Network-Aware Re-Scheduling: Towards Improving Network Performance of Virtual Machines in a Data Center

摘要

An effective virtual machine allocation and scheduling algorithm can improve the utilization of physical servers, lower energy cost and improve the overall performance of datacenters. Current virtual machine scheduling algorithms mainly focus on the initial allocation of VMs based on the CPU, memory and network bandwidth requirements. However, Caused by finish of jobs or expiration of lease, related virtual machines would be shut down and leave the system which generate plenty of resource fragments. Such fragments lead to unbalanced resource utilization and the communication performance may decline significantly. This paper studied the network influence on some typical applications in datacenters and proposed a self-adaptive network-aware virtual machine re-scheduling algorithm to maintain an optimal system-wide status. Our algorithm had two stages. In the first stage, we checked whether re-scheduling was necessary and in the second stage perform a heuristic re-scheduling to lower communication cost among VMs. We use two benchmarks in a real environment to examine network influence on different tasks. To evaluate the advantages of the proposed algorithm, we also build a cloud computing testbed. Real workload trace-driven simulations and testbed-based experiments show that, our algorithm greatly shortens the average finish time of map-reduce tasks and reduced time delay of web applications. Simulation results showed that our algorithm considerably reduced the amount of high-delay jobs, lowered the average traffic passed through high-level switches and improved the communication ability among virtual machines.