(An) energy efficient resource management based on computing power consumption-adaptive resource group control scheme in clouds

Abstract

With more and more workloads in the cloud data center, the power consumed by computing resources is rapidly increasing in cloud data center. As the cost of consumed power increases, it is necessary to efficiently manage the power consumption due to the computing resources in data center. However, it becomes difficult to manage computing resources energy efficiently because the performance of each computing resource varies according to the hardware specification and its energy consumption may be different. In order to predict the performance (application execution time) and energy consumption of computing resources, many variable relationships must be considered. Therefore, it is impossible to select a server within a time limit. To overcome these limitations, existing Power Efficiency Rank-based heuristic algorithms measure the performance and energy consumption of all computing resources according to various applications. Therefore, it is difficult to manage a lot of computing resources in data center. To overcome these problems, we propose a power consumption-adaptive resource group control scheme to manage cloud resources energy efficiently. Each server is represented by a performance index and a power index, and then servers having similar performance index or power index are grouped. In order to derive information on performance (application execution time) and power consumption of each group, a group profiling process is performed, and a genetic algorithm is applied based on the preliminary information to allocate the requested applications to the server. Our proposed scheme measures and analyzes the performance (task execution time) and power consumption for various operation patterns such as CPU and IO compared to the existing schemes. By reducing the time spent on the pre-profiling process for obtaining information of each server, it showed that the cloud resources can be managed energy-efficiently.