Efficient Cloud Resource Scheduling for Stochastic Demand with Heterogeneous Cost Models

摘要

Distributed cloud platforms facilitate service providers to deliver geographically dispersed online services to a large number of users all over the world, while the aggregated user requests introduce stochastic demands for various resources in different cloud data centers. Resource scheduling in cloud platform is of high complexity due to various nonlinear cost models and multidimensional demand stochasticity. Existed scheduling algorithms generally utilize linear cost model, thus is hard to utilize budget efficiently. We proposed an efficient Heterogeneous Cost models oriented cloud Resource Scheduling (HCRS) to address the problem. Experiments results show that HCRS increase revenue by up to 40% than that in previous mean demand based algorithm and is of low enough complexity. Therefore, HCRS can be used as a candidate for scheduling in cloud platforms with heterogeneous cost models.