BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)
How a few BGP routers can relay routes inside an AS so that not every router needs a session with every other.
Bates, T., Chen, E., and R. Chandra, "BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)", RFC 4456, DOI 10.17487/RFC4456, April 2006, <https://www.rfc-editor.org/info/rfc4456>.
@misc{rfc4456,
series = {Request for Comments},
number = 4456,
howpublished = {RFC 4456},
publisher = {RFC Editor},
doi = {10.17487/RFC4456},
url = {https://www.rfc-editor.org/info/rfc4456},
author = {T. Bates and E. Chen and R. Chandra},
title = {{BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)}},
pagetotal = 12,
year = 2006,
month = apr,
} RFC text
Contents
Network Working Group T. Bates
Request for Comments: 4456 E. Chen
Obsoletes: 2796, 1966 Cisco Systems
Category: Standards Track R. Chandra
Sonoa Systems
April 2006
BGP Route Reflection:
An Alternative to Full Mesh Internal BGP (IBGP)
Status of This Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2006).
Abstract
The Border Gateway Protocol (BGP) is an inter-autonomous system
routing protocol designed for TCP/IP internets. Typically, all BGP
speakers within a single AS must be fully meshed so that any external
routing information must be re-distributed to all other routers
within that Autonomous System (AS). This represents a serious
scaling problem that has been well documented with several
alternatives proposed.
This document describes the use and design of a method known as
"route reflection" to alleviate the need for "full mesh" Internal BGP
(IBGP).
This document obsoletes RFC 2796 and RFC 1966.
Table of Contents
1. Introduction ....................................................2
2. Specification of Requirements ...................................2
3. Design Criteria .................................................3
4. Route Reflection ................................................3
5. Terminology and Concepts ........................................4
6. Operation .......................................................5
7. Redundant RRs ...................................................6
8. Avoiding Routing Information Loops ..............................6
9. Impact on Route Selection .......................................7
10. Implementation Considerations ..................................7
11. Configuration and Deployment Considerations ....................7
12. Security Considerations ........................................8
13. Acknowledgements ...............................................9
14. References .....................................................9
14.1. Normative References ......................................9
14.2. Informative References ....................................9
Appendix A: Comparison with RFC 2796 ..............................10
Appendix B: Comparison with RFC 1966 ..............................10
1. Introduction
Typically, all BGP speakers within a single AS must be fully meshed
and any external routing information must be re-distributed to all
other routers within that AS. For n BGP speakers within an AS that
requires to maintain n*(n-1)/2 unique Internal BGP (IBGP) sessions.
This "full mesh" requirement clearly does not scale when there are a
large number of IBGP speakers each exchanging a large volume of
routing information, as is common in many of today's networks.
This scaling problem has been well documented, and a number of
proposals have been made to alleviate this [2,3]. This document
represents another alternative in alleviating the need for a "full
mesh" and is known as "route reflection". This approach allows a BGP
speaker (known as a "route reflector") to advertise IBGP learned
routes to certain IBGP peers. It represents a change in the commonly
understood concept of IBGP, and the addition of two new optional
non-transitive BGP attributes to prevent loops in routing updates.
This document obsoletes RFC 2796 [6] and RFC 1966 [4].
2. Specification of Requirements
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [7].
3. Design Criteria
Route reflection was designed to satisfy the following criteria.
o Simplicity
Any alternative must be simple to configure and easy to
understand.
o Easy Transition
It must be possible to transition from a full-mesh
configuration without the need to change either topology or AS.
This is an unfortunate management overhead of the technique
proposed in [3].
o Compatibility
It must be possible for noncompliant IBGP peers to continue to
be part of the original AS or domain without any loss of BGP
routing information.
These criteria were motivated by operational experiences of a very
large and topology-rich network with many external connections.
4. Route Reflection
The basic idea of route reflection is very simple. Let us consider
the simple example depicted in Figure 1 below.
+-------+ +-------+
| | IBGP | |
| RTR-A |--------| RTR-B |
| | | |
+-------+ +-------+
\ /
IBGP \ ASX / IBGP
\ /
+-------+
| |
| RTR-C |
| |
+-------+
Figure 1: Full-Mesh IBGP
In ASX, there are three IBGP speakers (routers RTR-A, RTR-B, and
RTR-C). With the existing BGP model, if RTR-A receives an external
route and it is selected as the best path it must advertise the
external route to both RTR-B and RTR-C. RTR-B and RTR-C (as IBGP
speakers) will not re-advertise these IBGP learned routes to other
IBGP speakers.
If this rule is relaxed and RTR-C is allowed to advertise IBGP
learned routes to IBGP peers, then it could re-advertise (or reflect)
the IBGP routes learned from RTR-A to RTR-B and vice versa. This
would eliminate the need for the IBGP session between RTR-A and RTR-B
as shown in Figure 2 below.
+-------+ +-------+
| | | |
| RTR-A | | RTR-B |
| | | |
+-------+ +-------+
\ /
IBGP \ ASX / IBGP
\ /
+-------+
| |
| RTR-C |
| |
+-------+
Figure 2: Route Reflection IBGP
The route reflection scheme is based upon this basic principle.
5. Terminology and Concepts
We use the term "route reflection" to describe the operation of a BGP
speaker advertising an IBGP learned route to another IBGP peer. Such
a BGP speaker is said to be a "route reflector" (RR), and such a
route is said to be a reflected route.
The internal peers of an RR are divided into two groups:
1) Client peers
2) Non-Client peers
An RR reflects routes between these groups, and may reflect routes
among client peers. An RR along with its client peers form a
cluster. The Non-Client peer must be fully meshed but the Client
peers need not be fully meshed. Figure 3 depicts a simple example
outlining the basic RR components using the terminology noted above.
/ - - - - - - - - - - - - - -
| Cluster |
+-------+ +-------+
| | | | | |
| RTR-A | | RTR-B |
| |Client | |Client | |
+-------+ +-------+
| \ / |
IBGP \ / IBGP
| \ / |
+-------+
| | | |
| RTR-C |
| | RR | |
+-------+
| / \ |
- - - - - /- - -\- - - - - - /
IBGP / \ IBGP
+-------+ +-------+
| RTR-D | IBGP | RTR-E |
| Non- |---------| Non- |
|Client | |Client |
+-------+ +-------+
Figure 3: RR Components
6. Operation
When an RR receives a route from an IBGP peer, it selects the best
path based on its path selection rule. After the best path is
selected, it must do the following depending on the type of peer it
is receiving the best path from
1) A route from a Non-Client IBGP peer:
Reflect to all the Clients.
2) A route from a Client peer:
Reflect to all the Non-Client peers and also to the Client
peers. (Hence the Client peers are not required to be fully
meshed.)
An Autonomous System could have many RRs. An RR treats other RRs
just like any other internal BGP speakers. An RR could be configured
to have other RRs in a Client group or Non-client group.
In a simple configuration, the backbone could be divided into many
clusters. Each RR would be configured with other RRs as Non-Client
peers (thus all the RRs will be fully meshed). The Clients will be
configured to maintain IBGP session only with the RR in their
cluster. Due to route reflection, all the IBGP speakers will receive
reflected routing information.
It is possible in an Autonomous System to have BGP speakers that do
not understand the concept of route reflectors (let us call them
conventional BGP speakers). The route reflector scheme allows such
conventional BGP speakers to coexist. Conventional BGP speakers
could be members of either a Non-Client group or a Client group.
This allows for an easy and gradual migration from the current IBGP
model to the route reflection model. One could start creating
clusters by configuring a single router as the designated RR and
configuring other RRs and their clients as normal IBGP peers.
Additional clusters can be created gradually.
7. Redundant RRs
Usually, a cluster of clients will have a single RR. In that case,
the cluster will be identified by the BGP Identifier of the RR.
However, this represents a single point of failure so to make it
possible to have multiple RRs in the same cluster, all RRs in the
same cluster can be configured with a 4-byte CLUSTER_ID so that an RR
can discard routes from other RRs in the same cluster.
8. Avoiding Routing Information Loops
When a route is reflected, it is possible through misconfiguration to
form route re-distribution loops. The route reflection method
defines the following attributes to detect and avoid routing
information loops:
ORIGINATOR_ID
ORIGINATOR_ID is a new optional, non-transitive BGP attribute of Type
code 9. This attribute is 4 bytes long and it will be created by an
RR in reflecting a route. This attribute will carry the BGP
Identifier of the originator of the route in the local AS. A BGP
speaker SHOULD NOT create an ORIGINATOR_ID attribute if one already
exists. A router that recognizes the ORIGINATOR_ID attribute SHOULD
ignore a route received with its BGP Identifier as the ORIGINATOR_ID.
CLUSTER_LIST
CLUSTER_LIST is a new, optional, non-transitive BGP attribute of Type
code 10. It is a sequence of CLUSTER_ID values representing the
reflection path that the route has passed.
When an RR reflects a route, it MUST prepend the local CLUSTER_ID to
the CLUSTER_LIST. If the CLUSTER_LIST is empty, it MUST create a new
one. Using this attribute an RR can identify if the routing
information has looped back to the same cluster due to
misconfiguration. If the local CLUSTER_ID is found in the
CLUSTER_LIST, the advertisement received SHOULD be ignored.
9. Impact on Route Selection
The BGP Decision Process Tie Breaking rules (Sect. 9.1.2.2, [1]) are
modified as follows:
If a route carries the ORIGINATOR_ID attribute, then in Step f)
the ORIGINATOR_ID SHOULD be treated as the BGP Identifier of the
BGP speaker that has advertised the route.
In addition, the following rule SHOULD be inserted between Steps
f) and g): a BGP Speaker SHOULD prefer a route with the shorter
CLUSTER_LIST length. The CLUSTER_LIST length is zero if a route
does not carry the CLUSTER_LIST attribute.
10. Implementation Considerations
Care should be taken to make sure that none of the BGP path
attributes defined above can be modified through configuration when
exchanging internal routing information between RRs and Clients and
Non-Clients. Their modification could potentially result in routing
loops.
In addition, when a RR reflects a route, it SHOULD NOT modify the
following path attributes: NEXT_HOP, AS_PATH, LOCAL_PREF, and MED.
Their modification could potentially result in routing loops.
11. Configuration and Deployment Considerations
The BGP protocol provides no way for a Client to identify itself
dynamically as a Client of an RR. The simplest way to achieve this
is by manual configuration.
One of the key component of the route reflection approach in
addressing the scaling issue is that the RR summarizes routing
information and only reflects its best path.
Both Multi-Exit Discriminators (MEDs) and Interior Gateway Protocol
(IGP) metrics may impact the BGP route selection. Because MEDs are
not always comparable and the IGP metric may differ for each router,
with certain route reflection topologies the route reflection
approach may not yield the same route selection result as that of the
full IBGP mesh approach. A way to make route selection the same as
it would be with the full IBGP mesh approach is to make sure that
route reflectors are never forced to perform the BGP route selection
based on IGP metrics that are significantly different from the IGP
metrics of their clients, or based on incomparable MEDs. The former
can be achieved by configuring the intra-cluster IGP metrics to be
better than the inter-cluster IGP metrics, and maintaining full mesh
within the cluster. The latter can be achieved by
o setting the local preference of a route at the border router to
reflect the MED values, or
o making sure the AS-path lengths from different ASes are
different when the AS-path length is used as a route selection
criteria, or
o configuring community-based policies to influence the route
selection.
One could argue though that the latter requirement is overly
restrictive, and perhaps impractical in some cases. One could
further argue that as long as there are no routing loops, there are
no compelling reasons to force route selection with route reflectors
to be the same as it would be with the full IBGP mesh approach.
To prevent routing loops and maintain consistent routing view, it is
essential that the network topology be carefully considered in
designing a route reflection topology. In general, the route
reflection topology should be congruent with the network topology
when there exist multiple paths for a prefix. One commonly used
approach is the reflection based on Point of Presence (POP), in which
each POP maintains its own route reflectors serving clients in the
POP, and all route reflectors are fully meshed. In addition, clients
of the reflectors in each POP are often fully meshed for the purpose
of optimal intra-POP routing, and the intra-POP IGP metrics are
configured to be better than the inter-POP IGP metrics.
12. Security Considerations
This extension to BGP does not change the underlying security issues
inherent in the existing IBGP [1, 5].
13. Acknowledgements
The authors would like to thank Dennis Ferguson, John Scudder, Paul
Traina, and Tony Li for the many discussions resulting in this work.
This idea was developed from an earlier discussion between Tony Li
and Dimitri Haskin.
In addition, the authors would like to acknowledge valuable review
and suggestions from Yakov Rekhter on this document, and helpful
comments from Tony Li, Rohit Dube, John Scudder, and Bruce Cole.
14. References
14.1. Normative References
[1] Rekhter, Y., Li, T., and S. Hares, "A Border Gateway Protocol 4
(BGP-4)", RFC 4271, January 2006.
14.2. Informative References
[2] Savola, P., "Reclassification of RFC 1863 to Historic", RFC
4223, October 2005.
[3] Traina, P., McPherson, D., and J. Scudder, "Autonomous System
Confederations for BGP", RFC 3065, February 2001.
[4] Bates, T. and R. Chandra, "BGP Route Reflection An alternative
to full mesh IBGP", RFC 1966, June 1996.
[5] Heffernan, A., "Protection of BGP Sessions via the TCP MD5
Signature Option", RFC 2385, August 1998.
[6] Bates, T., Chandra, R., and E. Chen, "BGP Route Reflection - An
Alternative to Full Mesh IBGP", RFC 2796, April 2000.
[7] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.
Appendix A: Comparison with RFC 2796
The impact on route selection is added.
The pictorial description of the encoding of the CLUSTER_LIST
attribute is removed as the description is redundant to the BGP
specification, and the attribute length field is inadvertently
described as one octet.
Appendix B: Comparison with RFC 1966
All the changes listed in Appendix A, plus the following.
Several terminologies related to route reflection are clarified, and
the reference to EBGP routes/peers are removed.
The handling of a routing information loop (due to route reflection)
by a receiver is clarified and made more consistent.
The addition of a CLUSTER_ID to the CLUSTER_LIST has been changed
from "append" to "prepend" to reflect the deployed code.
The section on "Configuration and Deployment Considerations" has been
expanded to address several operational issues.
Authors' Addresses
Tony Bates
Cisco Systems, Inc.
170 West Tasman Drive
San Jose, CA 95134
EMail: tbates@cisco.com
Ravi Chandra
Sonoa Systems, Inc.
3255-7 Scott Blvd.
Santa Clara, CA 95054
EMail: rchandra@sonoasystems.com
Enke Chen
Cisco Systems, Inc.
170 West Tasman Drive
San Jose, CA 95134
EMail: enkechen@cisco.com
Full Copyright Statement
Copyright (C) The Internet Society (2006).
This document is subject to the rights, licenses and restrictions
contained in BCP 78, and except as set forth therein, the authors
retain all their rights.
This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.
Acknowledgement
Funding for the RFC Editor function is provided by the IETF
Administrative Support Activity (IASA).
AI ELI5
On this page
§1The Full-Mesh Problem
Normally all BGP speakers inside one autonomous system must be fully meshed, and any external routing information must be passed on to every other router in that AS. For n speakers that means n*(n-1)/2 unique IBGP sessions. The RFC says this does not scale when many speakers exchange a large volume of routing information.
Route reflection is offered as another alternative to the full mesh. It lets a BGP speaker, the route reflector, advertise IBGP-learned routes to certain IBGP peers. That changes the commonly understood concept of IBGP and adds two new optional non-transitive BGP attributes to prevent loops.
If every kid in a big class had to phone every other kid, the phones would never stop ringing. Route reflection picks a few kids to pass messages along instead.
§3Design Criteria
Route reflection was designed to meet three criteria, motivated by operational experience of a very large, topology-rich network with many external connections.
- Simplicity: it must be simple to configure and easy to understand.
- Easy transition: it must be possible to move from a full mesh without changing topology or AS. The RFC calls this an unfortunate management overhead of the confederation technique in its reference [3].
- Compatibility: peers that do not support it must be able to stay in the AS without losing any BGP routing information.
The new idea had to be easy to set up, easy to switch to, and friendly to old equipment that did not know about it.
§4Reflectors, Clients and Clusters
The basic idea is simple. In an AS with three IBGP speakers (RTR-A, RTR-B and RTR-C), IBGP speakers normally do not re-advertise IBGP-learned routes to other IBGP speakers. If RTR-C is allowed to do so, it can reflect the routes learned from RTR-A to RTR-B and the other way round, so the IBGP session between RTR-A and RTR-B is no longer needed.
A speaker that does this is a route reflector (RR) and the route is a reflected route. The internal peers of an RR fall into two groups: client peers and non-client peers. An RR together with its clients forms a cluster. The non-client peers must be fully meshed, but the clients need not be.
The reflector is the class messenger. Its clients just tell the messenger, and the messenger tells everyone else.
§6How Reflection Works
When an RR receives a route from an IBGP peer, it first selects the best path using its normal path selection rule. What it does next depends on who sent the best path.
- From a non-client peer: reflect it to all the clients.
- From a client peer: reflect it to all the non-client peers and also to the other clients. That is why clients do not need to be fully meshed.
An AS can have many RRs, and an RR treats other RRs like any other internal BGP speaker. In a simple setup the backbone is split into clusters, the RRs are fully meshed as non-client peers of each other, and each client keeps an IBGP session only with the RR in its cluster.
Conventional BGP speakers that do not understand route reflection can be clients or non-clients, which allows a gradual migration. You can start with one designated RR and configure the others as ordinary IBGP peers, then add clusters over time.
A cluster usually has a single RR and is identified by that RR's BGP Identifier. Because that is a single point of failure, several RRs in one cluster can be given the same 4-byte CLUSTER_ID so that an RR can discard routes from other RRs in its cluster.
The messenger repeats what it hears to the right groups. Old-style kids who do not know the game can still join in, and a cluster can have a backup messenger.
| Best path learned from | Reflected to |
|---|---|
| Non-client IBGP peer | All the clients |
| Client peer | All the non-client peers and also the other clients |
§8Avoiding Routing Information Loops
Through misconfiguration, reflected routes could form redistribution loops. Two new optional, non-transitive attributes detect them.
ORIGINATOR_ID (type code 9) is 4 bytes long and is created by an RR when it reflects a route. It carries the BGP Identifier of the route's originator in the local AS. A speaker SHOULD NOT create one if it already exists, and a router that recognizes it SHOULD ignore a route that arrives with its own BGP Identifier as the ORIGINATOR_ID.
CLUSTER_LIST (type code 10) is a sequence of CLUSTER_ID values showing the reflection path the route has passed. An RR MUST prepend its local CLUSTER_ID when it reflects a route, creating the list if it is empty. If the local CLUSTER_ID is already in the list, the advertisement SHOULD be ignored.
Each message gets stamped with who started it and which groups it has passed through. If a messenger sees its own stamp, it knows the message has gone in a circle and drops it.
- 9ORIGINATOR_ID — 4 bytes; BGP Identifier of the route's originator in the local AS. Optional, non-transitive.
- 10CLUSTER_LIST — Sequence of CLUSTER_IDs the route has passed through; each RR prepends its own. Optional, non-transitive.
§9Impact on Route Selection
The BGP tie-breaking rules from RFC 4271 are modified in two ways. If a route carries ORIGINATOR_ID, then in step f) it SHOULD be treated as the BGP Identifier of the speaker that advertised the route. A new rule SHOULD be inserted between steps f) and g): prefer the route with the shorter CLUSTER_LIST. A route without the attribute counts as length zero.
When reflecting, an RR SHOULD NOT modify NEXT_HOP, AS_PATH, LOCAL_PREF or MED, and the loop-prevention attributes must not be modifiable by configuration, because changes could cause routing loops.
When two routes look equally good, the one that went through fewer messengers wins. Messengers must pass notes along without changing them.
§11Configuration and Deployment
BGP has no way for a client to identify itself dynamically as a client of an RR, so the simplest approach is manual configuration. The RR summarizes routing information by reflecting only its best path, which is a key part of how it addresses scaling.
Because MEDs are not always comparable and IGP metrics can differ for each router, some reflection topologies may not choose the same routes as a full IBGP mesh would. The RFC suggests keeping intra-cluster IGP metrics better than inter-cluster ones and keeping a full mesh within the cluster, and handling MEDs by setting local preference to reflect them, making AS-path lengths differ, or using community-based policies. It also notes that one could argue this is overly restrictive, as long as there are no routing loops.
The reflection topology should be congruent with the network topology when multiple paths exist for a prefix. A commonly used approach is reflection by Point of Presence (POP): each POP has its own route reflectors serving its clients, and all the reflectors are fully meshed.
You have to tell each kid who their messenger is. Put the messengers where the real roads are, so messages take sensible routes.
§12Security Considerations
The RFC says only that this extension to BGP does not change the underlying security issues inherent in existing IBGP. It points to RFC 4271 and RFC 2385 for those issues and adds no new security mechanism.
Passing messages through a messenger does not make the messages any safer or less safe than before.