A Framework for Optimized Bandwidth Allocation to LSPs in an MPLS Network

Abstract

Bandwidth allocation is a vital issue in the emerging MPLS technology in the area of computer networking. There is need to ensure an ecient and congestion free trac through suitable bandwidth allocation. Though some algorithms exist to address this issue, it is felt that more optimized algorithms can be benecial. Max-Min congestion control algorithm" [1] by Marty and Ali, proposed a basis of congestion control. The Proportional fair rate allocation algorithm"[2] and Bandwidth Allocation Algorithm"[3] addressed the issue of congestion control in MPLS networks. The above approaches used the concept of predened weights to the LSPs which means that bandwidth is allocated according to some presumptions. This may lead to some amount of unused bandwidth and a situation may arise where bandwidth is allocated to an LSP which doesnt utilize it fully but there exists another LSP which falls short of its current bandwidth requirement. To account for the changing bandwidth needs and also the current datarate of the LSPs, this paper proposes a framework for fair bandwidth allocation to the LSPs in a more optimized manner. In addition to the algorithm, we include a simulation of a static bandwidth allocation approach using RSVP-TE with MPLS in OMNET++ IDE integrated with INET framework. We compare the parameter of queue length for all interfaces of all LSRs in the network and for a particular interface at dierent datarate values. We show from our observations that with increasing datarate, the average queue length gradually decreases.