Efficient Analysis of Resource Availability for Cloud Computing Systems to Reduce SLA Violations

Resource availability of cloud computing systems is vital for today's information infrastructures. However, the constituent computing nodes are not fault free. The availability requirements on the delivered computing resources should be defined clearly by using a formal, contractual agreement, known as the Service Level Agreement (SLA), between providers and customers. To reduce the risk of various SLA violations, an efficient analysis of resource availability is important. This paper proposes a new analytical approach based on multi-valued decision diagrams (MDD) for the efficient resource availability analysis of cloud computing systems with heterogeneous, multi-state computing nodes. Particularly, a novel and efficient MDD construction method is presented to generate compact MDD models encoding different amounts of cumulative computing resources. Two detailed case studies are performed to illustrate basics and application of the proposed approach to reduce SLA violations and guarantee the availability requirements on the delivered computing resources. Benchmark studies are further conducted to show efficiency of the proposed MDD-based approach as compared with the continuous-time Markov chains-based method and the universal generation function-based method.