Abstract: With the vision of anytime anywhere computing, the availability management is playing an important role in pervasive computing. Most research focus on the qualitative analysis. This paper adopts a quantitative approach to attack this topic. Due to the CTMC and SMP can not be applied to the modeling of the availability of pervasive computing system, which has the features of user oriented, non-exponential distributions and associative analysis between user states and system states, this paper presents an MRGP(Markov regenerative process)-based measurement for the availability of pervasive computing system. First, a bidirectional extended user model is developed, which provides a base point for describing the availability in pervasive computing. Second, an MRGP model for availability is establised, and a stochastic process-based analysis is conducted for the availability of pervasive computing system. The transient and steady availability analysis are provided for this model, and a novel availability metric, called request availability of pervasive computing system, is suggested. Finally, numerical analysis has been carried out and the results argue that the proposed approach and availability metric can give reasonable evaluation for pervasive computing system.

Key words: user model, pervasive computing, MRGP, availability

摘要： 普适计算模式追求随时随地的计算境界，对相关可用性管理提出了很高的要求。现有研究大多基于定性的角度研究普适计算系统的可用性问题，本文尝试给出定量的可用性度量和分析。普适计算系统可用性具有典型的面向用户特点，为刻画用户行为需要引入具有非指数分布性质的用户状态，并且需要考虑用户态与系统态关联分析时的复杂情况。传统的连续时间马尔科夫链和半马尔科夫过程都不能很好的对以上情况给与分析。鉴于此，本文提出了一种基于马尔科夫重生过程（MRGP）的普适计算系统可用性度量方法。首先建立具有双扩展特点的用户模型，基于此构建了普适计算系统可用性度量的MRGP模型，对可用性进行了随机过程分析，给出了对应瞬态和稳态行为的度量方法，结合普适计算模式的特点定义了普适计算系统请求可用性度量。最后进行了数值分析，评价了用户和系统参数等因素对普适计算系统可用性度量的影响。

关键词: 用户模型, 普适计算, MRGP, 可用性度量