Traffic-aware Stress Testing of Distributed Real-Time Systems Based on UML Models in the Presence of Time Uncertainty
Traffic-aware Stress Testing of Distributed Real-Time Systems Based on UML Models in the Presence of Time Uncertainty
复制标题
DOI：
10.1109/icst.2008.7
复制
发表时间：
2008-04
期刊：
影响因子：
--
通讯作者：
V. Garousi
中科院分区：
文献类型：
--
作者：
V. Garousi
在以前的工作中，我们报告和实验的压力测试方法来检测网络流量相关的实时（RT）故障的分布式实时系统（DRTS）的设计UML模型的基础上。压力方法，称为时移压力测试方法（TSSTM），旨在增加发现RT故障的机会，起源于网络流量过载的DRTS。TSSTM使用了被测系统（SUT）的UML 2.0模型，增加了时间信息，并基于对UML序列图中控制流的分析。为了设计确定性的测试要求（从时间的角度来看），在SUT的网络流量方面产生最大的压力测试方案，TSSTM方法要求，时序图中的消息的定时信息是可用的，并尽可能精确。然而，在现实中，消息的定时信息并不总是可用的和精确的。正如我们在这项工作中使用的案例研究表明，TSSTM产生的压力测试用例的有效性是非常敏感的时间不确定性。换句话说，TSSTM可能会产生不精确的，不一定是最大的压力测试情况下，存在这样的时间不确定性，因此，它可能不是很有效地揭示RT故障。为了解决上述限制TSSTM，我们在这篇文章中提出了一个修改后的测试方法，可以用来压力测试系统时，消息的定时信息是不精确的或不可预测的。将新测试方法应用于原型DRTS的压力测试结果表明，在消息的定时信息存在不确定性的情况下，与我们以前的方法相比，新方法在检测RT故障方面更有效（即，TSSTM）以及基于操作概要的测试用例。
In a previous work, we reported and experimented with a stress testing methodology to detect network traffic- related real-time (RT) faults in distributed real-time systems (DRTSs) based on the design UML models. The stress methodology, referred to as time-shifting stress test methodology (TSSTM), aimed at increasing chances of discovering RT faults originating from network traffic overloads in DRTSs. The TSSTM uses the UML 2.0 model of a system under test (SUT), augmented with timing information, and is based on an analysis of the control flow in UML sequence diagrams. In order to devise deterministic test requirements (from time point of view) that yield the maximum stress test scenario in terms of network traffic in a SUT, the TSSTM methodology requires that the timing information of messages in sequence diagrams is available and as precise as possible. In reality, however, the timing information of messages is not always available and precise. As we demonstrate using a case study in this work, the effectiveness of the stress test cases generated by TSSTM is very sensitive to such time uncertainty. In other words, TSSTM might generate imprecise and not necessarily maximum stressing test cases in the presence of such time uncertainty and, thus, it might not be very effective in revealing RT faults. To address the above limitation of TSSTM, we present in this article a modified testing methodology which can be used to stress test systems when the timing information of messages is imprecise or unpredictable. The stress test results of applying the new test methodology to a prototype DRTS indicate that, in the presence of uncertainty in timing information of messages, the new methodology is more effective in detecting RT faults when compared to our previous methodology (i.e., TSSTM) and also test cases based on an operational profile.