Bi-objective online scheduling with quality of service for IaaS clouds

This paper focuses on the bi-objective experimental analysis of online scheduling in the Infrastructure as a Service model of Cloud computing. In this model, customer have the choice between different service levels. Each service level is associated with a price per unit of job execution time and a slack factor that determines the maximal time span to deliver the requested amount of computing resources. It is responsibility of the system and its scheduling algorithm to guarantee the corresponding quality of service for all accepted jobs. We do not consider any optimistic scheduling approach, that is, a job cannot be accepted if its service guarantee will not be observed assuming that all accepted jobs receive the requested resources. We analyze several scheduling algorithms with different cloud configurations and workloads and use the maximization of the provider income and minimization of the total power consumption of a schedule as additional objectives. Therefore, we cannot expect finding a unique solution to a given problem but a set of nondominated solutions also known as Pareto optima. Then we assess the performance of different scheduling algorithms by using a set coverage metric to compare them in terms of Pareto dominance. Based on the presented case study, we claim that a simple algorithm can provide the best energy and income trade-offs. This scheduling algorithm performs well in different scenarios with a variety of workloads and cloud configurations.