Energy Efficient Data Center Management System and Implementation of Data Center Emulator

Abstract

This thesis presents an energy efficient data center management system and implementation of data center emulator. The explosive growth of cloud services has increased the use of data centers, which are the technical base of cloud services. As the data center usage increases, problems in the data center have also been highlighted. Especially a large amount of electric energy consumption and carbon dioxide emissions are arising issues both economically and environmentally. The electrical energy used by the server is one of the main cause of the electrical energy consumed in the data center. The average server utilization rate is low because the entire data center computing resource must be ready according to the maximum service request amount. And, the idle server has low energy efficiency due to an idle power which is half of peak power. Therefore, various research to increase energy efficiency of data center. We propose the energy efficient data center management system and implement a data center emulator that can be applied and verified the data center management system. The proposed emulator generates each data center component based on open source which is actively used in the data center field. An emulator operator can generate and modify the data center through the Web dashboard, and web dashboard also provide graphical charts of the various results generated by the emulator. The proposed system estimates resources needed for the data center in the future through a probabilistic resource estimation module. The proposed system consolidates the host machines based on the estimation result, and manages the power of the host machines to reduce the power consumption of the data center. Software-defined network controller is used to monitor the data center network to select QoS-aware migration path during consolidation. The proposed system is applied and verified through the implemented data center emulator.