Abstract:

The energy consumption optimization of resources allocation in the cloud data center with heterogeneous physical servers is an NPHard combinatorial optimization problem. When the number of servers is large, the solution space is large as well. It is therefore quite difficult to get the optimal solution within reasonable time. We propose a scalable distributed scheduling method (SDSM) based on the "Divide and Conquer" idea from the aspect of scheduling model, which implements an efficient resource allocation algorithm in heterogeneous cloud environments. When the number of physical servers in cloud data centers is large, the servers will be divided into several server clusters, and then in every cluster we use Choco to model and solve the constraint satisfaction problem (CPS) of energy consumption optimization in heterogeneous cloud data centers, which can obtain the optimal resource allocation and greatly reduce its complexity. Finally, we compare the proposed SDSM and the nonscalable scheduling method through experiments, and the experimental results show that the SDSM has obvious advantage in largescale cloud resource allocation.

Key words: cloud computing;resource allocation;energy optimization;divide and conquer;constraint satisfaction problem (CPS)