Resource Allocation Technology Research In Cloud Computing Environment

A cloud computing system may include tens of thousands of nodes, its complexity, scale and dynamic bring great difficulty to data center maintenance personnel. Therefore, how to reduce manual management of the cloud computing resources, and let the system can be self-management according to its own management strategies, so that to achieve adaptive management of resources became urgent issue and research focus. This need to solve two major problems: first, how to predict the resource load; second, given a appropriate resource allocation policies depending on the load forecast result. Due to resource load forecasting is complex in cloud computing environment, simple prediction model cannot achieve satisfactory results. In addition, traditional single target virtual machine resource allocation optimization algorithm achieve better results only in aspects such as the migration times or the number of nodes, but they are lack of applicability due to poor combined effect.For the first issue, this paper propose a load prediction algorithm based on the similarity of the neural network. This algorithm make full use of the advantages of neural network, while combine the load clustering, finding the optimal neural network training data, so that it can improve forecast accuracy. For the second question, this paper presents the multi-objective genetic algorithm based on hybrid grouping algorithm to solve the virtual machine configuration problem. This algorithm consider two goals, the virtual machine migration times and the number of physical machines use. Experimental results show that, Elman neural network load forecasting algorithm based on similarity load can effectively improve the performance of resource load forecasting, and the multi-objective genetic algorithms based on hybrid grouping algorithm making the reduction of the number of virtual machine migration while reducing the number of physical machines.