基于仿生自主神经系统的节能高效云调度研究*

为了实现兼顾性能和能耗的高效云调度管理机制，提出了一种基于仿生自主神经系统（BANS）的云调度管理系统。建立了理论模型来评估和分析重要的性能和能耗指标，并利用纯利润优化模型均衡性能和能耗之间的制约关系。基于理论分析结果，进一步利用最优性分析和自主触发机制实现了动态灵活的局部资源管理，同时，采用启发式算法来获取面向用户请求分发的全局最优调度策略。实验结果展示了重要的性能-能耗制约关系，同时也表明，相比传统负载均衡调度机制，局部自主资源管理可以在纯利润上带来约60%的显著提升，全局请求调度还将进一步带来约15%的提升效果。

Research on energy-efficient cloud scheduling based on bionic autonomic nervous systems

For developing an efficient cloud scheduling and management mechanism taking performance and energy consumption into account, this paper proposed a cloud scheduling and management system based on bionic autonomic nervous systems (BANS). It presented a theoretical model for evaluating and analyzing important performance and energy consumption indies. Then, it built a pure profit optimization model for balancing the tradeoff between performance and energy consumption. According to theoretical analysis, this paper further adopted optimality analysis and an autonomic trigger mechanism for achieving dynamic and flexible management of local resources. Meanwhile, it developed a heuristic algorithm to obtain a global scheduling strategy for dispatching user requests. Experimental results illustrated the important tradeoff between performance and energy consumption. It also demonstrated that the autonomic resource management can bring a significant increase of around 60% in the pure profit compared with a traditional load balance scheduling mechanism, and the global request scheduling further leads to that the pure profit raises by about 15%.