Multicast scheduling algorithm in software defined fat-tree data center networks

Abstract

Multicast can not only shorten task completion time of applications, but also effectively reduce overall bandwidth consumption in data center networks (DCNs). However, load imbalance and sudden link blocking will seriously impact the network performance owing to the fundamental characteristics of huge traffic in DCNs. To address this problem, in this paper, we propose a novel multicast scheduling algorithm in software defined fat-tree DCNs, which can improve network performance by reducing the blocking of multicast traffic. In particular, the multicast scheduling algorithm selects the minimum blocking cost of links as optimal paths. By our scheduling algorithm, multicast flows are evenly distributed over the available links so as to achieve load balance. In addition, the DCNs are controlled globally by the software defined networking (SDN) technology, therefore, the multicast traffic cannot be interfered by the unexpected flow requests. Furthermore, the multicast scheduling algorithm can lead to asymptotically minimum network blocking probability, and has a low time complexity. Simulation results verify the effectiveness of our proposed multicast scheduling algorithm in different network traffic intensities.