To update network state or not?

Abstract

A Link-State Update Policy (LSUP) has the task to distribute information regarding the network resources, and is therefore considered to be an integral part of future Quality of Service (QoS) routing protocols. The argument is that in order to guarantee QoS to applications, one must know the available resources. Unfortunately, the high dynamics in available resources complicates the development of an LSUP, which on its turn will result in high deployment costs for Internet Service Providers. In this paper we will re-examine whether the gain in network performance, which is expected under the deployment of LSUPs, will outweigh its investment and complexity costs. To capture the complete range of possible LSUPs, we take a pragmatic approach and confine ourselves to examining two extreme strategies: routing with exact resource information and routing with no resource information. Our study comprises of an analytical exercise and extensive simulations on various network topologies under a large range of network loads. Our objective is to determine where static information provides acceptable network performance and where dynamic LSUPs are indispensable.