| Summary: | Cloud computing has become the main source for executing scientific experiments. It is an effective technique for distributing and processing tasks on virtual machines. Scientific workflows are complex and demand efficient utilization of cloud resources. Scheduling of scientific workflows is considered as NP-complete. The problem is constrained by some parameters such as Quality of Service (QoS), dependencies between tasks and users' deadlines, etc. There exists a strong literature on scheduling scientific workflows in cloud environments. Solutions include standard schedulers, evolutionary optimization techniques, etc. This article presents a hybrid algorithm for scheduling scientific workflows in cloud environments. In the first phase, the algorithm prepares tasks lists for PSO algorithm. Bottleneck tasks are processed on high priority to reduce execution time. In the next phase, tasks are scheduled with the PSO algorithm to reduce both execution time and monetary cost. The algorithm also monitors the load balance to efficiently utilize cloud resources. Benchmark scientific workflows are used to evaluate the proposed algorithm. The proposed algorithm is compared with standard PSO and specialized schedulers to validate the performance. The results show improvement in execution time, monetary cost without affecting the load balance as compared to other techniques.
|
|---|
