MPLS/VPN: Scaling the Solution

Many factors must be considered when planning a scalable network design. Many of these are not specific to the MPLS/VPN architecture and are therefore not covered in this book. However, certain areas within the network infrastructure are directly affected by the introduction of the MPLS/VPN architecture and must be addressed before implementation.

To understand the factors that may affect the scalability of the design, we must first consider how the MPLS/VPN service is achieved and what mechanisms are used to provision the service. This has been extensively covered in previous chapters, but in summary, we have seen that customer VPN routes are advertised across the backbone between PE-routers, and this information is then disseminated to other VPN sites. Whenever routes are advertised across a backbone network, BGP is the protocol of choice and indeed is a fundamental building block of the MPLS/VPN architecture. Therefore, we need to look at the scaling properties of this protocol, not only in terms of configuration, but also in terms of overhead for the routers that provide the BGP session endpoints and the devices that carry these sessions across the backbone.

We also need to understand how quickly routing information can be disseminated and synchronized between VPN sites in the event of a topology change in the network (for example, a link going down or coming back up)?this process is known as convergence. The convergence speed of the MPLS/VPN solution is of paramount importance to the customers of the VPN service who deploy their mission-critical and time-sensitive solutions across this infrastructure, so a thorough understanding of the factors that affect this convergence is imperative.

Finally, we have seen that many different sites can attach to the backbone through the same PE-device. This means that we need to investigate what issues this presents and how the PE-device can cope with deployment of additional sites.