Optimizing multi-dimensional resource utilization in virtual data center

Resource Scheduling is a centerpiece of data centers. However, most previous works concentrate only on one-dimensional model, which ignoring the fact that multiple resources such as CPU, memory and network bandwidth are consumed simultaneously. As cloud computing allows uncoordinated and heterogeneous users to share a data center, competition for multiple resources has become increasingly severe. Motivated by the differences on integrated utilization obtained from different packing schemes, in this paper we model this scheduling problem as a multi-dimensional combinatorial optimization problem with constraint satisfaction. Due to the NP hardness, a novel heuristic algorithm called Multi-Attribute Decision based Integrated Resource Scheduling (MADIRS) is presented to obtain the approximate optimal solution. The simulation results show that our algorithm significantly outperforms the existing approaches in terms of efficiency and performance for a diverse set of workloads.