A Multi-Stage Optimization Framework for Efficient Cloud Workflow Scheduling

Abstract

Cloud Workflow scheduling is still a challenging task with increasing workload over servers. This issue is due to dynamic nature of tasks arrived and their execution dependencies over heterogeneous resources. In such condition, efficient scheduling is required to virtual machines in cloud environment to reduce high latency requirements and efficient utilization of resources. Recently, researchers have contributed in this field and developed many optimization approaches to reduce the operational costs and make span time but still there is room for improvement as growing need of resources. Motivated by this, the paper presented a multi-stage and multi-objective based workflow scheduling algorithm. In first stage of the algorithm task priotization is performed using particle swarm optimization then resource allocation matrix is generated using ensemble learning and finally meta-heuristic optimization is used for allocating the optimal number of resources to respective priority queue tasks. The entire working model is developed on MATLAB and simulation is performed with varying number of tasks as well as virtual machines (VMs). The result was compared in terms of make span time with existing works and achieved better performance.