Multi-server optimal bandwidth monitoring for collaborative distributed retrieval

Accurate bandwidth monitoring is critical for multimedia delivery applications that guarantee QoS parameters like time limit for transmission. In our past work we monitored the bandwidth at several servers independently, then chose the most reliable links to transmit scalable multimedia information. The assumption behind our past work was that each server retrieves a fixed part of a multimedia object, and this part is determined after the bandwidth monitoring process and cannot be changed during the retrieval process. In this paper, we consider a retrieval strategy where the different servers collaborate in the retrieval process; i.e., if a server finishes its share of the retrieval task it helps with the retrieval task of other servers. Given this "collaborative" retrieval scenario, a new distributed monitoring algorithm is proposed. Simulation results show that the proposed algorithm is more effective than our past work in distributed retrieval.