Research About Virtual Resource Optimization Allocation For Cloud Computing

Cloud computing is envisioned as the future IT service paradigm and has attracted tremendous attention from academia and industry. With cloud computing, cloud customers are able to gain virtually infinite capability of computing and storage through simple devices.The increasing requirements on cloud computing entail building up large numbers of large-scale data centers which require a surprising number of energy. With the gradual depletion and price escalation of traditional energy, operating data center in an energy efficient way is an emerging urgent problem.Firstly, the thesis introduced the basic concepts, classification, characteristic and reaserch focus of cloud computing, research in virtual resource allocation at home and abroad and inadequate. With the limit of server performance index, to efficiently decrease the energy consumption of data center, the thesis model virtual resource allocation problem as a problem of path construction, improving elitist strategy for ant system(EAS) to optimize resource allocation scheme. Simulation results illustrate that strategy improve utilization of servers, thus efficiently decreasing the energy consumption of data center. To greater decrease the energy consumption of data center, the thesis introduce Dynamic Voltage and Frequency Scaling (DVFS) technique and formulate the energy efficiency virtual resource allocation for cloud computing as a multi-objective optimization problem, which is then solved by an excellent multi-objective evolutionary algorithm non-dominated sorting genetic algorithm Ⅱ (NSGA-Ⅱ). The simulation results reveal that the strategy can successfully generate schedule scheme of different numbers of servers-VMs with diverse characteristics in a reasonable time period and decrease the total operating energy of data center effectively. In the last part of the thesis, For the original NSGA-Ⅱ is difficult to improve the ability to search the local solution space effectively through of evolutionary procedure, resulting in precocious puberty, the thesis improve mutation of NSGA-Ⅱ, improve the ability to search the local solution space and maintain the diversity of the population, thus preventing precocious puberty. At the same time, the thesis improve multi-objective optimization model, introduce the turning on/off cost to be objective function. The simulation results reveal that the strategy can decrease the operating energy of data center more effectively under the premise to ensure SLAs.