Dynamic Multi-tree Switching for Multimedia Multicast in an OpenFlow-Based Fat-Tree Network

Abstract

In data center networks, multicast paradigm is significant for multimedia data transmission, e.g. distributed video transcoding [13]. However, single multicast tree generally cannot cope with network congestion and failures well. Although multi-tree multicast is able to achieve load balance and failure recovery by applying dynamic multi-tree switching mechanisms, it is hard to be deployed and implemented in the conventional networks. OpenFlow separates the control plane and data plane, and thus has better controllability and programmability. Therefore, in this paper, we aim to propose an adaptive filter based dynamic multi-tree switching algorithm for multimedia multicast in an OpenFlow-based Fat-Tree data center network. Firstly, for each multicast session, we select several multicast trees as candidates and calculate a priority value for each multicast tree by using the monitored network status. Secondly, adaptive filter is used to predict the priority value of the next time slot which is then used to design the dynamic hierarchical multi-tree switching strategy and the data distribution policy. Finally, experiments are carried out in Mininet to verify the feasibility and performance of our proposed algorithm.