云环境下基于神经网络和群搜索优化的资源分配机制

在云环境下，各种闲置资源可以通过池化形成资源池，进而利用虚拟化技术将资源池中的不同资源组合以服务的形式提供给用户使用，因此需要合理而有效的机制来分配资源.针对云环境下资源的特点，将经济学和智能方法相结合，提出了一种基于双向组合拍卖的智能资源分配机制.在该机制中，提出了基于体验质量（quality ofexperience，简称QoE）的威望系统，引入威望衰减系数和用户信誉度，降低拍卖中恶意行为造成的影响，为资源交易提供QoE 支持.对拍卖中的竞价决策，综合考虑多种因素，提出了基于BP 神经网络的竞标价格决策机制，不仅可以合理确定竞标价，而且使价格可以动态适应市场变化.最后，由于组合拍卖胜标确定问题是NP 完全的，因此引入群搜索优化算法，以市场盈余和总体威望为优化目标，得到资源分配方案.仿真研究结果表明，该机制是可行和有效的.

Resource Allocation Scheme Based on Neural Network and Group Search Optimization in Cloud Environment

In cloud environment, all kinds of idle resources can be pooled to establish a resource pool, and different kinds of resources can be combined as a service to the users through virtualization. Therefore, an effective scheme is necessary for managing and allocating the resources. In this paper, economic and intelligent methods are employed to form an intelligent resource allocation scheme based on double combinatorial auction with respect to the characteristics of resources in cloud environment. In the proposed scheme, a reputation system on the basis of quality of experience (QoE) is devised, and the reputation attenuation coefficient and the user credit degree are introduced to decrease the negative effects of malicious behaviors on resource auctions, providing QoE support to resource dealing. In order to determine bidding price rationally, a bidding price decision mechanism based on back propagation (BP) neural network is presented to comprehensively consider various influence factors to make price adapt to the fluctuating market. Finally, due to the fact that the problem of winner determination in combinatorial auction is NP-complete, a group search optimization algorithm is adopted to find the specific resource allocation solution with market surplus and total reputation optimized. Simulation studies are conducted to demonstrate the feasibility and effectiveness of the proposed scheme.