Information Technology Reference
In-Depth Information
To further increase RR scalability, you can create hierarchies. Between different levels of
RRs, you can use standard BGP communities to designate routes to be passed between dif-
ferent partitions. For example, PEs can attach a designated additional standard community
to a subset of the VPNv4 routes that need to be made available in the other RR groups. The
top-level RRs can then be configured to accept only the routes with the designated standard
community, so only routes matching that community can be passed between RR groups.
Selective filtering between a PE device and an RR can also be accomplished on the PE
side. An RT export map (discussed later, in the section “Route Target Design Examples”)
configured under a VRF can selectively export RTs that an RR will accept. A standard
community filter can also be used on a PE device. When compared to outbound filtering on
PE devices, inbound filtering on RRs generally requires less maintenance but increases
CPU usage on RRs.
In a network that carries both IPv4 and VPNv4 routes, RRs for both types of prefixes can
be used to increase scalability. When large numbers of Internet routes and VPN routes are
carried, it is desirable to have dedicated RRs for each address family. On a VPNv4 RR,
disable the default IPv4 prefix processing for all sessions.
Tuning an RR Router
Because an RR handles large numbers of routes, it is important that the router is at peak
performance for receiving and processing updates. You can use the following two
approaches:
•
Increase input hold-queue size for all the interfaces
•
Enable TCP path-MTU discovery.
Consult Chapter 3, “Tuning BGP Performance,” for details.
Design Guidelines for RDs
One RD is configured on a PE router for each VRF. A common RD format is local AS
number:customer ID. However, an RD might or might not be related to a site or a customer
VPN. The assignment of RDs affects how VPNv4 routes are installed. This section
discusses the implications for memory use.
Generally, there are three different ways to assign RDs in PE devices across the network:
•
The same RD for the same VPN
•
A unique RD for each VRF
•
A unique RD for each VRF per site
