Internet Engineering Task Force Jieyun (Jessica) Yu
INTERNET DRAFT UUNET
Expires May, 2000 November, 1999
IPv6 Multihoming with Route Aggregation
<draft-ietf-ipngwg-ipv6multihome-with-aggr-00.txt>
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that other
groups may also distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (1999). All Rights Reserved.
Abstract
With the growing requirements for reliable Internet connectivity,
increasing number of enterprises choose to acquire Internet
connectivity from more than one Internet Service Providers (ISPs) for
the purpose of connectivity redundancy and traffic load distribution.
The potential of large number of multi-connection sites impose direct
challenge on routing aggregation and consequently on scalability of
the global Internet routing. Hence a solution is highly desirable
which provides the benefit of multi-connection as well as the better
scalability of the global routing system. In addition, such solution
needs to be simple enough to be operationally manageable. With the
fast growth of ISP networks as well as enterprise networks, network
manageability is becoming an increasingly important requirement. This
Yu IPv6 Multihoming with Route Aggregation [Page 1]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
document describes a solution which achieves the stated goals.
1. Motivations
With the growing requirements for reliable Internet connectivity,
increasing number of enterprises choose to acquire Internet
connectivity from more than one Internet Service Providers (ISPs) for
the purpose of connectivity redundancy and traffic load distribution.
This type of Internet connection arrangement is referred to as
multihoming and such enterprise network is referred to as multi-homed
site.
Large number of Multi-homed sites impose direct challenge on routing
aggregation and consequently on global Internet routing scalability.
As is well known, the keys to scaling of huge global Internet routing
system including routing information abstraction and aggregation. The
IPv6 unicast address format described in [1] enables the strategy of
allocating a large block of IPv6 address space to a service provider
and letting the service provider further assigning sub-blocks of the
IP addresses to its customers. This provider based IP address
assignment strategy makes it possible for route aggregation at the
provider level and thus facilitates the scaling of the global
Internet routing system.
However, the current common mechanism to route a multi-homed site is
to make the route specifically associated with the site visible in
the global Internet routing system. This practice prevents route
aggregation at the provider level and imposes challenge to a scalable
global routing. Therefore, a solution is needed for IPv6 multihoming
which provides the desired benefits of a multihoming connection such
as redundancy and load sharing, and at the same time, enables better
scaling of the global global routing system. In addition, such a
solution needs to be simple enough to be operational manageable. With
the fast growth of ISP networks as well as enterprise networks,
network manageability is becoming an increasingly important
requirement. In today's Internet, manageability of a solution should
be one of the top considerations.
This document describes a scheme that supports IPv6 multihoming and
also achieves the followings:
a. Providing redundancy and load sharing for the multi-homed sites
b. Facilitating the scalability of the global IPv6 Internet
routing table
c. Simple and operationally manageable
This mechanism is a routing approach for multihoming. It uses
existing routing protocol and implementation thus no new protocol or
Yu IPv6 Multihoming with Route Aggregation [Page 2]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
changes are needed.
The mechanism described in this document can also be applied to IPv4
Internet.
2. The Multihoming Mechanism
Multihoming connections in general can be categorized into two
different types: a) a site multi-homed to a single ISP, commonly at
different geography locations and b) a site multi-homed to more than
one ISPs.
In scenario a), the specific routes associated with the multihomed
site will not be visible outside of the particular ISP network and
thus there is no real impact on the global routing. Therefore, no
special mechanism is needed for multihoming in this scenario. The
mechanism described in this document addresses the situation of a
multi-homed site connects to more than one ISPs.
The mechanism is described with an example of a multi-homed site with
two ISPs connections since two-connection multihoming represents the
majority of the multihoming cases and it simplifies the discussion.
The mechanism, however, can be extended to apply to multi-homed sites
with more than two ISP connections.
2.1. Address Assignment
To obtain IP addresses, a multi-homed site will designate one of its
ISPs as its primary ISP and receive IP address assignment from the
primary ISP's IPv6 aggregation block.
Figure-1 illustrates an example of a multi-homed site (Customer-A)
with connectivity to ISP-1 and ISP-2. ISP-1 is chosen as the primary
ISP for customer-A and assigns Addr-1-A from its address block
(Addr-1) to the customer.
ISP-3 ---- ISP-4
| / |
| / |
| / |
ISP-1 ---- ISP-2
\ /
link-1 \ / link-2
Customer-A
Yu IPv6 Multihoming with Route Aggregation [Page 3]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
Figure 1: Example of Multihomed Site
2.2. Routing
In order for Internet traffic destinated to Customer-A to reach the
targeted destinations, Customer-A will advertise addr-1-A to ISP-1
and ISP-2 respectively. ISP-2 will advertise Addr-1-A to ISP-1 and to
ISP-1 only. ISP-1 will, of course, advertise its own aggregation
Addr-1 to the entire Internet.
As a result of this routing advertisement, inbound traffic destinated
to Customer-A and originating within ISP-1 or ISP-2 will be forwarded
to Customer-A using link-1 or link-2 respectively. Traffic originated
from anywhere else in the Internet will first be forwarded to ISP-1
since it advertises the route to Addr-1 which contains Addr-1-A.
ISP-1 will then forward the traffic destined to Customer-A via its
connection(s) to ISP-2 and/or via its direct link to customer-A,
according to the preset routing polices. The commonly used policy is
to use the shortest exit by utilizing IGP metric as a tier break in
BGP route selection process. By using both connections to forward
traffic to Customer-A, load sharing among the multiple links used by
Customer-A for connecting to the Internet is achieved.
For outbound traffic originated from Customer-A, ISP-1 and ISP-2
would announce default route and/or a selected set of specific
prefixes to Customer-A based on the requirements of Customer-A. As a
result of the advertisement, traffic originated from Customer-A to
the Internet will be forwarded accordingly and load sharing can be
accomplished.
In the aspect of redundancy, when the link between customer-A and
ISP-2 (link-2 in Figure 1) fails, all traffic will go in and out via
the connection between Customer-A and ISP-1 (i.e. via link-1 as shown
in Figure 1). Likewise, when link-1 is experiencing an outage, link-2
will be used for transmitting the traffic. This is because ISP-1 will
continue announcing its aggregate block Addr-1 to the entire Internet
and ISP-2 will still advertise Addr-1-A to ISP-1. All of the inbound
traffic to the customer will utilize link-2 by taking the path of
ISP-1 -> ISP-2 -> Customer-A. Outbound traffic from Customer-A will
automatically fall to link-2. This way, when one of the two links
fails, the other will be used for traffic in and out from the multi-
homed site. Redundancy is thus accomplished.
As one would observe, with this mechanism, the specific route
associated with the multi-homed site is only visible to ISP-1 and
ISP-2 in the example. Only the multi-homed sites directly connected
ISPs, not the rest of the Internet will have to obtain the specific
Yu IPv6 Multihoming with Route Aggregation [Page 4]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
route(s) associated with a multi-homed site. This results in better
scaling of the global Internet routing.
The same mechanism can be extended to sites multihoming to more than
two ISPs. Again, only those ISPs that the customer directly connected
to would carry the more specific prefix assigned to the multi-homed
customer.
3. Discussions
Characteristics of a multihoming mechanism described in this document
include:
- Improved scaling of the global routing system without loosing the
benefits expected from multihoming.
- Does not require new protocol or routing software changes to deploy
the scheme.
- Simple and thus manageable. In addition, operationally, it has less
chances of generating errors compared to more complicated solutions.
- Due to its simpleness, the requirement of having sophisticated
network administrator onsite is greatly reduced, which can be an
attractive choice for a variety of multi-homed sites.
- The primary ISP of a multi-homed site such as ISP-1 in Figure-1
would need to do more work in terms of distributing traffic among the
other ISP the multi-homed site directly connect to and its direct
link to the site. It will also carry more traffic for the multi-homed
customer. However, this can be considered as a value added service
from the ISP to the customer and the primary ISP could charge for
such services accordingly.
- If the two involved ISPs has no direct connection, the more
specific route associated with the multi-homed site would need to be
carried by other ISP(s) in the path thus it would result in less
effective aggregation. However, it seems to be lesser of a problem
since ISP assigned with an Internet visible aggregate block or Top
Level Aggregator (TLA) usually are top tiered ISPs and all such ISPs
generally have direct connections to each other either via private
peering or public peering points.
- The primary ISP is the sole interface for the multi-homed customer
to the Internet with the exception of the ISPs the customer has
direct connection with. Outages such as one between ISP-1 and ISP-4
in Figure-1 would impact the reachability from customers of ISP-4 to
Customer-A even though ISP-2 still has good connection to ISP-4.
Yu IPv6 Multihoming with Route Aggregation [Page 5]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
However, if the primary ISP is a good quality ISP, this sort of
situation should rarely happen. It's common practice that an ISP,
especially a good quality one, to have multiple connections to other
big ISPs at different geographical locations. Good quality ISPs also
have robust internal network design to prevent any failure from
impacting the entire network. Choosing a good quality ISP as primary
ISP is a good practice for multi-homed sites adopting this solution.
4. Conclusions
It wouldn't be hard to understand that even those enterprises desire
multiple Internet connections may have different criterias and
resource constrains for implementing such connection. Some may
require absolute redundancy while most may only desire reasonable
redundancy. This document offers a viable multihoming solution for an
enterprise to choose based on its particular requirements and
constrains. The multihoming mechanism described in this document is
applicable to various multihoming scenarios, the most suitable
environment for deploying it are:
a. ISPs serving the multi-homed site have direct connection(s) to
each other. Although such direct connection is not required, it
would make arrangement simpler and will also improve aggregation
by limiting specific routes visible only to ISPs serving the
multi-homed site.
b. Enterprises with requirements for good redundancy but not
absolute redundancy.
c. Enterprises with limited to resource for onsite sophisticated
network administrators
d. Enterprises able to choose a robust ISP as primary provider.
Although not the main focus, the mechanism described in this
document can also be used to improve routing scalability within
networks shares one aggregation block or Top Level Aggregator
(TLA).
5. Security Considerations
BGP security applies to the work presented. No added security risk is
known.
6. Acknowledgements
Many thanks to Guy Davis, Robert J. Rockell and Akira Kato for their
insightful comments.
Yu IPv6 Multihoming with Route Aggregation [Page 6]
Internet Draft ipngwg-ipv6multihome-with-aggr-00.txt November 1999
7. References
[1] R. Hinden, M. O'Dell and S. Deering, "An IPv6 Aggregatable Global
Unicast Address Format." RFC2374, July 1998. ftp://ftp.isi.edu/in-
notes/rfc2374.txt
[RFC2546] A. Durand and B. Buclin, "6Bone Routing Practice."
RFC2546, March 1999.ftp://ftp.isi.edu/in-notes/rfc2546.txt.
Author's Address
Jieyun (Jessica) Yu
UUNET Technology
880 Technology Dr.
Ann Arbor, MI 48108
Phone: (734) 214-7314
EMail: jyy@uu.net
Yu IPv6 Multihoming with Route Aggregation [Page 7]
