A Dynamic Resource Allocation Algorithm in Cloud Computing Based on Workflow and Resource Clustering

Abstract

Since the complexity of large-scale and scientific computation, workflow has been used for task decomposition in cloud computing. A dynamic resource allocation algorithm based on workflow and resource clustering is proposed in this paper. The workflow is described by a directed acyclic graph, which represents the precedence relations and communication cost of subtasks. Fussy clustering algorithm is used to group nodes by the features, which includes the computing capability, transmission capability, storage capacity, cost and reliability. Subtasks are mapped to different resource in cluster by multi-objective optimization model. Resource reputation is added to feature vector by feedback mechanism for evaluating resource reliability. Simulation results show the algorithm reduces the completion time and cost. It improves the resource utilization and load balance.