FMEA and BBN for robustness analysis in web-based applications I. Canova Calori & T. Stålhane Dep. of Computer and Information Science, NTNU, 7491 Trondheim, Norway ABSTRACT: In this paper we present a general framework for conducting robustness analysis early in the development life cycle of web-based systems. This framework exploits the robustness failure modes and evaluates the impact of modifications that can be applied to reduce the severity of these failures. First, the system is systematically decomposed in its components using Jacobson’s analysis. Next, with Failure Modes and Effects Analysis (FMEA) we identify all failure modes, causes and effects. Finally, by using Bayesian Belief Networks (BBNs) we model each subsystem and evaluate failure severities and possible improve- ments. A more complex model of the system can also be built up by integrating the subsystem models, giving a better understanding of the overall system behavior. We present a practical example and we discuss the benefits of applying this framework and BBN models to analyze the robustness of web-based systems. 1 INTRODUCTION BBN has been applied successfully in many do- mains and, among them, to model and predict soft- ware reliability (Singh et al. 2001, Yacoub et al. 1999, Bai 2005), even during early stage of the de- velopment process (Smidts et al. 1998), and to ana- lyze other software characteristics (Houmb et al. 2005, Houmb 2005, Gran 2002b). However, there are still few studies dedicated to analyze robustness of web-based systems. Market pressure brings new challenges in terms of quality issues in web-based systems development. In order to improve the system in an inexpensive way it is crucial to be able to analyze qualities early in the development life cycle. Decisions have to be taken dealing with uncertainty. Uncertainties arise because data are incomplete and the outcomes of many ac- tions are not easy to predict, especially at an early stage of the development process. This paper describes a framework to perform an early robustness analysis of web-based systems by combining the methods proposed by (Lee 2001) and (Zhou & Stålhane 2004), using the Jacobson’s analysis, FMEA and BBN. The paper is organized as follows. Section 2 presents the background of the methods we use. In Section 3, we present in more detail the proposed robustness analysis framework. Section 4 describes a case study, while conclusions are given in Section 5. More specifically, we are interested in robustness. Robustness has been defined in several ways. Ac- cording to (Leveson 1995) there should be no ob- servable events that leave the program’s behavior indeterminate. In (Carlson & Doyle 2002) robust- ness means the ability of a system to maintain some desired characteristics despite fluctuations in the be- havior of its components or its environment. In (Gribble 2001), robustness is considered as the abil- ity to continue to operate correctly across a wide range of operational profiles. By robustness, we mean the property of a system or a component that is totally correct in respect to a complete specification, thus its behavior is predictable for all possible opera- tional environments. 2 BACKGROUND 2.1 Jacobson’s analysis In product development, it is useful to capture the intended behavior of the system. Jacobson’s analysis identifies system behavioral aspects already at an early stage of the development cycle when little in- formation about the system structure is available. Ja- cobson’s analysis method is based on use cases. By analyzing them we can identify the objects that play an important role in a use case and classified them into one of the categories depicted in Table 1. Due to the complexity of web-based applications and their development process, robustness analysis needs to cope with an incremental process and in- formal communication (Ziemer & Stålhane 2006), and thus be adaptable to increasing knowledge and frequent updates as well as provide a reliable and in- tuitive tool to the development team. In fact, as it has been pointed out in (Langseth & Portinale 2005), BBNs can be built up by smaller pieces and repre- sent a mathematically sound formalism, that allows rigorous reasoning, while still being easy to under- stand for the decision maker. According to (Zhou & Stålhane 2004), Jacob- son’s analysis provides a systematic method for de- composing a system, that is suitable to our case too. Refer to (Jacobson 1993, Rosenberg & Scott 2001) for further information.