Performance Modeling and Energy Conservation in Cloud Computing

Abstract

Cloud computing, a vibrant network service, is gaining worldwide popularity because of its convenience and flexibility. Although the cloud provides benefits for providers and users, its inefficient energy utilization has drawn international attention. In the report, a promising scheduling algorithm derived from a mathematical model was investigated to reduce power consumption in data centres while not degrading its performance. The methodology, Lazy Capacity Provisioning (LCP), achieved cost savings by dynamically adjusting the number of active servers based on the workload. The solution of online LCP method was compared with an offline optimal solution by using two real data load traces. Additionally, the cost reduction of LCP was evaluated by providing a conservative estimation from a static right-sizing scheme. The influence of workload shape was also studied. Even though the performance of LCP solution is affected by different factors, it generally can make a trade-off between cost reduction and quality of service in cloud computing applications.