Live Migration Planning of Virtual Machines in Hybrid SDN

Abstract

Software Defined Networking (SDN) is an emerging network architecture which intends to make networks into the age of the cloud. Nevertheless, the deployment of SDN is a big challenge in both technical and financial aspect. Hybrid Software Defined Network (HSDN) is a new method to slove the problems in the transition period and is even considered as the inexpensive substitutes of SDN. Virtual machine (VM) live migration is widely used for system management in data centers and SDN, which can facilitate resource optimization, load balancing, fault management and energy saving etc.. There have been continuous efforts in reducing the migration time. In fact, the similar VM management can also be realized in HSDN. In this paper, we explore the live migration planning of virtual machines in an inter-datacenter network, where SDN node is deployed incrementally. We compute the optimal migration task sequence and network bandwidth used for each migration task. This is the first work about Virtual Machine management in HSDN. We then formulate this problem as a mixed integer programming (MIP), which is NP-hard. And then we propose a fully polynomial time approximation (FPTA) algorithm to slove it. Extensive experiments show that our algorithm makes full use of the features of HSDN, and achieve the goals of live VM migration planning in terms of the total migration time and downtime.