g-EXTRACONNECTIVITY AND FAULT-TOLERANT UNICAST ALGORITHM OF RECURSIVE DATA CENTER NETWORKS

The research of data center network is a hotspot that has emerged in recent years. Compared with the traditional tree data center network, the recursive data center network has better fault tolerance and scalability. The k-dimensional recursive complete graph network deployed on n-port switch can be expressed as Xk,n. σ represents the number of edges connecting any vertex in the graph with other subgraphs of the same dimension. In practical situations, the extra connectivity of the network can measure the network’s fault tolerance more accurately than traditional connectivity. This paper proves that when k≥3, n≥3, σ∈1,n-1 and 0≤g≤2, the extra connectivity of Xk,n is (g+1)(kσ-1)+n, which is close to g+1 times of its connectivity. Furthermore, we propose a fault-tolerant unicast algorithm based on this case. In this paper, it is proved that the time complexity of the algorithm and the maximal length of the path constructed by the algorithm between any two nodes in the worst case. Through simulation experiments, it is proved that the algorithm is superior to the breadth-first search and depth-first search algorithms in execution efficiency, and the algorithm has good fault tolerance performance.