Debian Codegraph @

rfc1772.html

raw
1 <pre>Network Working Group Y. Rekhter
2 Request for Comments: 1772 T.J. Watson Research Center, IBM Corp.
3 Obsoletes: <a href="./rfc1655">1655</a> P. Gross
4 Category: Standards Track MCI
5 Editors
6 March 1995
7
8
9 <span class="h1">Application of the Border Gateway Protocol in the Internet</span>
10
11 Status of this Memo
12
13 This document specifies an Internet standards track protocol for the
14 Internet community, and requests discussion and suggestions for
15 improvements. Please refer to the current edition of the "Internet
16 Official Protocol Standards" (STD 1) for the standardization state
17 and status of this protocol. Distribution of this memo is unlimited.
18
19 Abstract
20
21 This document, together with its companion document, "A Border
22 Gateway Protocol 4 (BGP-4)", define an inter-autonomous system
23 routing protocol for the Internet. "A Border Gateway Protocol 4
24 (BGP-4)" defines the BGP protocol specification, and this document
25 describes the usage of the BGP in the Internet.
26
27 Information about the progress of BGP can be monitored and/or
28 reported on the BGP mailing list ([email protected]).
29
30 Acknowledgements
31
32 This document was originally published as <a href="./rfc1164">RFC 1164</a> in June 1990,
33 jointly authored by Jeffrey C. Honig (Cornell University), Dave Katz
34 (MERIT), Matt Mathis (PSC), Yakov Rekhter (IBM), and Jessica Yu
35 (MERIT).
36
37 The following also made key contributions to <a href="./rfc1164">RFC 1164</a> -- Guy Almes
38 (ANS, then at Rice University), Kirk Lougheed (cisco Systems), Hans-
39 Werner Braun (SDSC, then at MERIT), and Sue Hares (MERIT).
40
41 We like to explicitly thank Bob Braden (ISI) for the review of the
42 previous version of this document.
43
44 This updated version of the document is the product of the IETF BGP
45 Working Group with Phill Gross (MCI) and Yakov Rekhter (IBM) as
46 editors.
47
48
49
50
51
52 <span class="grey">Rekhter &amp; Gross [Page 1]</span></pre>
53 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
54 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
55
56
57 John Moy (Proteon) contributed <a href="#section-7">Section 7</a> "Required set of supported
58 routing policies".
59
60 Scott Brim (Cornell University) contributed the basis for <a href="#section-8">Section 8</a>
61 "Interaction with other exterior routing protocols".
62
63 Most of the text in <a href="#section-9">Section 9</a> was contributed by Gerry Meyer
64 (Spider).
65
66 Parts of the Introduction were taken almost verbatim from [<a href="#ref-3" title="&quot;Supernetting: an Address Assignment and Aggregation Strategy&quot;">3</a>].
67
68 We would like to acknowledge Dan Long (NEARNET) and Tony Li (cisco
69 Systems) for their review and comments on the current version of the
70 document.
71
72 The work of Yakov Rekhter was supported in part by the National
73 Science Foundation under Grant Number NCR-9219216.
74
75 <span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. Introduction</span>
76
77 This memo describes the use of the Border Gateway Protocol (BGP) [<a href="#ref-1" title="&quot;A Border Gateway Protocol 4 (BGP-4)">1</a>]
78 in the Internet environment. BGP is an inter-Autonomous System
79 routing protocol. The network reachability information exchanged via
80 BGP provides sufficient information to detect routing loops and
81 enforce routing decisions based on performance preference and policy
82 constraints as outlined in <a href="./rfc1104">RFC 1104</a> [<a href="#ref-2" title="&quot;Models of Policy Based Routing&quot;">2</a>]. In particular, BGP exchanges
83 routing information containing full AS paths and enforces routing
84 policies based on configuration information.
85
86 As the Internet has evolved and grown over in recent years, it has
87 become painfully evident that it is soon to face several serious
88 scaling problems. These include:
89
90 - Exhaustion of the class-B network address space. One
91 fundamental cause of this problem is the lack of a network
92 class of a size which is appropriate for mid-sized
93 organization; class-C, with a maximum of 254 host addresses, is
94 too small while class-B, which allows up to 65534 addresses, is
95 too large to be densely populated.
96
97 - Growth of routing tables in Internet routers are beyond the
98 ability of current software (and people) to effectively manage.
99
100 - Eventual exhaustion of the 32-bit IP address space.
101
102 It has become clear that the first two of these problems are likely
103 to become critical within the next one to three years. Classless
104 inter-domain routing (CIDR) attempts to deal with these problems by
105
106
107
108 <span class="grey">Rekhter &amp; Gross [Page 2]</span></pre>
109 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
110 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
111
112
113 proposing a mechanism to slow the growth of the routing table and the
114 need for allocating new IP network numbers. It does not attempt to
115 solve the third problem, which is of a more long-term nature, but
116 instead endeavors to ease enough of the short to mid-term
117 difficulties to allow the Internet to continue to function
118 efficiently while progress is made on a longer-term solution.
119
120 BGP-4 is an extension of BGP-3 that provides support for routing
121 information aggregation and reduction based on the Classless inter-
122 domain routing architecture (CIDR) [<a href="#ref-3" title="&quot;Supernetting: an Address Assignment and Aggregation Strategy&quot;">3</a>]. This memo describes the
123 usage of BGP-4 in the Internet.
124
125 All of the discussions in this paper are based on the assumption that
126 the Internet is a collection of arbitrarily connected Autonomous
127 Systems. That is, the Internet will be modeled as a general graph
128 whose nodes are AS's and whose edges are connections between pairs of
129 AS's.
130
131 The classic definition of an Autonomous System is a set of routers
132 under a single technical administration, using an interior gateway
133 protocol and common metrics to route packets within the AS and using
134 an exterior gateway protocol to route packets to other AS's. Since
135 this classic definition was developed, it has become common for a
136 single AS to use several interior gateway protocols and sometimes
137 several sets of metrics within an AS. The use of the term Autonomous
138 System here stresses the fact that, even when multiple IGPs and
139 metrics are used, the administration of an AS appears to other AS's
140 to have a single coherent interior routing plan and presents a
141 consistent picture of which destinations are reachable through it.
142
143 AS's are assumed to be administered by a single administrative
144 entity, at least for the purposes of representation of routing
145 information to systems outside of the AS.
146
147 <span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. BGP Topological Model</span>
148
149 When we say that a connection exists between two AS's, we mean two
150 things:
151
152 Physical connection: There is a shared Data Link subnetwork
153 between the two AS's, and on this shared subnetwork each AS has at
154 least one border gateway belonging to that AS. Thus the border
155 gateway of each AS can forward packets to the border gateway of
156 the other AS without resorting to Inter-AS or Intra-AS routing.
157
158 BGP connection: There is a BGP session between BGP speakers in
159 each of the AS's, and this session communicates those routes that
160 can be used for specific destinations via the advertising AS.
161
162
163
164 <span class="grey">Rekhter &amp; Gross [Page 3]</span></pre>
165 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
166 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
167
168
169 Throughout this document we place an additional restriction on the
170 BGP speakers that form the BGP connection: they must themselves
171 share the same Data Link subnetwork that their border gateways
172 share. Thus, a BGP session between adjacent AS's requires no
173 support from either Inter-AS or Intra-AS routing. Cases that do
174 not conform to this restriction fall outside the scope of this
175 document.
176
177 Thus, at each connection, each AS has one or more BGP speakers and
178 one or more border gateways, and these BGP speakers and border
179 gateways are all located on a shared Data Link subnetwork. Note that
180 BGP speakers do not need to be a border gateway, and vice versa.
181 Paths announced by a BGP speaker of one AS on a given connection are
182 taken to be feasible for each of the border gateways of the other AS
183 on the same shared subnetwork, i.e. indirect neighbors are allowed.
184
185 Much of the traffic carried within an AS either originates or
186 terminates at that AS (i.e., either the source IP address or the
187 destination IP address of the IP packet identifies a host internal to
188 that AS). Traffic that fits this description is called "local
189 traffic". Traffic that does not fit this description is called
190 "transit traffic". A major goal of BGP usage is to control the flow
191 of transit traffic.
192
193 Based on how a particular AS deals with transit traffic, the AS may
194 now be placed into one of the following categories:
195
196 stub AS: an AS that has only a single connection to one other AS.
197 Naturally, a stub AS only carries local traffic.
198
199 multihomed AS: an AS that has connections to more than one other
200 AS, but refuses to carry transit traffic.
201
202 transit AS: an AS that has connections to more than one other AS,
203 and is designed (under certain policy restrictions) to carry both
204 transit and local traffic.
205
206 Since a full AS path provides an efficient and straightforward way of
207 suppressing routing loops and eliminates the "count-to-infinity"
208 problem associated with some distance vector algorithms, BGP imposes
209 no topological restrictions on the interconnection of AS's.
210
211
212
213
214
215
216
217
218
219
220 <span class="grey">Rekhter &amp; Gross [Page 4]</span></pre>
221 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span>
222 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
223
224
225 <span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>. BGP in the Internet</span>
226
227 <span class="h3"><a class="selflink" id="section-3.1" href="#section-3.1">3.1</a> Topology Considerations</span>
228
229 The overall Internet topology may be viewed as an arbitrary
230 interconnection of transit, multihomed, and stub AS's. In order to
231 minimize the impact on the current Internet infrastructure, stub and
232 multihomed AS's need not use BGP. These AS's may run other protocols
233 (e.g., EGP) to exchange reachability information with transit AS's.
234 Transit AS's using BGP will tag this information as having been
235 learned by some method other than BGP. The fact that BGP need not run
236 on stub or multihomed AS's has no negative impact on the overall
237 quality of inter-AS routing for traffic that either destined to or
238 originated from the stub or multihomed AS's in question.
239
240 However, it is recommended that BGP be used for stub and multihomed
241 AS's as well. In these situations, BGP will provide an advantage in
242 bandwidth and performance over some of the currently used protocols
243 (such as EGP). In addition, this would reduce the need for the use
244 of default routes and in better choices of Inter-AS routes for
245 multihomed AS's.
246
247 <span class="h3"><a class="selflink" id="section-3.2" href="#section-3.2">3.2</a> Global Nature of BGP</span>
248
249 At a global level, BGP is used to distribute routing information
250 among multiple Autonomous Systems. The information flows can be
251 represented as follows:
252
253 +-------+ +-------+
254 BGP | BGP | BGP | BGP | BGP
255 ---------+ +---------+ +---------
256 | IGP | | IGP |
257 +-------+ +-------+
258
259 &lt;-AS A--&gt; &lt;--AS B-&gt;
260
261 This diagram points out that, while BGP alone carries information
262 between AS's, both BGP and an IGP may carry information across an AS.
263 Ensuring consistency of routing information between BGP and an IGP
264 within an AS is a significant issue and is discussed at length later
265 in <a href="#appendix-A">Appendix A</a>.
266
267 <span class="h3"><a class="selflink" id="section-3.3" href="#section-3.3">3.3</a> BGP Neighbor Relationships</span>
268
269 The Internet is viewed as a set of arbitrarily connected AS's.
270 Routers that communicate directly with each other via BGP are known
271 as BGP speakers. BGP speakers can be located within the same AS or in
272 different AS's. BGP speakers in each AS communicate with each other
273
274
275
276 <span class="grey">Rekhter &amp; Gross [Page 5]</span></pre>
277 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span>
278 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
279
280
281 to exchange network reachability information based on a set of
282 policies established within each AS. For a given BGP speaker, some
283 other BGP speaker with which the given speaker communicates is
284 referred to as an external peer if the other speaker is in a
285 different AS, while if the other speaker is in the same AS it is
286 referred to as an internal peer.
287
288 There can be as many BGP speakers as deemed necessary within an AS.
289 Usually, if an AS has multiple connections to other AS's, multiple
290 BGP speakers are needed. All BGP speakers representing the same AS
291 must give a consistent image of the AS to the outside. This requires
292 that the BGP speakers have consistent routing information among them.
293 These gateways can communicate with each other via BGP or by other
294 means. The policy constraints applied to all BGP speakers within an
295 AS must be consistent. Techniques such as using a tagged IGP (see
296 A.2.2) may be employed to detect possible inconsistencies.
297
298 In the case of external peers, the peers must belong to different
299 AS's, but share a common Data Link subnetwork. This common subnetwork
300 should be used to carry the BGP messages between them. The use of BGP
301 across an intervening AS invalidates the AS path information. An
302 Autonomous System number must be used with BGP to specify which
303 Autonomous System the BGP speaker belongs to.
304
305 <span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>. Requirements for Route Aggregation</span>
306
307 A conformant BGP-4 implementation is required to have the ability to
308 specify when an aggregated route may be generated out of partial
309 routing information. For example, a BGP speaker at the border of an
310 autonomous system (or group of autonomous systems) must be able to
311 generate an aggregated route for a whole set of destination IP
312 addresses (in BGP-4 terminology such a set is called the Network
313 Layer Reachability Information or NLRI) over which it has
314 administrative control (including those addresses it has delegated),
315 even when not all of them are reachable at the same time.
316
317 A conformant implementation may provide the capability to specify
318 when an aggregated NLRI may be generated.
319
320 A conformant implementation is required to have the ability to
321 specify how NLRI may be de-aggregated.
322
323 A conformant implementation is required to support the following
324 options when dealing with overlapping routes:
325
326 - Install both the less and the more specific routes
327
328 - Install the more specific route only
329
330
331
332 <span class="grey">Rekhter &amp; Gross [Page 6]</span></pre>
333 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span>
334 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
335
336
337 - Install the less specific route only
338
339 - Install neither route
340
341 Certain routing policies may depend on the NLRI (e.g. "research"
342 versus "commercial"). Therefore, a BGP speaker that performs route
343 aggregation should be cognizant, if possible, of potential
344 implications on routing policies when aggregating NLRI.
345
346 <span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>. Policy Making with BGP</span>
347
348 BGP provides the capability for enforcing policies based on various
349 routing preferences and constraints. Policies are not directly
350 encoded in the protocol. Rather, policies are provided to BGP in the
351 form of configuration information.
352
353 BGP enforces policies by affecting the selection of paths from
354 multiple alternatives and by controlling the redistribution of
355 routing information. Policies are determined by the AS
356 administration.
357
358 Routing policies are related to political, security, or economic
359 considerations. For example, if an AS is unwilling to carry traffic
360 to another AS, it can enforce a policy prohibiting this. The
361 following are examples of routing policies that can be enforced with
362 the use of BGP:
363
364 1. A multihomed AS can refuse to act as a transit AS for other
365 AS's. (It does so by only advertising routes to destinations
366 internal to the AS.)
367
368 2. A multihomed AS can become a transit AS for a restricted set of
369 adjacent AS's, i.e., some, but not all, AS's can use the
370 multihomed AS as a transit AS. (It does so by advertising its
371 routing information to this set of AS's.)
372
373 3. An AS can favor or disfavor the use of certain AS's for
374 carrying transit traffic from itself.
375
376 A number of performance-related criteria can be controlled with the
377 use of BGP:
378
379 1. An AS can minimize the number of transit AS's. (Shorter AS
380 paths can be preferred over longer ones.)
381
382 2. The quality of transit AS's. If an AS determines that two or
383 more AS paths can be used to reach a given destination, that AS
384 can use a variety of means to decide which of the candidate AS
385
386
387
388 <span class="grey">Rekhter &amp; Gross [Page 7]</span></pre>
389 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span>
390 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
391
392
393 paths it will use. The quality of an AS can be measured by such
394 things as diameter, link speed, capacity, tendency to become
395 congested, and quality of operation. Information about these
396 qualities might be determined by means other than BGP.
397
398 3. Preference of internal routes over external routes.
399
400 For consistency within an AS, equal cost paths, resulting from
401 combinations of policies and/or normal route selection procedures,
402 must be resolved in a consistent fashion.
403
404 Fundamental to BGP is the rule that an AS advertises to its
405 neighboring AS's only those routes that it uses. This rule reflects
406 the "hop-by-hop" routing paradigm generally used by the current
407 Internet.
408
409 <span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>. Path Selection with BGP</span>
410
411 One of the major tasks of a BGP speaker is to evaluate different
412 paths from itself to a set of destination covered by an address
413 prefix, select the best one, apply appropriate policy constraints,
414 and then advertise it to all of its BGP neighbors. The key issue is
415 how different paths are evaluated and compared. In traditional
416 distance vector protocols (e.g., RIP) there is only one metric (e.g.,
417 hop count) associated with a path. As such, comparison of different
418 paths is reduced to simply comparing two numbers. A complication in
419 Inter-AS routing arises from the lack of a universally agreed-upon
420 metric among AS's that can be used to evaluate external paths.
421 Rather, each AS may have its own set of criteria for path evaluation.
422
423 A BGP speaker builds a routing database consisting of the set of all
424 feasible paths and the list of destinations (expressed as address
425 prefixes) reachable through each path. For purposes of precise
426 discussion, it's useful to consider the set of feasible paths for a
427 set of destinations associated with a given address prefix. In most
428 cases, we would expect to find only one feasible path. However, when
429 this is not the case, all feasible paths should be maintained, and
430 their maintenance speeds adaptation to the loss of the primary path.
431 Only the primary path at any given time will ever be advertised.
432
433 The path selection process can be formalized by defining a complete
434 order over the set of all feasible paths to a set of destinations
435 associated with a given address prefix. One way to define this
436 complete order is to define a function that maps each full AS path to
437 a non-negative integer that denotes the path's degree of preference.
438 Path selection is then reduced to applying this function to all
439 feasible paths and choosing the one with the highest degree of
440 preference.
441
442
443
444 <span class="grey">Rekhter &amp; Gross [Page 8]</span></pre>
445 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span>
446 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
447
448
449 In actual BGP implementations, the criteria for assigning degree of
450 preferences to a path are specified as configuration information.
451
452 The process of assigning a degree of preference to a path can be
453 based on several sources of information:
454
455 1. Information explicitly present in the full AS path.
456
457 2. A combination of information that can be derived from the full
458 AS path and information outside the scope of BGP (e.g., policy
459 routing constraints provided as configuration information).
460
461 Possible criteria for assigning a degree of preference to a path are:
462
463 - AS count. Paths with a smaller AS count are generally better.
464
465 - Policy considerations. BGP supports policy-based routing based
466 on the controlled distribution of routing information. A BGP
467 speaker may be aware of some policy constraints (both within
468 and outside of its own AS) and do appropriate path selection.
469 Paths that do not comply with policy requirements are not
470 considered further.
471
472 - Presence or absence of a certain AS or AS's in the path. By
473 means of information outside the scope of BGP, an AS may know
474 some performance characteristics (e.g., bandwidth, MTU, intra-
475 AS diameter) of certain AS's and may try to avoid or prefer
476 them.
477
478 - Path origin. A path learned entirely from BGP (i.e., whose
479 endpoint is internal to the last AS on the path) is generally
480 better than one for which part of the path was learned via EGP
481 or some other means.
482
483 - AS path subsets. An AS path that is a subset of a longer AS
484 path to the same destination should be preferred over the
485 longer path. Any problem in the shorter path (such as an
486 outage) will also be a problem in the longer path.
487
488 - Link dynamics. Stable paths should be preferred over unstable
489 ones. Note that this criterion must be used in a very careful
490 way to avoid causing unnecessary route fluctuation. Generally,
491 any criteria that depend on dynamic information might cause
492 routing instability and should be treated very carefully.
493
494
495
496
497
498
499
500 <span class="grey">Rekhter &amp; Gross [Page 9]</span></pre>
501 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span>
502 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
503
504
505 <span class="h2"><a class="selflink" id="section-7" href="#section-7">7</a>. Required set of supported routing policies</span>
506
507 Policies are provided to BGP in the form of configuration
508 information. This information is not directly encoded in the
509 protocol. Therefore, BGP can provide support for very complex routing
510 policies. However, it is not required that all BGP implementations
511 support such policies.
512
513 We are not attempting to standardize the routing policies that must
514 be supported in every BGP implementation; we strongly encourage all
515 implementors to support the following set of routing policies:
516
517 1. BGP implementations should allow an AS to control announcements
518 of BGP-learned routes to adjacent AS's. Implementations should
519 also support such control with at least the granularity of a
520 single address prefix. Implementations should also support
521 such control with the granularity of an autonomous system,
522 where the autonomous system may be either the autonomous system
523 that originated the route, or the autonomous system that
524 advertised the route to the local system (adjacent autonomous
525 system). Care must be taken when a BGP speaker selects a new
526 route that can't be announced to a particular external peer,
527 while the previously selected route was announced to that peer.
528 Specifically, the local system must explicitly indicate to the
529 peer that the previous route is now infeasible.
530
531 2. BGP implementations should allow an AS to prefer a particular
532 path to a destination (when more than one path is available).
533 At the minimum an implementation shall support this
534 functionality by allowing to administratively assign a degree
535 of preference to a route based solely on the IP address of the
536 neighbor the route is received from. The allowed range of the
537 assigned degree of preference shall be between 0 and 2^(31) -
538 1.
539
540 3. BGP implementations should allow an AS to ignore routes with
541 certain AS's in the AS_PATH path attribute. Such function can
542 be implemented by using the technique outlined in [<a href="#ref-2" title="&quot;Models of Policy Based Routing&quot;">2</a>], and by
543 assigning "infinity" as "weights" for such AS's. The route
544 selection process must ignore routes that have "weight" equal
545 to "infinity".
546
547 <span class="h2"><a class="selflink" id="section-8" href="#section-8">8</a>. Interaction with other exterior routing protocols</span>
548
549 The guidelines suggested in this section are consistent with the
550 guidelines presented in [<a href="#ref-3" title="&quot;Supernetting: an Address Assignment and Aggregation Strategy&quot;">3</a>].
551
552
553
554
555
556 <span class="grey">Rekhter &amp; Gross [Page 10]</span></pre>
557 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span>
558 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
559
560
561 An AS should advertise a minimal aggregate for its internal
562 destinations with respect to the amount of address space that it is
563 actually using. This can be used by administrators of non-BGP 4 AS's
564 to determine how many routes to explode from a single aggregate.
565
566 A route that carries the ATOMIC_AGGREGATE path attribute shall not be
567 exported into either BGP-3 or EGP2, unless such an exportation can be
568 accomplished without exploding the NLRI of the route.
569
570 <span class="h3"><a class="selflink" id="section-8.1" href="#section-8.1">8.1</a> Exchanging information with EGP2</span>
571
572 This document suggests the following guidelines for exchanging
573 routing information between BGP-4 and EGP2.
574
575 To provide for graceful migration, a BGP speaker may participate in
576 EGP2, as well as in BGP-4. Thus, a BGP speaker may receive IP
577 reachability information by means of EGP2 as well as by means of
578 BGP-4. The information received by EGP2 can be injected into BGP-4
579 with the ORIGIN path attribute set to 1. Likewise, the information
580 received via BGP-4 can be injected into EGP2 as well. In the latter
581 case, however, one needs to be aware of the potential information
582 explosion when a given IP prefix received from BGP-4 denotes a set of
583 consecutive A/B/C class networks. Injection of BGP-4 received NLRI
584 that denotes IP subnets requires the BGP speaker to inject the
585 corresponding network into EGP2. The local system shall provide
586 mechanisms to control the exchange of reachability information
587 between EGP2 and BGP-4. Specifically, a conformant implementation is
588 required to support all of the following options when injecting BGP-4
589 received reachability information into EGP2:
590
591 - inject default only (0.0.0.0); no export of any other NLRI
592
593 - allow controlled deaggregation, but only of specific routes;
594 allow export of non-aggregated NLRI
595
596 - allow export of only non-aggregated NLRI
597
598 The exchange of routing information via EGP2 between a BGP speaker
599 participating in BGP-4 and a pure EGP2 speaker may occur only at the
600 domain (autonomous system) boundaries.
601
602 <span class="h3"><a class="selflink" id="section-8.2" href="#section-8.2">8.2</a> Exchanging information with BGP-3</span>
603
604 This document suggests the following guidelines for exchanging
605 routing information between BGP-4 and BGP-3.
606
607 To provide for graceful migration, a BGP speaker may participate in
608 BGP-3, as well as in BGP-4. Thus, a BGP speaker may receive IP
609
610
611
612 <span class="grey">Rekhter &amp; Gross [Page 11]</span></pre>
613 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-12" ></span>
614 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
615
616
617 reachability information by means of BGP-3, as well as by means of
618 BGP-4.
619
620 A BGP speaker may inject the information received by BGP-4 into BGP-3
621 as follows.
622
623 If an AS_PATH attribute of a BGP-4 route carries AS_SET path
624 segments, then the AS_PATH attribute of the BGP-3 route shall be
625 constructed by treating the AS_SET segments as AS_SEQUENCE segments,
626 with the resulting AS_PATH being a single AS_SEQUENCE. While this
627 procedure loses set/sequence information, it doesn't affect
628 protection for routing loops suppression, but may affect policies, if
629 the policies are based on the content or ordering of the AS_PATH
630 attribute.
631
632 While injecting BGP-4 derived NLRI into BGP-3, one needs to be aware
633 of the potential information explosion when a given IP prefix denotes
634 a set of consecutive A/B/C class networks. Injection of BGP-4 derived
635 NLRI that denotes IP subnets requires the BGP speaker to inject the
636 corresponding network into BGP-3. The local system shall provide
637 mechanisms to control the exchange of routing information between
638 BGP-3 and BGP-4. Specifically, a conformant implementation is
639 required to support all of the following options when injecting BGP-4
640 received routing information into BGP-3:
641
642 - inject default only (0.0.0.0), no export of any other NLRI
643
644 - allow controlled deaggregation, but only of specific routes;
645 allow export of non-aggregated NLRI
646
647 - allow export of only non-aggregated NLRI
648
649 The exchange of routing information via BGP-3 between a BGP speaker
650 participating in BGP-4 and a pure BGP-3 speaker may occur only at
651 the autonomous system boundaries. Within a single autonomous system
652 BGP conversations between all the BGP speakers of that autonomous
653 system have to be either BGP-3 or BGP-4, but not a mixture.
654
655 <span class="h2"><a class="selflink" id="section-9" href="#section-9">9</a>. Operations over Switched Virtual Circuits</span>
656
657 When using BGP over Switched Virtual Circuit (SVC) subnetworks it may
658 be desirable to minimize traffic generated by BGP. Specifically, it
659 may be desirable to eliminate traffic associated with periodic
660 KEEPALIVE messages. BGP includes a mechanism for operation over
661 switched virtual circuit (SVC) services which avoids keeping SVCs
662 permanently open and allows it to eliminates periodic sending of
663 KEEPALIVE messages.
664
665
666
667
668 <span class="grey">Rekhter &amp; Gross [Page 12]</span></pre>
669 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-13" ></span>
670 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
671
672
673 This section describes how to operate without periodic KEEPALIVE
674 messages to minimise SVC usage when using an intelligent SVC circuit
675 manager. The proposed scheme may also be used on "permanent"
676 circuits, which support a feature like link quality monitoring or
677 echo request to determine the status of link connectivity.
678
679 The mechanism described in this section is suitable only between the
680 BGP speakers that are directly connected over a common virtual
681 circuit.
682
683 <span class="h3"><a class="selflink" id="section-9.1" href="#section-9.1">9.1</a> Establishing a BGP Connection</span>
684
685 The feature is selected by specifying zero Hold Time in the OPEN
686 message.
687
688 <span class="h3"><a class="selflink" id="section-9.2" href="#section-9.2">9.2</a> Circuit Manager Properties</span>
689
690 The circuit manager must have sufficient functionality to be able to
691 compensate for the lack of periodic KEEPALIVE messages:
692
693 - It must be able to determine link layer unreachability in a
694 predictable finite period of a failure occurring.
695
696 - On determining unreachability it should:
697
698 - start a configurable dead timer (comparable to a
699 typical Hold timer value).
700
701 - attempt to re-establish the Link Layer connection.
702
703
704 - If the dead timer expires it should:
705
706 - send an internal circuit DEAD indication to TCP.
707
708 - If the connection is re-established it should:
709
710 - cancel the dead timer.
711
712 - send an internal circuit UP indication to TCP.
713
714 <span class="h3"><a class="selflink" id="section-9.3" href="#section-9.3">9.3</a> TCP Properties</span>
715
716 A small modification must be made to TCP to process internal
717 notifications from the circuit manager:
718
719 - DEAD: Flush transmit queue and abort TCP connection.
720
721
722
723
724 <span class="grey">Rekhter &amp; Gross [Page 13]</span></pre>
725 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-14" ></span>
726 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
727
728
729 - UP: Transmit any queued data or allow an outgoing TCP call to
730 proceed.
731
732 <span class="h3"><a class="selflink" id="section-9.4" href="#section-9.4">9.4</a> Combined Properties</span>
733
734 Some implementations may not be able to guarantee that the BGP
735 process and the circuit manager will operate as a single entity; i.e.
736 they can have a separate existence when the other has been stopped or
737 has crashed.
738
739 If this is the case, a periodic two-way poll between the BGP process
740 and the circuit manager should be implemented. If the BGP process
741 discovers the circuit manager has gone away it should close all
742 relevant TCP connections. If the circuit manager discovers the BGP
743 process has gone away it should close all its connections associated
744 with the BGP process and reject any further incoming connections.
745
746 <span class="h2"><a class="selflink" id="section-10" href="#section-10">10</a>. Conclusion</span>
747
748 The BGP protocol provides a high degree of control and flexibility
749 for doing interdomain routing while enforcing policy and performance
750 constraints and avoiding routing loops. The guidelines presented here
751 will provide a starting point for using BGP to provide more
752 sophisticated and manageable routing in the Internet as it grows.
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780 <span class="grey">Rekhter &amp; Gross [Page 14]</span></pre>
781 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-15" ></span>
782 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
783
784
785 <span class="h2"><a class="selflink" id="appendix-A" href="#appendix-A">Appendix A</a>. The Interaction of BGP and an IGP</span>
786
787 This section outlines methods by which BGP can exchange routing
788 information with an IGP. The methods outlined here are not proposed
789 as part of the standard BGP usage at this time. These methods are
790 outlined for information purposes only. Implementors may want to
791 consider these methods when importing IGP information.
792
793 This is general information that applies to any generic IGP.
794
795 Interaction between BGP and any specific IGP is outside the scope of
796 this section. Methods for specific IGP's should be proposed in
797 separate documents. Methods for specific IGP's could be proposed for
798 standard usage in the future.
799
800 Overview
801
802 By definition, all transit AS's must be able to carry traffic which
803 originates from and/or is destined to locations outside of that AS.
804 This requires a certain degree of interaction and coordination
805 between BGP and the Interior Gateway Protocol (IGP) used by that
806 particular AS. In general, traffic originating outside of a given AS
807 is going to pass through both interior gateways (gateways that
808 support the IGP only) and border gateways (gateways that support both
809 the IGP and BGP). All interior gateways receive information about
810 external routes from one or more of the border gateways of the AS via
811 the IGP.
812
813 Depending on the mechanism used to propagate BGP information within a
814 given AS, special care must be taken to ensure consistency between
815 BGP and the IGP, since changes in state are likely to propagate at
816 different rates across the AS. There may be a time window between the
817 moment when some border gateway (A) receives new BGP routing
818 information which was originated from another border gateway (B)
819 within the same AS, and the moment the IGP within this AS is capable
820 of routing transit traffic to that border gateway (B). During that
821 time window, either incorrect routing or "black holes" can occur.
822
823 In order to minimize such routing problems, border gateway (A) should
824 not advertise to any of its external peers a route to some set of
825 exterior destinations associated with a given address prefix X via
826 border gateway (B) until all the interior gateways within the AS are
827 ready to route traffic destined to these destinations via the correct
828 exit border gateway (B). In other words, interior routing should
829 converge on the proper exit gateway before/advertising routes via
830 that exit gateway to external peers.
831
832
833
834
835
836 <span class="grey">Rekhter &amp; Gross [Page 15]</span></pre>
837 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-16" ></span>
838 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
839
840
841 <span class="h3"><a class="selflink" id="appendix-A.2" href="#appendix-A.2">A.2</a> Methods for Achieving Stable Interactions</span>
842
843 The following discussion outlines several techniques capable of
844 achieving stable interactions between BGP and the IGP within an
845 Autonomous System.
846
847 <span class="h4"><a class="selflink" id="appendix-A.2.1" href="#appendix-A.2.1">A.2.1</a> Propagation of BGP Information via the IGP</span>
848
849 While BGP can provide its own mechanism for carrying BGP information
850 within an AS, one can also use an IGP to transport this information,
851 as long as the IGP supports complete flooding of routing information
852 (providing the mechanism to distribute the BGP information) and one
853 pass convergence (making the mechanism effectively atomic). If an IGP
854 is used to carry BGP information, then the period of
855 desynchronization described earlier does not occur at all, since BGP
856 information propagates within the AS synchronously with the IGP, and
857 the IGP converges more or less simultaneously with the arrival of the
858 new routing information. Note that the IGP only carries BGP
859 information and should not interpret or process this information.
860
861 <span class="h4"><a class="selflink" id="appendix-A.2.2" href="#appendix-A.2.2">A.2.2</a> Tagged Interior Gateway Protocol</span>
862
863 Certain IGPs can tag routes exterior to an AS with the identity of
864 their exit points while propagating them within the AS. Each border
865 gateway should use identical tags for announcing exterior routing
866 information (received via BGP) both into the IGP and when propagating
867 this information to other internal peers (peers within the same AS).
868 Tags generated by a border gateway must uniquely identify that
869 particular border gateway--different border gateways must use
870 different tags.
871
872 All Border Gateways within a single AS must observe the following two
873 rules:
874
875 1. Information received from an internal peer by a border gateway
876 A declaring a set of destination associated with a given
877 address prefix to be unreachable must immediately be propagated
878 to all of the external peers of A.
879
880 2. Information received from an internal peer by a border gateway
881 A about a set of reachable destinations associated with a given
882 address prefix X cannot be propagated to any of the external
883 peers of A unless/until A has an IGP route to the set of
884 destinations covered by X and both the IGP and the BGP routing
885 information have identical tags.
886
887 These rules guarantee that no routing information is announced
888 externally unless the IGP is capable of correctly supporting it. It
889
890
891
892 <span class="grey">Rekhter &amp; Gross [Page 16]</span></pre>
893 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-17" ></span>
894 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
895
896
897 also avoids some causes of "black holes".
898
899 One possible method for tagging BGP and IGP routes within an AS is to
900 use the IP address of the exit border gateway announcing the exterior
901 route into the AS. In this case the "gateway" field in the BGP UPDATE
902 message is used as the tag.
903
904 An alternate method for tagging BGP and IGP routes is to have BGP and
905 the IGP agree on a router ID. In this case, the router ID is
906 available to all BGP (version 3 or higher) speakers. Since this ID
907 is already unique it can be used directly as the tag in the IGP.
908
909 <span class="h4"><a class="selflink" id="appendix-A.2.3" href="#appendix-A.2.3">A.2.3</a> Encapsulation</span>
910
911 Encapsulation provides the simplest (in terms of the interaction
912 between the IGP and BGP) mechanism for carrying transit traffic
913 across the AS. In this approach, transit traffic is encapsulated
914 within an IP datagram addressed to the exit gateway. The only
915 requirement imposed on the IGP by this approach is that it should be
916 capable of supporting routing between border gateways within the same
917 AS.
918
919 The address of the exit gateway A for some exterior destination X is
920 specified in the BGP identifier field of the BGP OPEN message
921 received from gateway A (via BGP) by all other border gateways within
922 the same AS. In order to route traffic to destination X, each border
923 gateway within the AS encapsulates it in datagrams addressed to
924 gateway A. Gateway A then performs decapsulation and forwards the
925 original packet to the proper gateway in another AS.
926
927 Since encapsulation does not rely on the IGP to carry exterior
928 routing information, no synchronization between BGP and the IGP is
929 required.
930
931 Some means of identifying datagrams containing encapsulated IP, such
932 as an IP protocol type code, must be defined if this method is to be
933 used.
934
935 Note that, if a packet to be encapsulated has length that is very
936 close to the MTU, that packet would be fragmented at the gateway that
937 performs encapsulation.
938
939 <span class="h4"><a class="selflink" id="appendix-A.2.4" href="#appendix-A.2.4">A.2.4</a> Pervasive BGP</span>
940
941 If all routers in an AS are BGP speakers, then there is no need to
942 have any interaction between BGP and an IGP. In such cases, all
943 routers in the AS already have full information of all BGP routes.
944 The IGP is then only used for routing within the AS, and no BGP
945
946
947
948 <span class="grey">Rekhter &amp; Gross [Page 17]</span></pre>
949 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-18" ></span>
950 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
951
952
953 routes are imported into the IGP.
954
955 For routers to operate in this fashion, they must be able to perform
956 a recursive lookup in their routing table. The first lookup will use
957 a BGP route to establish the exit router, while the second lookup
958 will determine the IGP path to the exit router.
959
960 Since the IGP carries no external information in this scenario, all
961 routers in the AS will have converged as soon as all BGP speakers
962 have new information about this route. Since there is no need to
963 delay for the IGP to converge, an implementation may advertise these
964 routes without further delay due to the IGP.
965
966 <span class="h4"><a class="selflink" id="appendix-A.2.5" href="#appendix-A.2.5">A.2.5</a> Other Cases</span>
967
968 There may be AS's with IGPs which can neither carry BGP information
969 nor tag exterior routes (e.g., RIP). In addition, encapsulation may
970 be either infeasible or undesirable. In such situations, the
971 following two rules must be observed:
972
973 1. Information received from an internal peer by a border gateway
974 A declaring a destination to be unreachable must immediately be
975 propagated to all of the external peers of A.
976
977 2. Information received from an internal peer by a border gateway
978 A about a reachable destination X cannot be propagated to any
979 of the external peers of A unless A has an IGP route to X and
980 sufficient time has passed for the IGP routes to have
981 converged.
982
983 The above rules present necessary (but not sufficient) conditions for
984 propagating BGP routing information to other AS's. In contrast to
985 tagged IGPs, these rules cannot ensure that interior routes to the
986 proper exit gateways are in place before propagating the routes to
987 other AS's.
988
989 If the convergence time of an IGP is less than some small value X,
990 then the time window during which the IGP and BGP are unsynchronized
991 is less than X as well, and the whole issue can be ignored at the
992 cost of transient periods (of less than length X) of routing
993 instability. A reasonable value for X is a matter for further study,
994 but X should probably be less than one second.
995
996 If the convergence time of an IGP cannot be ignored, a different
997 approach is needed. Mechanisms and techniques which might be
998 appropriate in this situation are subjects for further study.
999
1000
1001
1002
1003
1004 <span class="grey">Rekhter &amp; Gross [Page 18]</span></pre>
1005 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-19" ></span>
1006 <span class="grey"><a href="./rfc1772">RFC 1772</a> BGP-4 Application March 1995</span>
1007
1008
1009 References
1010
1011 [<a id="ref-1">1</a>] Rekhter Y., and T. Li, "A Border Gateway Protocol 4 (BGP-4), <a href="./rfc1771">RFC</a>
1012 <a href="./rfc1771">1771</a>, T.J. Watson Research Center, IBM Corp., cisco Systems,
1013 March 1995.
1014
1015 [<a id="ref-2">2</a>] Braun, H-W., "Models of Policy Based Routing", <a href="./rfc1104">RFC 1104</a>,
1016 Merit/NSFNET, June 1989.
1017
1018 [<a id="ref-3">3</a>] Fuller, V., Li, T., Yu, J., and K. Varadhan, "Supernetting: an
1019 Address Assignment and Aggregation Strategy", <a href="./rfc1519">RFC1519</a>, BARRNet,
1020 cisco, MERIT, OARnet, September 1993.
1021
1022 Security Considerations
1023
1024 Security issues are not discussed in this memo.
1025
1026 Authors' Addresses
1027
1028 Yakov Rekhter
1029 T.J. Watson Research Center IBM Corporation
1030 P.O. Box 704, Office H3-D40
1031 Yorktown Heights, NY 10598
1032
1033 Phone: +1 914 784 7361
1034 EMail: [email protected]
1035
1036
1037 Phill Gross
1038 MCI Data Services Division
1039 2100 Reston Parkway, Room 6001,
1040 Reston, VA 22091
1041
1042 Phone: +1 703 715 7432
1043 EMail: [email protected]
1044
1045
1046 IETF IDR WG mailing list: [email protected]
1047 To be added: [email protected]
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060 Rekhter &amp; Gross [Page 19]
1061 </pre>