Selecting Among Replicated Adaptive Multicast Servers
Ammar, Mostafa H. (Mostafa Hamed)
Zegura, Ellen W.
MetadataShow full item record
Server replication and multicasting are well-established techniques for increasing capacity of a networked service and improving client performance. In this paper, we consider the combination of these two techniques. Specifically, we investigate the problem of selecting amongst rate-adaptive multicast servers, which adjust their sending rate based on network conditions and/or feedback from clients. Effective server rate adaptation can lead to efficient utilization of network resources and performance improvement perceived by clients. In this initial study of adaptive multicast server selection, we explore some fundamental issues and study the implications of different selection strategies on the performance perceived by clients. We first define the Static Multicast Selection Problem, in which there are static sets of clients and servers, and one needs to establish a set of multicast trees with one tree for each server. We explore several optimization problems based on different performance measures. We prove that the general problem is NP-hard and then present two interesting special cases with an optimal polynomial-time solution in each case. We design a heuristic for the general case and show that it can improve the performance over some simple strategies. We also consider the Dynamic Multicast Selection Problem, in which clients may join and leave multicast trees already established. We design a heuristic for this dynamic case by which clients can select a tree to join. We investigate the performance of the heuristic through simulation.