A framework for a cost-efficient cloud ecosystem

Abstract

Cloud Computing has developed rapidly over the last ten years with worldwide spending on public and private cloud hosting passing the $32bn. mark this year, and savings to businesses using the cloud are typically stated as better than 30% due to the cloud's ability to better take advantage of economies of scale. Yet, in actuality, in commercial data centres the utilization of resources still remain low. In this paper we introduce a new approach from CloudWave which uses user-defined performance goals as well as monitoring across the Cloud ecosystem as a means for guiding runtime adaptation of both the cloud infrastructure and the hosted service to improve overall performance. In addition, inspired by the current industry trend of DevOps methodology, CloudWave feeds runtime performance and monitoring insights back to the DevOps engineers to enable identification and refactoring of problematic code sections. The potential impact for improving data centre efficiency is huge - as Google researcher John Wilkes has pointed out that at such large scales, "A few percent here, a few percent there, and all of the sudden, you're talking about huge amounts of money". After two years of research, the CloudWave platform is starting to test its ideas on actual applications, which in turn is exposing new avenues for research.