Utility-Based VM Assignment in Cloud Scenarios

The problem of Virtual Machine (VM) assignment in cloud computing is getting more and more attention each year. In this paper, we introduce the utility cost model, which combines both job time cost and the VM rent cost in the Cloud Scenarios. We investigate the interrelationship between the time cost and the VM rent cost, and formalize it as the parallel speedup pattern. We firstly propose the time first algorithm only considering the time cost, with the objective of minimizing the average completion time of all jobs. Additionally, we propose the price first algorithm, which focus on minimizing the rent cost. Based on that two algorithms, we introduce the policy shifting scheduling algorithm, which combines both time cost and rent price at the same time. We then formulate three group-based algorithms by adopting the idea of minimizing the utility cost. There are also time complexity and performance tradeoffs among the three group-based algorithms. Our experimental results demonstrate that our algorithms can achieve very good average utility gains in the real setting.