Deploying applications in leased cloud infrastructure is increasingly considered by a variety of business and service integrators. However, the challenge of selecting the leasing strategy - larger or faster instances? on-demand or reserved instances? etc.- and to configure the leasing strategy with appropriate scheduling policies is still daunting for the (potential) cloud user. In this work, we investigate leasing strategies and their policies from a broker's perspective. We propose, CoH, a family of Cloud-based, online, Hybrid scheduling policies that minimizes rental cost by making use of both on-demand and reserved instances. We formulate the resource provisioning and job allocation policies as Integer Programming problems. As the policies need to be executed online, we limit the time to explore the optimal solution of the integer program, and compare the obtained solution with various heuristics-based policies; then automatically pick the best one. We show, via simulation and using multiple real-world traces, that the hybrid leasing policy can obtain significantly lower cost than typical heuristics-based policies. © 2013 Springer-Verlag.
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||19th International Conference on Parallel Processing, Euro-Par 2013|
|Period||26/08/13 → 30/08/13|