Dealing with structural changes on provisioning resources for deadline-constrained workflow

Abstract

Cloud computing has received an increasing attention in the past years thanks to its new model of resources provisioning. One well-known challenge in this context is to make an appropriate decision when mapping tasks to resources considering multiple objectives that are often contradictory. This problem has become more complex, mainly for workflow applications which impose dependencies and order constraints between tasks. The resources provisioning problem for workflow applications has been widely studied in the literature. Nevertheless, the existing works consider only static workflows. They neglect the need to change workflow instances while they are being executed. This functionality has become a major requirement to deal with unusual situations and evolution. In fact, the strong competition in which companies are involved often lead them to adapt their workflows to face new regulation laws, changes in the customer behavior and some exceptional situations. In this paper, we present a new provisioning algorithm based on the meta-heuristic optimization technique, particle swarm optimization. It takes into account dynamic structural changes of a workflow, while satisfying some performance criteria defined by the user. We address general flow structures including sequential, parallel, choice and loop patterns. We conducted our experiments using CloudSim and various well-known scientific workflows of different sizes. Experimental results show that our approach has a promising performance.