Latency-Aware Resource-Efficient Virtual Network Embedding in Software Defined Networking

Abstract

Software defined networking (SDN) is a key technology in building future generation networks by decoupling the control and data plane. By allowing multiple virtual networks (VN) to coexist in the same underlying SDN, virtual software defined networking (vSDN) can further improve the resource utilization and satisfy the diverse quality of service (QoS) requirements of users. Different from traditional VN embedding, there are more types of resources need to be allocated in vSDN. In addition, the control plane latency impacts on the performance of vSDN, especially when control flow and data flow share the same substrate links. Therefore, a key problem is minimizing the control latency while maintaining load balance. In this paper, we propose a heuristic latency-aware resource-efficient algorithm to solve it, after formulating the vSDN embedding problem as the integer linear programming model. In our method, we first aggregate the candidate physical nodes having similar resources proportion with the virtual node based on K-Means clustering. Then, the optimal physical node is selected out from the cluster according to the control latency derived from network calculus. Finally, the K-shortest path algorithm executes to map virtual links to the substrate paths between the corresponding substrate node. Simulation results demonstrate that the proposed algorithm improves the vSDN acceptance ratio by around 20% than an existing greedy algorithm with the controller-to-switch latency guarantee.