Dynamic Cloud Workflow Scheduling with a Heuristic-Based Encoding Genetic Algorithm

Abstract

Cloud computing is a powerful and scalable computing platform that enables the virtualization, share and on-demand use of computing resources. Scientific workflows on clouds are promising for handling computational-intensive and complex scientific computing tasks. The scientific workflow scheduling problem has been regarded as an intractable optimization problem that determines the performance of a scientific cloud workflow management system. The problem becomes even more challenging if the dynamic and heterogeneous characteristics of cloud workflows are taken into account. In order to adapt to the dynamic environment, this paper proposes a hybrid genetic algorithm (HGA) algorithm. Different from the traditional evolutionary algorithms for workflow scheduling that uses a direct encoding scheme, the proposed HGA uses an indirect encoding scheme, i.e., a schedule is encoded as a sequence of heuristic rules. Since there have been some widely-studied heuristic information for scheduling on a directed acyclic graph, this heuristic information is adopted by HGA to improve performance. In addition, under the dynamic batch-processing environment, it is found that the results returned by HGA in the form of heuristic-based can still adaptive to the changes. The experimental results validate that HGA is promising.