Grant: Research Tools and Instruments - Category 1 (<$150,000)

Cloud computing has recently emerged as an attractive paradigm for providing services over the Internet. Despite its short history, it has become a multi-billion dollar industry that attracts many companies like Amazon, Google, Microsoft, Oracle and IBM. However, the research on cloud computing is still in its infancy. Many key issues in cloud computing, including multi-tenancy, resource and performance management, energy management, security and pricing, are still largely unsolved. As a result, cloud computing has become a hot research topic in the systems community in the past two years. Our current research focuses on resource management issues in cloud computing environments. Specifically, we study the workload characteristics and scheduling algorithms for cloud datacenters. Workload characterization has profound impact on the design of scheduling policies, server consolidation algorithms and datacenter capacity management. At present, we have performed our analysis using traces from production datacenters. Our on-going work is to develop clustering algorithms which generate distinct VM classes, such that VMs in each class have similar scheduling characteristics. Based on the findings in our workload characterization work, we aim at devising fair and efficient job scheduling policies for cloud computing applications. The validation of our methods requires a large cluster testbed to conduct realistic experiments. This application aims at building such a cluster which is necessary for producing high quality research results.