Data intensive, computing and network aware (DCN) cloud VMs scheduling algorithm

Abstract

The emergence of cloud computing technology aims at sharing resources such as storage, knowledge, computation and information for scientific research at an expanded scale. The application about associated data are deployed by the cloud users as paying the bills when they get due. Such data-intensive applications are normally commanded by the virtual machines (VMs). Data at a large scale are analyzed by data intensive applications and their replications are made for diffusing them among various geographical sites. In case the very spot of execution of a job gets no data replication, then data are streamed from a distant site. The overall execution of the job will deteriorate with such data transfer from remote sites. The decisive factors in the performance of these applications are workload volume, network status between storage nodes SNs and CNs, workload types I/O computation or I/O data-intensive and CPU attributes into computing node CN. Thus, the completion time differs according to the application jobs in workload on the basis of retrieval of vast data and decision of VM placement. Our proposal for obtaining elevated performance in the completion time of overall jobs along with alleviating the throughput of cloud links is VMs placement that takes both the I/O data and computation resources into consideration. This algorithm tries to diminish the completion time of overall jobs (including both time for data transfer and computing time). The CloudSim Simulator results show that our algorithm with the ability of significantly increasing and decreasing the performance of overall performance and the completion time of average jobs respectively instead of earlier proposal for VMs placement in literature review.