Toward an Autonomic Energy Efficient Data Center

Autonomic computing is a promising solution to managing complex distributed computing systems, including helping to reduce energy consumption. In this paper, we have studied the use of policy-based autonomic computing techniques to make data centers self-optimized in terms of energy consumption while still trying to meet SLA requirements. Our work has focused on using two-level autonomic managers which use dynamic frequency scaling (DFS) and dynamic server provisioning (DSP) as tools to optimize energy consumption. Based on predefined policies and numbers of SLA violations, an autonomic manager can change CPU working frequency or change the number of active servers available to an application. Experiments are run on ADCMSim, an evolving simulation platform for studying autonomic algorithms for managing energy consumption in data centers. The result shows that multi-level policy based solution yields promising results in terms of reducing energy consumption and limiting the number of SLA violation.