Multi-objective scheduling for scientific workflow in multicloud environment

Providing resources and services from multiple clouds is becoming an increasingly promising paradigm. Workflow applications are becoming increasingly computation-intensive or data-intensive, with its resource requirement being maintained from multicloud environment in terms of pay-per-use pricing mechanism. Existing works of cloud workflow scheduling primarily target optimizing makespan or cost. However, the reliability of workflow scheduling is also a critical concern and even the most important metric of QoS (quality of service). In this paper, a multi-objective scheduling (MOS) algorithm for scientific workflow in multicloud environment is proposed, the aim of which is to minimize workflow makespan and cost simultaneously while satisfying the reliability constraint. The proposed MOS algorithm is according to particle swarm optimization (PSO) technology, and the corresponding coding strategy takes both the tasks execution location and tasks order of data transmission into consideration. On the basis of real-world scientific workflow models, extensive simulation experiments demonstrate the significant multi-objective performances improvement of MOS algorithm over the CMOHEFT algorithm and the RANDOM algorithm.