Selective Task Scheduling for Time-Targeted Workflow Execution on Cloud

With increasing demand of data-intensive applications, a workflow with multiple distributed parallel applications becomes a useful way to build various large scale data processing applications on cloud environments. However, previous scheduling approaches for executing workflows are focused on estimating minimum make span of executions on pre-configured clusters. In this paper, we present two time-targeted selective task scheduling schemes which consider malleability and rigidity of each task, named SDER (Sequential Deployment with Expanded Resources) and DDMS (Distributed Deployment on Multiple Streams). These scheduling algorithms deploy all task of workflow in sequential and distributed ways each, and estimate optimized number of resources for completing workflow execution within given deadline. We show the experimental results which compares the performance of two scheduling schemes for various workflows.