摘要

Application layer multicast (ALM) provides a low-cost solution for multicast over the Internet. It overcomes the deployment hurdle of IP multicast by moving all multicast related functions from network routers to end-hosts. However, since packet replication is performed on end-hosts, the system performance of an ALM is limited by the bandwidth of end-hosts. Therefore, degree-constrained QoS-aware multicast routing becomes one of the key concerns for implementing realtime multicast services, such as continuous streaming applications. In this paper, we claim that the QoS gained by most users will be better evaluated using the overall latency, and we explore the optimization of Degree-Constrained Minimum Overall Latency Spanning Tree (DCMOLST). The process for optimizing the overall latency is divided into two phases, i.e., the initialization phase and the dynamic adjustment phase. In the former phase, we present a heuristic DCMOLST algorithm which negotiates both transmission delay and node bandwidth simultaneously, so as to avoid QoS degradation caused by any single metrics. In the later phase, we define a set of distributed iterative optimizing operations to swap the position between nearby end-hosts for further optimization. Experimental results show that the proposed degree-constrained QoS-aware routing algorithm could improve the overall performance of application layer multicast services.