Caching techniques for rapid provisioning of virtual servers in cloud environment

摘要

Provisioning a virtual server instance in cloud goes through an elaborate workflow, characterized by user request submission, search for the requested image template in image repository, transfer of the template file to compute hosts, followed by expansion and boot up. Optimizing any step in this workflow is crucial in reducing the service time. In this work, we focus on reducing average service time by masking the template file transfer time from repository, and preparing a VM instance beforehand to service a request instantaneously. We use a strategy of pre-provisioning multiple VM instances from different templates. The instances to be pre-provisioned are determined based on request history. We show the benefits of our method using request trace from an enterprise grade cloud environment. Simulation results show more than 50% improvement in reducing average service time while delivering a server instance.