Performance modeling and analysis of cloud computing systems

Abstract

Cloud computing has offered a fundamental change in how applications and workloads are being provisioned and managed. It is now possible to employ the power of virtualization to dynamically provision software profiles across different types of cloud infrastructures - public, private and hybrid clouds. This change is responsible for increased utilization of resources at cost effective rates, support for multiple and diverse workloads and the ability to scale resources up and down as necessary. Many companies - both private and open-source have deployed cloud solutions. Of particular interest in recent times have been Openstack, Cloudstack, Eucalyptus and VMware vCloud. While these platforms promise much, it is of interest to truly understand the virtualization overhead and to quantify it, so that an informed decision can be made on whether to provision these platforms for any purposes. The objective of this report is to perform a comprehensive performance analysis – a benchmark of CPU, Disk I/O, Memory I/O and Database workloads under the virtualized environment provided by these platforms. Information was collected through deploying each of these platforms and running these benchmarks under the same conditions. It is also the hope to show without doubt that the performance of these cloud platforms outweighs the overhead which is inherently associated with virtualization.