Assessment of Performance in Data Center Network Based on Maximum Flow

Recently, data center networks (DCN) have received significant attention from the academic and industry. However, researches of DCN are mainly concentrated on the improvement of network architectures and the design of routing protocols, or the performance evaluation from the perspective of node importance. In contrast to existing solutions, in this paper, we propose using maximum-flow theory to assess the network performance. Firstly, we abstract two kinds of typical DCN architectures and then formulate and convert the performance analysis of those architectures into a maximum-flow problem including a supersource and a supersink. Secondly, we get the value of maximum-flow by using Edmonds and Goldberg algorithm. Last but not the least, based on the theory of maximum-flow and Minimal cut sets, we get the critical edges for each architecture. Extended experiments and analysis show that our method is effective and indeed introduce low overhead on computation. In addition, the method and issues observed in this paper is generic and can be widely used in newly proposed DCN architectures.