数据中心多维感知的资源联合优化调度策略

Multi-Dimensional Aware Scheduling for Co-optimizing Utilization in Data Center

Resource Scheduling is crucial to data centers. However, most previous works focus only on onedimensional resource models which ignoring the fact that multiple resources simultaneously utilized, including CPU, memory and network bandwidth. 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 take the scheduling problem as a multidimensional combinatorial optimization problem with constraint satisfaction. With NP hardness, we present Multiple attribute decision based Integrated Resource Scheduling (MIRS), and a novel heuristic algorithm to gain the approximate optimal solution. Refers to simulation results, in face of various workload sets, our algorithm has significant superiorities in terms of efficiency and performance compared with previous methods.