考虑需求多样性的云服务最优分级定价策略研究

In practice, pricing tiered cloud services is one of the common pricing strategies used by cloud service prividers (CSPs). However, differential pricing strategy on cloud service, supporting the CSP make an optimal policy of service level (SL) and price for cloud service, is rarely discussed in academia. From a perspective of dynamic and long-term cloud market, optimal strategy on tiered pricing the cloud service, where demand diversity is taken into account, is investigated. First of all, the SL for the cloud service is measured by metrics including service capability (e.g., service functionality, maximum concurrent users) and quality of service (e.g., response time, availiability). Secondly, demand diversity is measured in terms of customers’ preference of SL and its distribution. Two kinds of functions of willingness to pay for SL and three distribution functions of customers’ preference are constructed to represent the heterogenous demand. Finally, with the consideration of cost caused by violations, optimal strategies of SL and price are sequentially studied in three scenarios of cloud service market, i.e., CSP offers single SL or multiple tiered SLs in a monolist market and the market with dual competitors who simultaneously provide tiered SLs. Behaviors of the CSP/CSPs and customers are formulated in a bilevel programming model. Moreover, the optimal strategy is numerically constructed by utilizing coevolutionary algorithms due to the complexity of the defined models.This project is targeted to the national economic development planning and the long-term demand of cloud service industries by delivering a theoretical foundation and suggestions to cloud service firms.

分级定价是产业界云服务提供商(cloud service providers, CSPs)普遍采用的定价策略，但云服务最优服务等级及价格策略的研究还很少。考虑用户需求多样性，从动态、长期视角研究云服务的最优分级定价策略。首先，定义云服务的两维服务等级指标,即服务能力(如,服务提供的功能、最大并发用户数)及服务质量(如,响应时间、可用性)。然后，从用户的服务等级偏好、不同类型用户的市场分布两个基本特征度量需求多样性，建立两种用户支付意愿函数、以及三种用户类型概率分布函数。最后，在垄断CSP只提供一个服务等级、同时提供多个服务等级以及竞争CSPs同时提供多个服务等级三种管理情境下，分别研究考虑违约成本的最优服务等级及价格策略，将问题分别表示成两层规划模型，并对模型用协同进化算法数值求解。选题符合国家重大战略规划和产业发展需求，能对云计算产业发展提供理论支持。

服务等级协议（Service level agreements, SLA）是云服务提供商（Cloud Service Provider，CSP）与用户间就所提供云服务关键性能指标的约定，其中规定了云服务的功能、价格、服务等级或服务质量（Quality of Service, QoS）、以及服务保障（或违约补偿）等性能参数。本项目围绕SLA中的价格以及QoS这两个关键参数，就以下五个方面的内容进行了研究。（1）云服务的服务质量预测，为度量云服务的违约概率并建立用户效用函数，选取响应时间作为云服务QoS度量指标，并开发了QoS预测方法；（2）考虑分级违约补偿的云服务定价及资源配置决策；（3）基于SLA的IaaS云服务分级定价策略；（4）基于SLA的垄断/竞争云服务提供商价格及服务保障提供策略；（5）聚集与平台型CSP的差异化策略，研究了资源型以及非资源型两个非对称云市场对SaaS的最优开放策略及价格决策。本研究丰富和拓展了云服务运营、信息系统经济学等相关理论体系，可以为CSPs的差异化经营决策提供理论依据和技术支持，同时有助于引导云服务产业的全面协调可持续发展。