C-IRR: An Adaptive Engine for Cloud Storage Provisioning Determined by Economic Models with Workload Burstiness Consideration
Abstract
Being the long dreamed vision of computing as a utility, cloud enables convenient and on-demand access to a large centralized pool of resources via network. The emerging of cloud storage offers a rather feasible solution to handle the sheer amount of information. It is maturing and becoming an alternative for on-premise storage. Thus, for IT enterprises with high demand of storage, a big concern is to determine whether it is more cost-effective to lease storage service over clouds. In this paper, we introduce a cloud storage provisioning engine called C-IRR to help users rationally evaluate the benefits of purchasing new disk drives and comparing it against leasing cloud storage offered by Infrastructure as a service (IaaS) providers. We also discuss issues regarding workload burstiness to achieve potential benefit for each applications in local data centers of SaaS providers. The C-IRR migrates the bursty workloads to the clouds and keeps stable ones locally. Such hybrid storage method achieves at least 20% costs saving in total for SaaS companies by experimental evaluation. In addition, C-IRR engine is of adaptivity about fluctuation of storage pricing and manpower cost increasing after the sensitivity studying.
