Performance Modeling and Energy Conservation in Cloud Computing

Abstract

Cloud Computing, one of the hot network services, has attracted extensive users worldwide as a result of various advantages. However, a large amount of energy is consumed in data center behind cloud computing every year with low energy efficiency at most of the time. Thus, most energy consumed in data centers are wasted in keep idle servers running. This project report investigates the amount of cost saved by a Dynamic Right-Sizing Algorithm, Lazy Capacity Provisioning (LCP(w)), working by dynamically adapting the number of active servers according to the workload. The impact factors on the cost reduction are exploited. A general mathematical model is proposed and workloads are from two data centers in real life. The total cost obtaining from the Static Right-Sizing Scheme are used as the cost benchmark to validate the LCP(w) algorithm. Comparisons between LCP(w) and Optimal Offline Solution have been made as well. The validation result shows that cost reduction can be achieved by LCP(w) algorithm, and the amount of energy cost saved by LCP(w) depends on the window size. In the worst case, no future arrival being predicted, LCP(0) can still reduce a certain amount of cost.