Virtual Machine based Hybrid Auto-Scaling for Large Scale Scientific Workflows in Cloud Computing

Abstract

Scheduling the Scientific Workflows (SWf) task onto ideal cost-effective resources in a distributed system remains a significant problem that demand for auto-scaling in a cloud computing environment. Even though cloud computing has considered as a quintessential platform for scientific users to perceive deadline constrained large-scale computations. However, complexity increases with dynamic increase in network size which makes mapping of resources a NP-hard problem. To overcome this situation, a novel hybrid auto-scaling strategy is proposed. The hybrid auto-scaling comprises of on-demand and spot instances pricing model for SWf computation under deadline and budget constraint. Many auto-scaling strategy have been proposed in the earlier works, but nevertheless, there is an ample scope for new auto-scaling strategy for efficient scheduling SWf in cloud environment. Moreover, hybrid auto-scaling approach for SWf considering spot and on-demand instances pose a challenge, since the approach has to estimate the proper amount and type of virtual machine instances to acquire and dynamically allocate the number of instances under spot or on-demand pricing model depending on the SWf need. A flexible hybrid auto scaling policy is proposed for scheduling SWf efficiently. Experimental results reveal the promising potential on the proposed algorithm with regard to minimization in makespan and cost of SWf under deadline and budget constraint in cloud environment.