]> Deprecation of AS_SET and AS_CONFED_SET in BGP Google, Inc.
1600 Amphitheatre Parkway Mountain View CA 94043 United States of America +1 571 748 4373 warren@kumari.net
USA NIST
100 Bureau Drive Gaithersburg MD 20899 United States of America +1 301 975 3973 ksriram@nist.gov
USA NIST
100 Bureau Drive Gaithersburg MD 20899 United States of America +1 301 975 3259 lilia.hannachi@nist.gov
Juniper Networks, Inc.
1133 Innovation Way Sunnyvale CA 94089 United States of America jhaas@juniper.net
RTG idr BCP 172 (i.e., RFC 6472) recommends not using AS_SET and AS_CONFED_SET AS_PATH segment types in the Border Gateway Protocol (BGP). This document advances that recommendation to a standards requirement in BGP; it prohibits the use of the AS_SET and AS_CONFED_SET path segment types in the AS_PATH. This is done to simplify the design and implementation of BGP and to make the semantics of the originator of a BGP route clearer. This will also simplify the design, implementation, and deployment of various BGP security mechanisms. This document updates RFC 4271 by deprecating the origination of BGP routes with AS_SET (Type 1 AS_PATH segment) and updates RFC 5065 by deprecating the origination of BGP routes with AS_CONFED_SET (Type 4 AS_PATH segment). Finally, it obsoletes RFC 6472.
Introduction recommends not using AS_SET and AS_CONFED_SET AS_PATH path segment types in the Border Gateway Protocol (BGP). This document advances the BCP recommendation to a standards requirement in BGP; it prohibits the use of the AS_SET and AS_CONFED_SET types of path segments in the AS_PATH. The purpose is to simplify the design and implementation of BGP and to make the semantics of the originator of a BGP route clearer. This will also simplify the design, implementation, and deployment of various BGP security mechanisms. In particular, the prohibition of AS_SETs and AS_CONFED_SETs removes any ambiguity about the origin AS in RPKI-based Route Origin Validation (RPKI-ROV) . The AS_SET path segment in the AS_PATH attribute (Sections and of ) is created by a router that is performing route aggregation and contains an unordered set of Autonomous Systems (ASes) that contributing prefixes in the aggregate have traversed. The AS_CONFED_SET path segment in the AS_PATH attribute is created by a router that is performing route aggregation and contains an unordered set of Member AS Numbers in the local confederation that contributing prefixes in the aggregate have traversed. It is very similar to an AS_SET but is used within a confederation. By performing aggregation, a router is combining multiple BGP routes for more specific destinations into a new route for a less specific destination (see ). Aggregation may blur the semantics of the origin AS for the prefix being announced by producing an AS_SET or AS_CONFED_SET. Such sets can cause operational issues, such as not being able to authenticate a route origin for the aggregate prefix in new BGP security technologies such as those that take advantage of X.509 extensions for IP addresses and AS identifiers (see , , , , and ). This could result in reachability problems for the destinations covered by the aggregated prefix. From analysis of historical Internet routing data, it is apparent that aggregation that involves AS_SETs is very seldom used in practice on the public Internet (see ). When it is used, it is often used incorrectly; only a single AS in the AS_SET is the most common case . Also, very often the same AS appears in the AS_SEQUENCE and the AS_SET in the BGP update. The occurrence of reserved AS numbers is also somewhat frequent.
Requirements Language The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.
Updates to the Requirements of RFCs 4271 and 5065 Unless explicitly configured by a network operator to do otherwise (e.g., during a transition phase), BGP speakers:
  • MUST NOT advertise BGP UPDATE messages containing AS_SETs or AS_CONFED_SETs and
  • MUST use the "treat-as-withdraw" error handling behavior per upon reception of BGP UPDATE messages containing AS_SETs or AS_CONFED_SETs in the AS_PATH or AS4_PATH .
Per the above specifications, this document updates and by deprecating AS_SET (see ) and AS_CONFED_SET (see ), respectively.
Treatment of Routes with AS_SET in RPKI-Based BGP Security Resource Public Key Infrastructure (RPKI) uses X.509 extensions for IP addresses and AS identifiers . RPKI-ROV is a BGP security technology that never allows a route with AS_SET to be considered Valid. BGPsec and Autonomous System Provider Authorization (ASPA) are also BGP security technologies based on RPKI. BGPsec does not support AS_SETs. In ASPA-based AS_PATH verification, a route with AS_SET is always considered Invalid and hence ineligible for route selection.
BGP AS_PATH "Brief" Aggregation Sections and of describe BGP aggregation procedures. describes a generally less utilized "Complex AS_PATH Aggregation" procedure. describes the ATOMIC_AGGREGATE Path Attribute and notes that:
When a BGP speaker aggregates several routes for the purpose of advertisement to a particular peer, the AS_PATH of the aggregated route normally includes an AS_SET formed from the set of ASes from which the aggregate was formed. In many cases, the network administrator can determine if the aggregate can safely be advertised without the AS_SET, and without forming route loops. If an aggregate excludes at least some of the AS numbers present in the AS_PATH of the routes that are aggregated as a result of dropping the AS_SET, the aggregated route, when advertised to the peer, SHOULD include the ATOMIC_AGGREGATE attribute.
When BGP AS_PATH aggregation is done according to the procedures in , and any resulting AS_SETs are discarded, it is typically referred to as "brief" aggregation in implementations. That terminology is adopted here: In this document, brief aggregation refers to what is described in this section, in contrast to consistent brief aggregation as described in . Brief aggregation results in an AS_PATH that has the following property (from ):
[D]etermine the longest leading sequence of tuples (as defined above) common to all the AS_PATH attributes of the routes to be aggregated. Make this sequence the leading sequence of the aggregated AS_PATH attribute.
The ATOMIC_AGGREGATE Path Attribute is subsequently attached to the BGP route, if AS_SETs are dropped.
Issues with "Brief" AS_PATH Aggregation and RPKI-ROV While brief AS_PATH aggregation has the desirable property of not containing AS_SETs, the resulting aggregated AS_PATH may contain an unpredictable origin AS. This is because the aggregating AS may be different from the purported origin AS (for the aggregate), which may vary as explained below. Such unpredictable origin ASes may result in RPKI-ROV validation issues:
  • Depending on the contributing routes to the aggregate route, the resulting origin AS may vary.
  • The presence of expected contributing routes may be unpredictable due to route availability from BGP neighbors.
  • In the presence of such varying origin ASes, it would be necessary for the resource holder to register ROAs for each potential origin AS that may result from the expected aggregated AS_PATHs.
Recommendations to Mitigate Unpredictable AS_PATH Origins for RPKI-ROV Purposes To ensure a consistent BGP origin AS is announced for aggregate BGP routes for implementations of "brief" BGP aggregation, the implementation MUST be configured to truncate the AS_PATH after the right-most instance of the desired origin AS for the aggregate. The desired origin AS could be the aggregating AS itself. A ROA would be necessary for the aggregate prefix with the desired origin AS. This form of brief aggregation is referred to as "consistent brief" BGP aggregation. If the resulting AS_PATH would be truncated from the otherwise expected result of BGP AS_PATH aggregation (an AS_SET would not be generated and possibly some ASes are removed from the "longest leading sequence" of ASes), the ATOMIC_AGGREGATE Path Attribute SHOULD be attached. This is consistent with the intent of .
Operational Considerations This section provides advice to operators regarding deployment and configuration.
Implementing Consistent Brief Aggregation When aggregating prefixes, network operators MUST use consistent brief aggregation as described in . In consistent brief aggregation, the AGGREGATOR and ATOMIC_AGGREGATE Path Attributes are included, but the AS_PATH does not have AS_SET or AS_CONFED_SET path segment types. See for examples of brief aggregation while keeping the origin AS unambiguous and generating appropriate ROAs.
Not Advertising Aggregate Routes to Contributing ASes An aggregate prefix SHOULD NOT be announced to the contributing ASes. Instead, more specific prefixes (from the aggregate) SHOULD be announced to each contributing AS, excluding any that were learned from the contributing AS in consideration. See for an example of this filtering policy.
Mitigating Forwarding Loops When both less specific and more specific destinations are present, it's possible to create forwarding loops between networks, as discussed in . As a reminder, Rule #2 in requires that BGP implementations performing aggregation discard packets that match the aggregate route but do not match any of the more specific routes. Further discussion of forwarding loops and their relationship to AS_SETs can be found in .
Security Considerations This document deprecates the use of aggregation techniques that create AS_SETs or AS_CONFED_SETs. Obsoleting these path segment types from BGP and the removal of the related code from implementations would potentially decrease the attack surface for BGP. Deployments of new BGP security technologies (e.g., , , and ) benefit greatly if AS_SETs and AS_CONFED_SETs are not used in BGP.
IANA Considerations This document has no IANA actions.
References Normative References Recommendation for Not Using AS_SET and AS_CONFED_SET in BGP Informative References Special-Purpose Autonomous System (AS) Numbers IANA Detailed analysis of AS_SETs in BGP updates commit ef3f4a9
Example of Route Filtering for Aggregate Routes and Their Contributors The illustration presented below shows how an AS_SET is not used when aggregating and how data plane route loops are avoided. Consider that p1/24 (from AS 64501), p2/24 (from AS 64502), p3/24 (from AS 64503), and p4/24 (from AS 64504) are aggregated by AS 64505 to p/22. AS_SET is not used with the aggregate p/22 but AGGREGATOR and ATOMIC AGGREGATE are used. Data plane route loops are avoided by not announcing the aggregate p/22 to the contributing ASes, i.e., AS 64501, AS 64502, AS 64503, and AS 64504. Instead, as further illustrated, p1/24, p2/24, and p4/24 are announced to AS 64503. The routing tables (post aggregation) of each of the ASes are depicted in the diagram below. ( ) <--+ | | ( AS64505 ) | +----------------> ( ) <----------------+ p/22 | V AS 64501 AS 64502 ========================== ========================== p1/24 AS_PATH "" p1/24 AS_PATH "64505 64501" p2/24 AS_PATH "64505 64502" p2/24 AS_PATH "" p3/24 AS_PATH "64505 64503" p3/24 AS_PATH "64505 64503" p4/24 AS_PATH "64505 64504" p4/24 AS_PATH "64505 64504" AS 64503 AS 64504 ========================== ========================== p1/24 AS_PATH "64505 64501" p1/24 AS_PATH "64505 64501" p2/24 AS_PATH "64505 64502" p2/24 AS_PATH "64505 64502" p3/24 AS_PATH "" p3/24 AS_PATH "64505 64503" p4/24 AS_PATH "64505 64504" p4/24 AS_PATH "" AS 64505 ========================== p/22 AS_PATH "" AGGREGATOR 64505 ATOMIC_AGGREGATE p1/24 AS_PATH "64501" p2/24 AS_PATH "64502" p3/24 AS_PATH "64503" p4/24 AS_PATH "64504"]]>
Examples of Consistent and Inconsistent BGP Origin AS Generated by Brief Aggregation The examples below illustrate how brief aggregation may result in an inconsistent origin AS. AS 64500 aggregates more specific routes into 192.0.2.0/24. Consider the following scenarios where brief aggregation is done by AS 64500 and what the resultant origin ASes would be. ( AS64500 )<-------------+ ( ) | | (announcing | aggregate 192.0.2.0/24) \/]]>
Scenario 1: First one route, then another, each with a fully disjoint AS_PATH Receive R1. Aggregate 192.0.2.0/24 AS_PATH "64501" Alternate "bug?": Aggregate 192.0.2.0/24 AS_PATH "[ 64501 ]" (Note: AS numbers within square brackets represent an AS_SET.) Receive R2. Aggregate 192.0.2.0/24 AS_PATH "[ 64501 64502 ]" If brief aggregation is in use, the AS_PATH would be truncated to the empty AS_PATH, "". The resulting AS_PATH is thus not stable and depends on the presence of specific routes.
Scenario 2: First one route, then another, and the AS_PATHs overlap at the origin AS Receive R1. Aggregate 192.0.2.0/24 AS_PATH "64501" Receive R4. Aggregate 192.0.2.0/24 AS_PATH "[ 64504 64501 ]" If brief aggregation is in use, the AS_PATH is truncated to "". The resulting AS_PATH is thus not stable and depends on the presence of specific routes.
Scenario 3: First one route, then another, and the AS_PATHs overlap at the neighbor AS Receive R3. Aggregate 192.0.2.0/24 AS_PATH "64504 64501" Receive R4. Aggregate 192.0.2.0/24 AS_PATH "64504 [ 64501 64502 ]" If brief aggregation is in use, the AS_PATH is truncated to "64504". The resulting AS_PATH is thus not stable and depends on the presence of specific routes.
Achieving Consistent Origin AS During Aggregation In the three scenarios above, the aggregating AS 64500 is using brief aggregation. This results in inconsistent origin ASes as the contributing routes are learned. This motivates the "consistent brief" BGP aggregation mentioned in and discussed further with examples below. The trivial solution to addressing the issue is to simply discard all of the ASes for the contributing routes. In simple BGP aggregation topologies, this is likely the correct thing to do. The AS originating the aggregate, 192.0.2.0/24 in this example, is likely the resource holder for the route in question. In such a case, simply originating the route to its BGP upstream neighbors in the Internet with its own AS, 64500, means that a consistent ROA could be registered in the RPKI for this prefix. This satisfies the need for a consistent (unambiguous) origin AS. If the contributing ASes are themselves multihomed to the Internet outside of their connections to AS 64500, then additional ROAs would need to be created for each of the more specific prefixes. In more complex proxy aggregation scenarios, there may be a desire to permit some stable (i.e., common) portion of the contributing AS_PATHs to be kept in the aggregate route. Consider the case for Scenario 3, where the neighbor AS is the same for both R3 and R4 -- AS 64504. In such a case, an implementation may permit the aggregate's brief AS_PATH to be "64504", and a ROA would be created for the aggregate prefix with 64504 as the origin AS.
Discussion on Forwarding Loops and AS_SETs Although BGP-4 was designed to carry Classless Inter-Domain Routing (CIDR) routes, does not discuss the installation of "discard" or "null" routes when implementing its aggregation procedures. Implementations could originate an aggregate prefix without a covering route for a more specific prefix (subsumed by the aggregate prefix) present in the local routing table. When aggregating more specific routes according to the aggregation procedures of , the aggregating BGP speaker will place contributing routes into the generated AS_PATH, perhaps using AS_SETs. As a result, a contributing AS will not install the aggregated route into its RIB since the route is an AS_PATH loop. This provides a form of protection against forwarding loops created by BGP aggregation. When brief aggregation methods are used, a BGP speaker may receive a route containing a less specific destination covering a local more specific destination and install it in its routing table since it is not prevented from doing so by BGP AS_PATH loop detection. This gives rise to the possibility of forwarding loops. To help prevent forwarding loops, it is critical to adhere to the following:
  1. Rule #2 in :
A router that generates an aggregate route for multiple, more-specific routes must discard packets that match the aggregate route, but not any of the more-specific routes. In other words, the "next hop" for the aggregate route should be the null destination.
  1. Not advertising aggregate routes to contributing ASes as specified in of this document (also see ).
Acknowledgements The authors would like to thank , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , and for comments and suggestions. The comments and suggestions received from the IESG reviewers are also much appreciated.