云计算机环境资源配置技术研究

摘要

For the problem of how to reduce the manual operation in the process of cloud computer resources management to achieve resource adaptive management,a resource management strategy of neural network load forecasting algorithm based on load similarity and multi⁃objective genetic algorithm based on hybrid grouping encoding is proposed. The simulation experiments for physical nodes of different scale and different neural networks were conducted in the environments of Matlab and CloudSim respectively. The experimental results show that the Elman neural network load forecasting algorithm based on load similarity adapts to the dynamic characteristics of cloud computer system,and can effectively improve the accuracy of resource load fore⁃casting. The resource management strategy of multi⁃objective genetic algorithm based on hybrid grouping encoding can reduce the frequency of the virtual machine migration,and optimize the use quantity of physical machines.%针对云计算机资源管理过程中如何减少人工操作，达到资源自适应管理这一问题，提出了基于负载相似度的神经网络负载预测算法和基于混合分组编码的多目标遗传算法的资源管理策略。针对不同神经网络和不同规模物理节点，分别在Matlab和CloudSim环境下进行了仿真实验。实验结果表明，基于负载相似度的Elman神经网络负载预测算法适应云计算机系统的动态特点，可以有效提高资源负载预测的准确性；基于混合分组编码的多目标遗传算法的资源管理策略能在减少虚拟机迁移次数的同时优化物理机使用数量。