A group leader election scheme to enhance the session throughput in tree-based reliable multicast

Abstract

In this thesis, we present a group leader election scheme to enhance session throughput in tree-based reliable multicast. Tree-based reliable multicast protocols organize senders and receivers into one or more groups. Each group has a group leader that is responsible for loss recovery of its group members. Existing work elects a node that places the closest to a sender and has low loss rate to promote local recovery. However, in tree-based reliable multicast protocols, the performance of a session is mostly affected by the node that has the lowest throughput, called "weakest link." Thus, a node that minimizes the throughput degradation caused by the weakest link in a group should be elected as a group leader. For this, we present a metric to estimate throughput degradation caused by the weakest link. The proposed scheme elects a node that enhances session throughput by estimating expected session throughput degradation caused by weakest link. Through simulation, we showed that the proposed scheme outperforms the existing work in terms of session throughput.