Expert System for Evaluating the Behavior of Concurrent System Prototy

在软件原型开发中，存在一个称为原型开发周期的周期，该周期分为原型构建、执行和评估三个阶段。软件设计人员应该尽可能快地围绕整个周期执行原型活动，以便验证、修改和改进需求和设计。在复杂的并发软件中，视觉检查不足以检查原型的行为。为了提高原型开发的快速性，需要在原型评价阶段为原型行为的检验提供更有效的工具。本研究在诊断型专家系统的基础上，研究了有效的工具。这些工具可以检测瓶颈和死锁，并提供避免其根源的改进计划。死锁是并发软件系统功能的致命影响。瓶颈是影响并发软件系统性能的致命因素。死锁专家系统是基于线性整数规划来获得系统进入死锁状态的激发序列。瓶颈专家系统是基于定性推理的瓶颈改进专家系统。研究了基于专家系统的启发式诊断方法和基于算法的分析诊断方法的互补作用，以及定性诊断和定量诊断的互补作用。

In software prototyping, there exists a cycle named prototyping cycle, which has 3 phases, i. e., prototype building, execution and evaluation phases. Software designers should perform prototyping activity around the cycle as fast as possible in order to validate, modify and refine requirements and design. In complex concurrent software, the visual inspection is not enough to examine the behavior of the prototype. In order to prompt the rapidness of prototyping, more effective tools should be provided at the prototype evaluation phase for the examination of the prototype behavior. This research studies the effective tools on the basis on diagnosis type of expert systems. These tools can detects the bottleneck and the deadlock and provide improvement plans for avoiding their sources. Deadlock is the fatal effect of function of concurrent software systems. Bottleneck is the fatal effect of performance of concurrent software systems. Expert system for deadlock is based on the linear integer programming for obtaining firing sequences by which the system falls into the deadlock state. Expert system for bottleneck is based on the qualitative reasoning for improving bottlenecks. This research also studies the complementary role of heuristic diagnosis method on the basis of expert system and the analytical diagnosis method on the basis of algorithm, and the complementary role of qualitative diagnosis and quantitative diagnosis.