Dynamic Network Scheduling for Virtual Routers

Abstract

Network scheduling is important to satisfy the bandwidth requirements of virtual networks that consist of virtual machines in the end-hosts and the virtual routers connecting them. However, existing studies have focused on developing bandwidth allocation techniques for end-host virtual machines, but do not consider the network performance of virtual routers. In this article, we propose a new network scheduling framework for virtual routers-CreditBank. CreditBank dynamically allocates network resources to virtual routers according to bandwidth requirements, and it adapts to changing network environments without adding significant overhead. CreditBank offers three scheduling policies: minimum bandwidth reservation, weight-based proportional sharing, and hybrid scheduling. In addition, CreditBank supports an efficient work-conserving method to maximize network utilization. We implement CreditBank based on the Xen and Kernel-based Virtual Machine (KVM) hypervisors and evaluate its performance. The evaluation results indicate that CreditBank satisfies bandwidth requirements of the virtual routers while utilizing up to 99% of network resources.