A Novel way of Improving CPU Utilization In Cloud

Abstract

Cloud computing is the collection of best software practices derived from dis-tributed computing systems like cluster and grid. Emerging as a superior design amongst them. Cloud not only has the capability of batch processing and massive massive parallel data computation like its parents but also, things hitherto was not imagined. The main concentration of our work is on IaaS(Infrastructure as a service) aspect of cloud. Wherein, the customer avails the computation resource over simple web interface. though the cloud has most of the good design choices of its parents, certain drawbacks of them have crept in. This may be the result of the lack of ecient preexisting tools or naive implementation/setup of system at certain layer. In our work, we've optimized the performance by identifying some of those implementation choices making the overall system more ecient. This thesis is an extension of the work on Nephele. Which is a parallel data processing framework(still experimental). Nephele facilitates on demand alloacation of resource by deducing few infor-mations from the users. Using these informations, Nephele splits the overall job to smaller stages and resource for the same is allocated. By doing so, unnecessary resource allocation like in conventional systems is being avoided. thus, eventually increasing the CPU utilization. In our proposal, we deduce few more informations from the user which helps in making more optimal scheduling of the resources.