrementing 1203 the offset. Starting with the saved head of the old zeroth list, each 1204 route would be reevaluated and used, disposed of entirely or requeued 1205 if it were not ready for reuse. If a route is used, it must be 1206 treated as if it were a new route advertisement as described in 1207 Section 4.8.3. 1209 5 Implementation Experience 1211 The first implementations of ``route flap damping'' were the route 1212 server daemon (rsd) coding by Ramesh Govindan (ISI) and the Cisco IOS 1213 implementation by Ravi Chandra. Both implementations first became 1214 available in 1995 and have been used extensively. The rsd 1215 implementation has been in use in route servers at the NSF funded 1216 Network Access Points (NAPs) and at other major Internet 1217 interconnects. The Cisco IOS version has been in use by Internet 1218 Service Providers worldwide. The rsd implementation has been 1219 integrated in releases of gated (see http://www.gated.org) and is 1220 available in commercial routers using gated. 1222 There are now more than 2 years of BGP route damping deployment 1223 experience. Some problems have occurred in deployment. So far these 1224 are solvable by careful implementation of the algorithm and by careful 1225 deployment. In some topologies coordinated deployment can be helpful 1226 and in all cases disclosure of the use of route damping and the param- 1227 eters used is highly beneficial in debugging connectivity problems. 1229 Some of the problems have occurred due to subtle implementation 1230 errors. Route damping should never be applied on IBGP learned routes. 1231 To do so can open the possibility for persistent route loops. 1232 Implementations should disallow this configuration. Penalties for 1233 flapping should only be applied when a route is removed or replaced 1234 and not when a route is added. If damping parameters are applied 1235 consistently, this implementation constraint will result in a stable 1236 secondary path being preferred over an unstable primary path due to 1237 damping of the primary path near the source. 1239 In topologies where multiple AS paths to a given destination exist 1240 flapping of the primary path can result in suppression of the 1241 secondary path. This can occur if no damping is being done near the 1242 cause of the route flap or if damping is being applied more 1243 aggressively by a distant AS. This problem can be solved in one of two 1244 ways. Damping can be done near the source of the route flap and the 1245 damping parameters can be made consistent. Alternately, a distant AS 1246 which insists on more aggressive damping parameters can disable 1247 penalizing routes on AS path change, penalizing routes only if they 1248 are withdrawn completely. In order to do so, the implementation must 1249 support this option (as described in Section 4.4.3). 1251 Route flap should be damped near the source. Single homed 1252 destinations can be covered by static routes. Aggregation provides 1253 another means of damping. Providers should damp their own internal 1254 problems, however damping on IGP link state origination is not yet 1255 implemented by router vendors. Providers which use multiple AS within 1256 their own topology should damp between their own AS. Providers should 1257 damp adjacent providers AS. 1259 Damping provides a means to limit propagation excessive route change 1260 when connectivity is highly intermittent. Once a problem is 1261 corrected, select damping state can be manually cleared. In order to 1262 determine where damping may have occurred after connectivity problems, 1263 providers should publish their damping parameters. Providers should 1264 be willing to manually clear damping on specific prefixes or AS paths 1265 at the request of other providers when the request is accompanied by 1266 assurance that the problem has truly been addressed. 1268 By damping their own routing information, providers can reduce their 1269 own need to make requests of other providers to clear damping state 1270 after correcting a problem. Providers should be pro-active and 1271 monitor what prefixes and paths are suppressed in addition to 1272 monitoring link states and BGP session state. 1274 Acknowledgements 1276 This work and this document may not have been completed without the 1277 advise, comments and encouragement of Yakov Rekhter (Cisco). Dennis 1278 Ferguson (MCI) provided a description of the algorithms in the gated 1279 BGP implementation and many valuable comments and insights. David 1280 Bolen (ANS) and Jordan Becker (ANS) provided valuable comments, 1281 particularly regarding early simulations. Over four years elapsed 1282 between the initial draft presented to the BGP WG (October 1993) and 1283 this iteration. At the time of this writing there is significant 1284 experience with two implementations, each having been deployed since 1285 1995. One was led by Ramesh Govindan (ISI) for the NSF Routing Ar- 1286 biter project. The second was led by Ravi Chandra (Cisco). Sean Doran 1287 (Sprintlink) and Serpil Bayraktar (ANS) were among the early independent 1288 testers of the Cisco pre-beta implementation. Valuable comments and im- 1289 plementation feedback were shared by many individuals on the IETF IDR WG 1290 and the RIPE Routing Work Group and in NANOG and IEPG. 1292 References 1294 [1] P. Gross and Y. Rekhter. Application of the border gateway proto- 1295 col in 1296 the internet. Request for Comments (Draft Standard) RFC 1268, In- 1297 ternet Engineering Task Force, October 1991. (Obsoletes RFC1164); 1298 (Obsoleted by RFC1655). ftp://ds.internic.net/rfc/rfc1268.txt. 1300 [2] ISO/IEC. Iso/iec 10747 - information technology - telecommunica- 1301 tions and information exchange between systems - protocol for 1302 exchange of inter-domain routeing information among intermediate 1303 systems to support forwarding of iso 1304 8473 pdus. Technical report, International Organization for Stan- 1305 dardization, August 1994. ftp://merit.edu/pub/iso/idrp.ps.gz. 1307 [3] K. Lougheed and Y. Rekhter. A border gateway protocol 3 (BGP-3). 1308 Request for Comments (Draft Standard) RFC 1267, In- 1309 ternet Engineering Task Force, October 1991. (Obsoletes RFC1163). 1310 ftp://ds.internic.net/rfc/rfc1267.txt. 1311 [4] Y. Rekhter and P. Gross. Application of the border gateway proto- 1312 col in the internet. Request for Comments (Draft Standard) 1313 RFC 1772, Internet Engineering Task Force, March 1995. (Obsoletes 1314 RFC1655). ftp://ds.internic.net/rfc/rfc1772.txt. 1316 [5] Y. Rekhter and T. Li. A border 1317 gateway protocol 4 (BGP-4). Request for Comments (Draft Standard) 1318 RFC 1771, Internet Engineering Task Force, March 1995. (Obsoletes 1319 RFC1654). ftp://ds.internic.net/rfc/rfc1771.txt. 1321 [6] Y. Rekhter and C. Topolcic. Exchanging routing information across 1322 provider boundaries in the CIDR environment. Request for Comments 1323 (Informational) RFC 1520, Internet Engineering Task Force, 1324 September 1993. ftp://ds.internic.net/rfc/rfc1520.txt. 1326 [7] P. Traina. BGP-4 protocol analysis. Request for Comments (Infor- 1327 mational) RFC 1774, Internet Engineering Task Force, March 1995. 1328 ftp://ds.internic.net/rfc/rfc1774.txt. 1330 [8] P. Traina. Experience with the BGP-4 protocol. Request for Com- 1331 ments (Informational) RFC 1773, 1332 Internet Engineering Task Force, March 1995. (Obsoletes RFC1656). 1333 ftp://ds.internic.net/rfc/rfc1773.txt. 1335 Security Considerations 1337 The practices outlined in this document do not further weaken the 1338 security of the routing protocols. Denial of service is possible in 1339 an already insecure routing environment but these practices only 1340 contribute to the persistence of such attacks and do not impact the 1341 methods of prevention and the methods of determining the source. 1343 Author's Addresses 1345 Curtis Villamizar 1347 ANS Communications 1349 1351 Ravi Chandra 1353 Cisco Systems 1355 1357 Ramesh Govindan 1359 ISI 1361