Reliability Analysis and Algorithm of Data Center Network

Folded Clos networks in data center networks architecture have received widespread attention in recent years, and the performance analysis in data center networks is a very important issue. In this paper, we analyze system reliability of folded Clos network, which is defined as the probability that at least one effective subgraph is operating. An effective subgraph is a subgraph of the network where all leaf nodes connect to at least one root node. Next, we define a basic subtree to be an effective subgraph having all nodes contributed to the reliability and show that the system reliability could be calculated by using basic subtrees of folded Clos network. In any data center network, the number of basic subtrees is less than 6% of the number of effective subgraphs, significantly reducing the running time of the accurate algorithm. Given that the calculation of the reliability is NP-hard, we propose a Monte Carlo algorithm to estimate it. The simulations show that the proposed system reliability describes the capability of the network to resist failures well and the proposed algorithms are effective and efficient.