A study of feedback based congestion control algorithm in MPLS network

Abstract

Considerable interest has arisen in congestion control from the knowledge that although sensible provisioning of network infrastructure is needed, together with sufficient underlying capacity, these are not sufficient to deliver the Quality of Service required. This is due to the dynamic variations in load. In operational Internet Protocol (IP) networks, it has been difficult to incorporate effective congestion control due to the limited capability of IP technology. In principle, Multiple Label Switch (MPLS), a connection-oriented label swapping technology, offers new possibilities in addressing this limitation by allowing the operator to use a sophisticated traffic control mechanism. Due to the reasons mentioned there is a strong requirement to improve network performance and efficiency by congestion control for MPLS network. This thesis proposes a scheme to dynamically manage traffic flows through the network by efficient rate allocation. It proposes two kinds of feedback based congestion control algorithms that will allow label switch router(LSR) to utilize explicit routing: Binary feedback congestion control algorithm and explicit rate congestion control algorithm. In the binary feedback congestion control, ingress LER can increase or decrease transmission rates using single bit information which indicates congestion state of network. The explicit rate feedback control, on the other hand, attempt to directly compute transmission rate based on available bandwidth and LSP state information. Since most feedback based congestion control suffers from the large time constant associated with the adaptive process, this thesis investigate the effect of large propagation delay using the Quasi-Birth-Death process. Numerical result show that both algorithm make it possible to improve network performance in terms of loss rate and throughput. This means that our feedback based congestion control algorithm is applicable to MPLS network.