Information Technology Reference
In-Depth Information
consider the following points:
•
Minimizing resource consumption on PE devices, including proper RD design and RT
policy to minimize PE memory use
•
Separating VPNv4 and IPv4 routing information
•
Using route reflection properly
•
Filtering at inter-AS borders properly
•
Using label maintenance methods
Many of these points are discussed throughout this chapter. This section concentrates on the
following topics:
•
Resource consumption on PE devices
•
Route reflection with MPLS VPN
•
Design guidelines for RDs
Resource Consumption on PE Devices
This discussion on resource consumption on PE devices focuses on CPU usage and
memory consumption required to store various structures. This section describes factors
that can affect resource consumption and some guidelines for consideration during actual
evaluation. Follow current Cisco documentation on exact usage numbers, which can change
from release to release.
The use of CPU resources is dependent on a variety of factors, including the following:
•
Number of backbone (toward P routers) BGP peers
—More peers lead to more
processing. The use of peer groups reduces the per-peer processing overhead.
•
Number of provisioned VRFs
—A higher number of VRFs configured locally
requires more maintenance.
•
Number of VPN routes
—More VPN routes require more processing.
•
PE-CE connectivity type
—Different protocols result in different processing
overhead. For example, eBGP might require less processing than OSPF.
•
The type of CPU
—Higher-powered CPUs obviously have better performance.
•
Hardware platforms
—Hardware platforms might require different maintenance
tasks to be performed.
Several structures can consume a significant amount of memory in a PE:
•
A global IP RIB to hold the provider internal networks and Internet routes
•
A VPN BGP table to hold VPNv4 structures
•
CEF and LFIB for the global routes and VPN routes
•
A VRF IP RIB to hold per-VRF routing information
