Multi-resource Allocation for Virtual Machine Placement in Video Surveillance Cloud

Abstract

Video surveillance cloud is an emerging cloud computing paradigm which can provide the elastic resource management ability for surveillance video processing tasks. The video processing tasks usually require extensive computing resources, and different tasks have different resource configuration requirements. It is challenging to find the optimal fine-grained resource configuration for various video processing tasks. In this paper, we study how to map the heterogeneous virtual machine requests to the heterogeneous physical machines. First, we design a video surveillance cloud platform architecture. The cloud platform can be seamlessly integrated with the video surveillance systems that comply with the ITU standard. Second, we propose a multi-resource virtual machine allocation algorithm named Dominant Resource First Allocation DRFA. Our aim is to maximize the resource utilization in heterogeneous cloud computing environment. By computing the dominant resource under multiple resource dimensions, our proposed algorithm DRFA can make full advantage of the heterogeneous physical resources. Finally, we implement the cloud platform and develop some typical video surveillance services on the cloud platform. The experimental results show that our resource allocation approach outperforms other widely used approaches.