Abstract: To solve the problems of frequent violation of Quality of Service (QoS) constraints,insufficient resource utilization,and low revenue of cloud application provider in cloud resource allocation,an optimal cloud resource allocation strategy for cloud application provider’ revenue was proposed.In this strategy,the cloud application provider obtained the probability distribution of visitor volume in a certain period of time based on experience.Then,based on the idea of newsvendor model for reference,the method of minimum loss expectation was adopted to calculate the optimal amount of cloud resource allocation.Fixed,segmented,and dynamic prices were common cloud resource pricing methods.Based on the above methods,the fixed price cloud resource configuration strategy,segmented price cloud resource configuration strategy and dynamic price cloud resource configuration strategy were designed.When visitor volume was random,the optimal cloud resource allocation amount obtained by this strategy could maximize the benefits of cloud application provider and make full use of cloud resources.There was no violation of the QoS constraint at the level of Infrastructure as a Service (IaaS) and Platform as a Service (PaaS).

Key words: cloud application provider, cloud resource, revenue-driven, allocation strategy, cloud computing system

摘要： 为解决云资源配置中存在服务质量（QoS）约束违反频繁、资源利用率不充分、云应用提供商收益低的问题,提出面向云应用提供商收益的最佳云资源配置策略。在该策略中,云应用提供商根据经验得出一定时间内用户访问量的概率分布;借鉴报童模型思想,采用损失期望值最小原则计算最佳云资源配置量;固定价格、分段价格、动态价格是较为常见的云资源计费模式,进而基于上述方法设计出固定价格的云资源配置策略、分段价格云资源配置策略、动态价格云资源配置策略。当用户访问量随机时,通过该策略求得的最佳云资源配置量能够使软件即服务提供商收益最大,能充分利用云资源,在基础设施即服务（IaaS）、平台即服务（PaaS）层面不存在违反服务质量约束。

关键词: 云应用提供商, 云资源, 收益驱动, 配置策略, 云计算系统