Virtual Instance Resource Usage Modeling: A Method for Efficient Resource Provisioning in the Cloud

摘要

Cloud computing is a promising framework providing a variety of solutions, ranging from software services to infrastructure services through the mechanism of customizable virtual instances. The cloud manager is responsible for resource provisioning for these instances to provide guaranteed performance but at the same time avoiding underutilization of the platform. In this paper, we introduce a novel method for modeling the resource usage of VIs which allows for better VI placement with more efficient resource usage in the physical infrastructure. Our proposed framework uses the mixture of Gaussians to model each virtual instance resource usage. Then for placement, a modified probabilistic bin packing method is been proposed to take advantage of modeling for placing virtual instances. We compared our scheme with other bin packing methods that use rigid statistical models, and the results support the efficiency and accuracy of our method which leads to more than 50% resource saving while preserving the given performance guarantee.