Since the proposal for the six object-oriented metrics by Chidamber and Kemerer (1994), several studies have been conducted to validate their metrics and have discovered some deficiencies. Consequently, many new metrics for object-oriented systems have been proposed. Among the various measurements of objectoriented characteristics, we focus on the metrics of class inheritance hierarchies in design and maintenance. As such, we propose two simple and heuristic metrics for the class inheritance hierarchy for the maintenance of object-oriented software.In this paper we investigate the work of Chidamber and Kemerer (1994) and Li (1998), and extend their work to apply specifically to the maintenance of a class inheritance hierarchy. In doing so, we suggest new metrics for understandability and modifiability of a class inheritance hierarchy. The main contribution here includes the various comparisons that we have made. We discuss the advantages over Chidamber and Kemerer's (1994) metrics and Henderson-Sellers's (1996) metrics in the context of maintaining class inheritance hierarchies. Figure 6(a): (1 + 1 + 2 + 2 + 4)/5 = 2 AM of Figure 6(a): 2 + (2/2 + 2/2 + 1/2)/5 = 2.5 AU of Figure 6(b): (1 + 2 + 2 + 2)/4 = 1.75 AM of Figure 6(b): 1.75 + (3/2)/4 = 2.13 The understandability and modifiability of Figure 6(b) is better than Figure 6(a) in our metrics. AU of

No abstract

Abstract. The use of formal model based (FMB) methods to evaluate the quality of components is an important research area. Except for a growing number of exceptions, FMB methods are still not really used in practice. This chapter presents two case studies that illustrate the value of FMB approaches for developing and evaluating componentbased software. In the first study, Zed (or Z) and Statecharts are used to evaluate (a priori) the software requirement specification of a Guidance Control System for completeness, consistency and fault-tolerance. The second study evaluates (post-priori) the reliability of a complex vehicle system using Stochastic Activity Networks (SANs). The FMB approach presented here provides further evidence that such methods can indeed be useful by showing how these two different industrial strength systems were assessed and the results. Clearly, future investigations of this nature will help to convince software system developers using component based approaches that such FMB methods should be considered as a valuable tool toward improving the software product lifecycle (quality, schedule and cost).

No abstract

Very faithful but intractable Faithful but relatively intractable Relatively faithful and tractable Unfaithful but very tractable System Under Study (Proposed / Exisitng) REAL SYSTEMChallenges: (1) large state space, (2) model stiffness (relative order of magnitude among parameters) and, (3) the memoryless property assumes that events are independent and identically distributed.A model is always a compromise between faithfulness and simplicity. Results should be precise and correct (model predictions should be realistic and accurate based on a well founded mathematical theory, such as reliability theory). Model 1 Model 2 Model 3 Model 4 Fig. 1: Precise realistic models. ABSTRACTThe increasingly ubiquitous use of embedded systems to manage and control our technologically (ever-increasing) complex lives makes us more vulnerable than ever before. Knowing how reliable such systems are is absolutely necessary especially for safety, mission and infrastructure critical applications. This paper presents a structured compositional modeling method for assessing reliability based on characteristic data and stochastic models. We illustrate this using a classic embedded control system (sensor-inputs | processing | actuator-outputs), Anti-lock Braking System (ABS) and empirical data. Special emphasis is laid on modeling extra-functional characteristics of severity of failures, coincident failures and usage-profiles with the goal of developing a modeling strategy that is realistic, generic and extensible. The validation approach compares the results from the two separate models. The results are comparable and indicate the effect of coincident failures, failure severity and usage-profiles is predictable.

No abstract