Simple Virtual Aggregation (S-VA)
Reduce routing table size using virtual aggregation prefixes.
Raszuk, R., Heitz, J., Lo, A., Zhang, L., and X. Xu, "Simple Virtual Aggregation (S-VA)", RFC 6769, DOI 10.17487/RFC6769, October 2012, <https://www.rfc-editor.org/info/rfc6769>.
@misc{rfc6769,
series = {Request for Comments},
number = 6769,
howpublished = {RFC 6769},
publisher = {RFC Editor},
doi = {10.17487/RFC6769},
url = {https://www.rfc-editor.org/info/rfc6769},
author = {R. Raszuk and J. Heitz and A. Lo and L. Zhang and X. Xu},
title = {{Simple Virtual Aggregation (S-VA)}},
pagetotal = 8,
year = 2012,
month = oct,
} RFC text
Contents
Internet Engineering Task Force (IETF) R. Raszuk
Request for Comments: 6769 NTT MCL
Category: Informational J. Heitz
ISSN: 2070-1721 Ericsson
A. Lo
Arista
L. Zhang
UCLA
X. Xu
Huawei
October 2012
Simple Virtual Aggregation (S-VA)
Abstract
All BGP routers in the Default-Free Zone (DFZ) are required to carry
all routes in the Default-Free Routing Table (DFRT). This document
describes a technique, Simple Virtual Aggregation (S-VA), that allows
some BGP routers not to install all of those routes into the
Forwarding Information Base (FIB).
Some routers in an Autonomous System (AS) announce an aggregate (the
VA prefix) in addition to the routes they already announce. This
enables other routers not to install the routes covered by the VA
prefix into the FIB as long as those routes have the same next-hop as
the VA prefix.
The VA prefixes that are announced within an AS are not announced to
any other AS. The described functionality is of very low operational
complexity, as it proposes a confined BGP speaker solution without
any dependency on network-wide configuration or requirement for any
form of intra-domain tunneling.
Status of This Memo
This document is not an Internet Standards Track specification; it is
published for informational purposes.
This document is a product of the Internet Engineering Task Force
(IETF). It represents the consensus of the IETF community. It has
received public review and has been approved for publication by the
Internet Engineering Steering Group (IESG). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc6769.
Copyright Notice
Copyright (c) 2012 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction ....................................................3
1.1. Scope of This Document .....................................3
1.2. Requirements Notation ......................................3
1.3. Terminology ................................................3
2. Operation of S-VA ...............................................4
3. Deployment Considerations .......................................6
4. Security Considerations .........................................7
5. Acknowledgements ................................................7
6. Normative References ............................................7
7. Informative References ..........................................7
1. Introduction
This document describes a technique called Simple Virtual Aggregation
(S-VA). It allows some routers not to store some routes in the
Forwarding Information Base (FIB) while still advertising and
receiving the full Default-Free Routing Table (DFRT) in BGP.
A typical scenario is as follows. Core routers in the ISP maintain
the full DFRT in the FIB and Routing Information Base (RIB). Edge
routers maintain the full DFRT in the BGP Local RIB (Loc-RIB), but do
not install certain routes in the RIB and FIB. Edge routers may
install a default route to core routers, to Area Border Routers (ABR)
that are installed on the Point of Presence (POP), to core boundary
routers, or to Autonomous System Border Routers (ASBRs).
S-VA must be enabled on an edge router that needs to save its RIB and
FIB space. The core routers must announce a new prefix called
Virtual Aggregate (VA prefix).
1.1. Scope of This Document
The VA prefix is not intended to be announced from one AS into
another, only between routers of the same AS.
S-VA can be used for both IPv4 unicast and multicast address families
and IPv6 unicast and multicast address families.
S-VA does not need to operate on every router in an AS.
1.2. Requirements Notation
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 [RFC2119].
1.3. Terminology
RIB/FIB-Installing Router (FIR): A router that does not suppress any
routes and announces the VA prefix. Typically, a core router, a
POP to core boundary router, or an ASBR would be configured as an
FIR.
RIB/FIB-Suppressing Router (FSR): An S-VA router that installs the
VA prefix, but does not install routes that are covered by and
have the same next-hop as the VA prefix into its FIB. Typically,
an edge router would be configured as an FSR.
Suppress: Not to install a route that is covered by the VA prefix
into the global RIB or FIB.
Legacy Router: A router that does not run S-VA and has no knowledge
of S-VA.
Global Routing Information Base (RIB): All routing protocols in a
router install their selected routes into the RIB. The routes in
the RIB are used to resolve next-hops for other routes, to be
redistributed to other routing protocols, and to be installed into
the FIB.
Local/Protocol Routing Information Base (Loc-RIB): The Loc-RIB
contains the routes that have been selected by the local BGP
speaker's Decision Process as in [RFC4271].
NLRI: Network Layer Reachability Information [RFC4271]
2. Operation of S-VA
There are three types of routers in S-VA: FIB-Installing routers
(FIR), FIB-Suppressing routers (FSR), and, optionally, legacy
routers. While any router can be an FIR or an FSR, the simplest form
of deployment is for AS border routers to be configured as FIRs and
for customer facing edge routers to be configured as FSRs.
When a FIR announces a VA prefix, it sets the path attributes as
follows. The ORIGIN MUST be set to INCOMPLETE (value 2). The
NEXT_HOP MUST be set to the same value as that of the routes that are
intended to be covered by the VA prefix. The ATOMIC_AGGREGATE and
AGGREGATOR attributes SHOULD NOT be included. The FIR MUST attach a
NO_EXPORT community attribute [RFC1997]. The NLRI SHOULD be 0/0.
A FIR SHOULD NOT FIB-suppress any routes.
An FSR must detect the VA prefix or prefixes (including 0/0) and
install them in all of Loc-RIB, RIB, and FIB. The FSR MAY suppress
any more-specific routes that carry the same next-hop as the VA
prefix.
Generally, any more-specific route that carries the same next-hop as
the VA prefix is eligible for suppression. However, provided that
there is at least one less-specific prefix with a different next-hop
between the VA prefix and the suppressed prefixes, then those
suppressed prefixes must be reinstalled.
An example with three prefixes can be considered where the VA-prefix
(prefix 1) is the least specific and covers prefix 2 and prefix 3.
Prefix 2 is less specific than prefix 3 and covers the latter. If
all three have the same next-hop, then only the bigger one, i.e.,
VA-Prefix, is announced. However, if prefix 2 has a different
next-hop, then it will need to be announced separately. In this
case, it is important to also announce prefix 3 separately.
Similarly, when Internal BGP (IBGP) multipath is enabled, and when
multiple VA prefixes form a multipath, only those more-specific
prefixes of which the set of next-hops are identical to the set of
next-hops of the VA prefix multipath are subject to suppression.
The expected behavior is illustrated in Figure 1. This figure shows
an AS with a FIR, FIR1, and an FSR, FSR1. FSR1 is an ASBR and is
connected to two external ASBRs, EP1 and EP2.
+------------------------------------------+
| Autonomous System | +----+
| | |EP1 |
| /---+---| |
| To ----\ +----+ +----+ / | +----+
| Other \|FIR1|----------|FSR1|/ |
|Routers /| | | |\ |
| ----/ +----+ +----+ \ | +----+
| \---+---|EP2 |
| | | |
| | +----+
+------------------------------------------+
Figure 1
Suppose that FSR1 has been enabled to perform S-VA. Originally, it
receives all routes from FIR1 (doing next-hop-self) as well as from
EP1 and EP2. FIR1 now will advertise a VA prefix 0/0 with the next-
hop set to itself. This will cause FSR1 to suppress all routes with
the same next-hop as the VA prefix. However, FSR1 will not suppress
any routes received from EP1 and EP2, because their next-hops are
different from that of the VA prefix.
Several FIRs may announce different S-VA prefixes. For example, in a
POP, each edge router can announce into the POP an S-VA prefix that
covers the addresses of the customers it services.
Several FIRs may announce the same S-VA prefix. In this case, an FSR
must choose to install only one of them. For example, two redundant
ASBRs, both of which announce the complete DFRT, may each also
announce the default route as an S-VA prefix into the AS.
S-VA may be used to split traffic among redundant exit routers. For
example, suppose in Figure 1 that EP1 and EP2 are two redundant ASBRs
that announce the complete DFRT. Each may also announce two S-VA
prefixes into the AS: 0/1 and 128/1. EP1 might announce 0/1 with
higher preference and EP2 might announce 128/1 with higher
preference. FIR1 will now install into its FIB 0/1 pointing to EP1
and 128/1 pointing to EP2. If either EP1 or EP2 were to fail, then
FSR1 would switch the traffic to the other exit router with a single
FIB installation of one S-VA prefix.
3. Deployment Considerations
BGP routes may be used to resolve next-hops for static routes or
other BGP routes. Because the default route does not imply
reachability of any destination, a router can be configured to not
resolve next-hops using the default route. In this case, S-VA should
not suppress a route that may be used to resolve a next-hop for
another route from installation into the RIB. It may still suppress
it from installation into the FIB.
Selected BGP routes in the RIB may be redistributed to other
protocols. If they no longer exist in the RIB, they will not be
redistributed. This is especially important when the conditional
redistribution is taking place based on the length of the prefix,
community value, etc. In those cases where a redistribution policy
is in place, S-VA implementation should refrain from suppressing
installation into the RIB routes matching such policy. It may still
suppress them from installation into the FIB.
A router may originate a network route or an aggregate route into
BGP. Some addresses covered by such a route may not exist. If this
router were to receive a packet for an unreachable address within an
originated route, it must not send that packet to the VA prefix
route. There are several ways to achieve this. One way is to have
the FIR aggregate the routes instead of the FSR. Another way is to
install a black hole route for the nonexistent addresses on the
originating router. This issue is not specific to S-VA, but
applicable to the general use of default routes.
Like any aggregate, an S-VA prefix may include more address space
than the sum of the prefixes it covers. As such, the S-VA prefix may
provide a route for a packet for which no real destination exists.
An FSR will forward such a packet to the FIR.
If an S-VA prefix changes its next-hop or is removed, then many
routes may need to be downloaded into the FIB to achieve convergence.
4. Security Considerations
The authors are not aware of any new security considerations due to
S-VA. The local nature of the proposed optimization eliminates any
external exposure of the functionality. The presence of more
specifics that are used as VA prefixes is also a normal BGP behavior
in current networks.
5. Acknowledgements
The concept for Virtual Aggregation comes from Paul Francis. In this
document, the authors only simplified some aspects of its behavior to
allow simpler adoption by some operators.
The authors would like to thank Clarence Filsfils, Nick Hilliard, S.
Moonesamy, and Tom Petch for their review and valuable input.
6. Normative References
[RFC1997] Chandra, R., Traina, P., and T. Li, "BGP Communities
Attribute", RFC 1997, August 1996.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC5082] Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C.
Pignataro, "The Generalized TTL Security Mechanism
(GTSM)", RFC 5082, October 2007.
7. Informative References
[RFC4271] Rekhter, Y., Ed., Li, T., Ed., and S. Hares, Ed., "A
Border Gateway Protocol 4 (BGP-4)", RFC 4271, January
2006.
Authors' Addresses
Robert Raszuk
NTT MCL
101 S Ellsworth Avenue Suite 350
San Mateo, CA 94401
USA
EMail: robert@raszuk.net
Jakob Heitz
Ericsson
300 Holger Way
San Jose, CA 95134
USA
EMail: jakob.heitz@ericsson.com
Alton Lo
Arista Networks
5470 Great America Parkway
Santa Clara, CA 95054
USA
EMail: altonlo@aristanetworks.com
Lixia Zhang
UCLA
3713 Boelter Hall
Los Angeles, CA 90095
USA
EMail: lixia@cs.ucla.edu
Xiaohu Xu
Huawei Technologies
Huawei Building, No.3 Xinxi Rd.,
Shang-Di Information Industry Base, Hai-Dian District
Beijing 100085
P.R. China
Phone: +86 10 82836073
EMail: xuxh@huawei.com
AI ELI5
On this page
§1The Routing Table Size Problem
The global routing table has grown to millions of prefixes (IP address ranges). Routers in the Default-Free Zone must carry nearly all of them to deliver traffic to any destination. This exhausts memory on edge routers and increases lookup latency. Simple Virtual Aggregation (S-VA) is a technique that allows some routers to operate without storing all routes, while still advertising and receiving the full routing table via BGP.
S-VA is purely local to a single autonomous system. It requires no inter-domain negotiation, no tunneling, and no changes to routers that do not support it. It relies only on existing BGP functionality.
Modern routers are drowning in information about every network in the world. S-VA lets some routers get away with less detail.
§2How S-VA Works
S-VA defines two types of routers. FIB-Installing Routers (FIR) maintain the complete forwarding table and advertise virtual aggregate (VA) prefixes alongside normal routes. FIB-Suppressing Routers (FSR) receive all routes via BGP but suppress installation into the forwarding table of any more-specific routes that have the same next-hop as the VA prefix.
The VA prefix is typically the default route (0/0) or a large block, and the FIR sets its next-hop attribute to point toward itself. When an FSR receives both the VA prefix and a more-specific route pointing to the same next-hop, it installs only the VA prefix. Traffic destined for the more-specific routes follows the VA prefix to the FIR, which then forwards using the more-specific route.
One smart router says 'I know the whole map', and edge routers say 'I'll just remember that one route and ask you about details when I need them'.
§3Deployment Model
Typically, AS boundary routers (ASBRs) are configured as FIRs, and customer-facing edge routers are configured as FSRs. The FIR announces one or more VA prefixes with NO_EXPORT community to ensure they remain within the AS. An FSR may have a default route to the FIR or to a point-of-presence (POP) boundary router.
Multiple FIRs can announce different VA prefixes, or the same VA prefix. An FSR will suppress routes based on whichever VA prefix it installs. When internal BGP multipath is enabled, the FSR uses the set of next-hops from the VA prefix multipath to determine which more-specific routes to suppress. This enables traffic splitting among redundant exit routers.
The backbone routers advertise 'I have all the routes', and edge routers listen and discard their local copies of redundant information.
- VA prefix 0/0, next-hop self
- Specific routes (e.g., 203.0.113.0/24)
- More-specific routes covered by VA prefix
- Suppress specific routes into FIBRoutes with same next-hop as VA prefix are not installed
- Receive Route
- Same Next-Hop as VA?
- Covered by VA?
- Install in FIB
- Suppress from FIB
| From | To | Event |
|---|---|---|
| Receive Route | Same Next-Hop as VA? | check next-hop |
| Same Next-Hop as VA? | Covered by VA? | No or different |
| Same Next-Hop as VA? | Covered by VA? | Yes, same |
| Covered by VA? | Install in FIB | No |
| Covered by VA? | Suppress from FIB | Yes |
§4Operational Considerations
Operators must be aware of several edge cases. If a BGP route is used to resolve the next-hop of another route or is redistributed to other protocols, suppression into the forwarding table may break redistribution. The solution is to suppress from the FIB only, not from the RIB. When an FSR originates a route or aggregate that does not fully correspond to received routes, unreachable addresses within the aggregate may result in packets being sent to the FIR.
If a VA prefix changes next-hop or is withdrawn, many suppressed routes must be installed quickly to maintain convergence. The FIR SHOULD NOT suppress any routes, and should use ORIGIN INCOMPLETE to signal aggregation. No ATOMIC_AGGREGATE or AGGREGATOR attributes should be included on the VA prefix announcement.
The system works smoothly when everyone follows the rules about how to announce these aggregate routes and what happens when they change.
§5Security Considerations
S-VA has no known security vulnerabilities beyond those of normal BGP. The local nature of the optimization means it creates no external exposure. More-specific routes may be present as normal BGP behavior, and the suppression is transparent to external routers. An operator might use GTSM (Generalized TTL Security Mechanism) to protect IBGP sessions, as with any internal BGP deployment.
Because S-VA operates entirely within an AS and is invisible to other networks, it does not introduce new attack surfaces or require additional security infrastructure.
This optimization does not introduce any new ways for attackers to compromise routers beyond normal BGP attacks.