Task-Aware Cloud Instance Selection

Abstract

Commercially available cloud computing offers a viable alternative to grid computing. Researchers and organizations can utilize cloud computing to process analytical tasks without the need to invest in hardware. When the number of tasks rises, having automated means of deploying these tasks to a commercial cloud, huge cost savings and availability benefits can be reaped. This research sets out to devise a decision mechanism to deploy analytical tasks to commercial cloud offerings in a cost-efficient way. This is done by a literature review, analysis of the portfolios of cloud providers and an experiment where cloud instances are benchmarked. The results of these experiments were incorporated in a decision tree, a decision algorithm and a proof of concept cloud decision tool, supporting the cloud instance decision.