基于仿生自主神经系统的节能高效云调度研究

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

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及能耗指标,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。

For developing an e ffic ie n t cloud scheduling and management mechanism ta kin g perform ance and energy consum ption in to a cco u n t, this paper proposed a cloud scheduling and management system based on bio n ic autonom ic nervous systems(B A N S ) . I t presented a the ore tical m odel fo r eva lu ating and analyzing im po rtan t perform ance and energy consum ption inde xs.T h e n , it b u ilt a pure p ro fit op tim ization m odel fo r ba la ncing the tra deo ff between perform ance and energy consum ption. A c cordingto the ore tical an a lysis, th is paper fu rth e r adopted o p tim a lity analysis and an autonom ic trig ger mechanism fo r achievingdynam ic and fle x ib le management o f local resources. M e a n w h ile , it developed a h e u ristic algo rithm to obtain a global scheduling strategy fo r dispa tching user requests. E xperim enta l results illu s tra te the im po rtan t tra deo ff between perform ance and energyconsum ption. I t also demonstrates th a t the autonom ic resource management can b rin g a sig n ifica n t increase o f around 6 0 %in the pure p ro fit com pared w ith a tra d itio n a l load balance scheduling m e chanism , and the global request scheduling fu rth e rleads to that the pure p ro fit raises by about 15%.