Addressing OSPF Load in IP/MPLS Networks

There is an increasing trend towards deploying IP/MPLS networks that provide value added services such as virtual private networks, guaranteed bandwidth pipes, etc. to network users. Such networks are required to be highly reliable and robust even under peak load and/or unanticipated situations of link/router failures. TE mechanisms typically involve path computation based on information obtained through OSPF or other link state routing mechanisms and then using a signaling mechanism such as RSVP-TE to set up the computed path in the network. As routers in such a network have to endure a higher computation and protocol processing load compared to routers in a best effort IP network, these mechanisms will have to be significantly scalable in robust large-scale TE networks. Path computation mechanisms have been studied extensively and it is generally accepted that such mechanisms can be implemented within the means of current router architectures. The link state routing overhead however, can still be significant even when using intelligent path computation mechanisms. Proposed mechanisms to decrease this overhead require changes to the protocols and are therefore not easily deployable given the large installed base of protocol implementations. In this paper, we show that OSPF protocol overhead is indeed a significant component of the load offered on routers and that this load can become unacceptable as the degree of the node increases. We then outline an implementation mechanism that takes advantage of the evolving router architectures which helps scale OSPF implementations significantly.