Multi-Resource Packing for Job Scheduling in Virtual Machine Based Cloud Environment

To efficiently schedule jobs with highly diverse resource requirements along CPU, memory and bandwidth for job performance and resource utilization in a virtual machine based cloud environment, the multi-resource job scheduler is proposed to pack tasks to virtual machines under the notion of fairness and efficiency. Given the definition of job scheduling proportional fairness and utility function, the multi-resource job scheduling algorithm which fulfills capacity constraints of virtual machines is conducted. Comparative analysis illustrates our scheme improves average job completion time by preferentially grouping jobs that has different resource requirements. Compared to existing methods, multi-resource packing algorithm significantly improves the cloud system's resource utilization, yet with a substantial reduction of average job completion times.