面向Deadline的云代理资源预留策略

Strategy of deadline-oriented cloud resource allocation via cloud broker

由于传统的代理模式在应对短期用户大量到达高峰和实例利用率上存在优化空间,为节约资源和节省开支,提出了一种新的面向Deadline的云服务代理构想,利用中间件技术整合用户请求和实例资源。分析发现,面向Deadline的云代理能有效平滑需求高峰,并大幅度提高实例利用率。针对用户deadline相同与不同2种模式,分别设计了贪心算法和EDF平移算法,实现了高峰平滑和利用率提升,降低了开支。实验表明,相比于未采用云代理,面向Deadline云服务代理处理大量用户请求时能减少50%以上的费用,具有良好的应用前景。

A new cloud service broker was introduced,which can save cost and resource through the integration of user needs, purchasing for long-term low-cost calculation instances for users to save resources and achieving cost savings purposes. On this basis, Deadline for cloud service proxy instance reserve Methods was proposed：The long-term strategies into short-term studies, and by Greedy and EDF （earliest deadline first） approach to user requests for processing. The two approaches can effectively smooth the request and improve the usage of instances, thereby reducing the cost. The simulation experiments show that, the Deadline-oriented cloud broker service can reduce more than 50% of the cost of a large number of user requests, and have good prospects.