Clustering and Constraints for Real-time Multicast

Real-time multicast is widely applied on Internet. Unfortunately, current systems share the same problem that they are not able to maintain high and stable performance when the scale grows unexpectedly large or the nodes´ performance fluctuation drastically. This paper attempts to solve the problem by means of network structure clustering and constraints. We show that the multicast network can achieve much higher optimizing speed through dynamically clustering nodes with similar performance records. We propose a cluster-based multicast model and constraints on it. We further provide Regular-Balanced constraint as the recommended constraint, which is related to higher ability of scale and failure tolerance. We also test our model by contrast experiments with CoolStreaming and NICE.