A scheduling algorithm towards bandwidth guarantee for virtual cluster in the cloud

Due to the sharing of network resources in multi-tenant cloud data centers, minimum bandwidth guarantee has become one of the important methods to provide predictable performance for cloud applications. Efficient virtual network allocation helps accommodate a larger number of virtual clusters and improve the resource utilization in the cloud. Towards the demand for network bandwidth guarantee, this paper proposes a novel backtracking algorithm for scheduling virtual clusters in the cloud. For the typical tree network topology of a data center, this algorithm firstly judges whether there exists a valid solution inside each sub-tree in the network topology. After determining the sub-tree for the virtual cluster, it recursively searches for the detailed solution inside the sub-tree based on a backtracking algorithm, thus avoiding the problems of fake allocation or rejecting the request by mistake existing in related works. The experimental results show that the exact search based on backtracking can help increase the acceptance ratio of virtual cluster requests. Compared with existing algorithms, it can reduce the rejection ratio by 10% on average, which can contribute to improving the resource utilization in the cloud.