Performance Evaluation Methodology for Cloud Computing using Data Envelopment Analysis

Leonardo Menezes De Souza, Marcial Porto Fernandez
Cloud Computing is a new distributed computing model based on the Internet infrastructure. The computational power, infrastructure, applications, and even collaborative content distribution is provided to users through the Cloud as a service, anywhere, anytime. The adoption of Cloud Computing systems in recent years is remarkable, and it is gradually gaining more visibility. The resource elasticity with the cost reduction has been increasing the adoption of cloud computing among organizations.
more » ... ong organizations. Thus, critical analysis inherent to cloud's physical characteristics must be performed to ensure consistent system deployment. Some applications demand more computer resources, other requests more storage or network resource. Therefore, it is necessary to propose an approach to performance measurement of Cloud Computing platforms considering the effective resource performance, such as processing rate, memory buffer refresh rate, disk I/O transfer rate, and the network latency. It is difficult to discover the amount of resources are important to a particular application. This work proposes a performance evaluation methodology considering the importance of each resource in a specific application. The evaluation is calculated using two benchmark suites: High-Performance Computing Challenge (HPCC) and Phoronix Test Suite (PTS). To define the weight for each resource, the Data Envelopment Analysis (DEA) methodology is used. The methodology is tested in a simple application evaluation, and the results are analyzed.