Building a multicasting tree in a high-speed network

Abstract

To build a multicasting tree in a multicomputer network, the authors propose three strategies based on voting constructing a minimum spanning tree, and repeatedly constructing multiple minimum spanning trees. Typically, performance metrics to evaluate a multicast solution include time and traffic. Simultaneously optimizing both metrics is computationally intractable because the problem is NP-complete. The first scheme always guarantees the use of the shortest path from the source node to each destination. which makes it time-optimal. The other two schemes do not guarantee this but try to reduce the traffic as much as possible. To demonstrate these strategies' effectiveness, the authors apply them to hypercubes, star graphs, and star graphs with some faults. The report shows experimental results to evaluate the performance of these solutions.