Performance Estimation of Fault-prone Infrastructure-as-a-Service Cloud Computing Systems and their Cost-aware Optimal Performance Determination

Abstract

The cloud computing paradigm enables elastic resources to be scaled at run time satisfy customers' demand. Cloud computing provisions on-demand service to users following a pay-per-use pattern. This novel paradigm enables cloud users or tenant users to afford computational resources in the form of virtual machines (VMs) as utilities, just like electricity, instead of paying for and building computing infrastructures by their own. Performance estimation of clouds is one of key research challenges and draws great research interests. For this purpose, we develop a comprehensive stochastic framework for estimation of performance of IaaS clouds with fault-prone instantiation and retrials of faulty instantiation. Our proposed approach is capable of analyzing several performance metrics under variable system conditions. A comparative study based on an actual campus cloud is carried out and its corresponding confidence interval validation suggests the correctness and accuracy of theoretical performance results. To optimize cloud performance, we also formulate the developed stochastic model into an optimal responsiveness determination problem with the aim of minimizing averaged system responsiveness with rejection rate and system cost constraints. An intelligent algorithm is introduced to obtain near-optimal solutions.