Promoting and Balancing Resource Utilization via Integration by Exploiting the Time-varying Property of Resource Requirements

Abstract

Virtualization technology improves the flexibility of a cloud data center network (CDCN) in resource allocation and leads to a dramatic increase in the resource utilization. Virtual network embedding or virtual machines placement refers to the problem of providing resources for tenants in CDCN. Most prior literature considers the conservative resource allocation strategy where the maximum resource requirement of virtual requests is always satisfied during their life time regardless of the resource requirement fluctuation. Such methods will definitely result in the resource wastage. By taking into account the time-varying resource requirement, we propose a novel virtual machine placement method, named MIE (Matching-based Integration for Embedding) in this paper. The key part of MIE is to enhance the resource utilization of physical machines (PM) and reduce their utilization fluctuation over time dimension by integrating the requirements with different variation tendencies. Our simulation experiments show that MIE can not only improve the resource utilization of physical machines, but also maintain the resource usage more stable. What's more, the number of active PMs is reduced and the acceptance ratio of virtual requests is enhanced.