Enhancing fork and join based workflow scheduling model in cloud computing

Enhancing fork and join based workflow scheduling model in cloud computing

연구 분야: Networking

학회: Cluster Computing

In the modern era, scientific computing demands efficient workflow execution for timely results and optimal resource utilization. With the rise of Infrastructure as a Service (IaaS) cloud platforms, researchers and practitioners focus on innovating workflow execution, particularly in minimizing makespan—the duration to complete a workflow. In this paper Modified Fork and Join Scheduling algorithm is proposed which is termed as MFJS. Where the study is focused on the Fork-Join heuristic method to compute the priority of the tasks. Here, tasks are drawn from real scientific workflows represented as Directed Acyclic Graphs, including Montage, SPIHT, AIRSN, and sets of randomly generated workflows in the Cloud Computing Environment (CCE). Additionally, the entry task of the workflow is duplicated across all virtual machines. By considering the workflow’s dependency on resource constraints, the proposed methods aim to achieve an optimal makespan compared to state-of-the-art algorithms such as HEFT, CPOP, ALAP, QLHEFT and Modified Min-Min. In the example examined in this paper, MFJS methods reduce makespan by approximately 13% for HEFT, 3% for ALAP, and 29% for CPOP. Furthermore, various Quality of Service (QoS) metrics, including Speedup, Efficiency, and Average Resource Utilization, are used for performance analysis. Result of the proposed algorithm is validate with the statistical analysis using ANOVA testing. In each case, simulation result shows that the MFJS algorithm outperforms the state-of-art algorithms.