Process placement scheme optimization for energy-efficient data center

The energy cost of a typical data center is usually much higher than those standard office buildings. Inefficiency in energy management can thus hurt a company. Currently, there are many commercial software that emphasize how to measure and control the energy usage. However, most of these available packages are not able to accurately predict behaviors of various applications submitted to a data center resulting in difficulties in energy-efficient optimization. A robust energy-efficient framework is then proposed. The framework allows a process to migrate among machines during execution. This paper presents an algorithm related to a process placement scheme generation employed in such a framework. The algorithm is a heuristic method designed based on Artificial Bee Colony and Pareto-domination. A simulation-based experiment is designed to study our method efficiency during live migration. The results suggested that our generated placement scheme provides 21-38% averaged improvement when compared with traditional scheduling algorithms used in data centers.