Showing papers by "Steffen Becker published in 2004"

Performance prediction of component-based systems

Abstract: Performance predictions of component assemblies and the ability of obtaining system-level performance properties from these predictions are a crucial success factor when building trustworthy component-based systems. In order to achieve this goal, a collection of methods and tools to capture and analyze the performance of software systems has been developed. These methods and tools aim at helping software engineers by providing them with the capability to understand design trade-offs, optimize their design by identifying performance inhibitors, or predict a system's performance within a specified deployment environment. In this paper, we analyze the applicability of various performance prediction methods for the development of component-based systems and contrast their inherent strengths and weaknesses in different engineering problem scenarios. In so doing, we establish a basis to select an appropriate prediction method and to provide recommendations for future research activities, which could significantly improve the performance prediction of component-based systems.

Towards an engineering approach to component adaptation

Abstract: Component adaptation needs to be taken into account when developing trustworthy systems, where the properties of component assemblies have to be reliably obtained from the properties of its constituent components. Thus, a more systematic approach to component adaptation is required when building trustworthy systems. In this paper, we illustrate how (design and architectural) patterns can be used to achieve component adaptation and thus serve as the basis for such an approach. The paper proposes an adaptation model which is built upon a classification of component mismatches, and identifies a number of patterns to be used for eliminating them. We conclude by outlining an engineering approach to component adaptation that relies on the use of patterns and provides additional support for the development of trustworthy component-based systems.

Classifying Software Component Interoperability Errors to Support Component Adaption

Abstract: This paper discusses various classifications of component interoperability errors. These classifications aim at supporting the automation of component adaptation. The use of software components will only demonstrate beneficial, if the costs for component deployment (i.e., acquisition and composition) are considerably lower than those for custom component development. One of the main reasons for the moderate progress in component-based software engineering are the high costs for component deployment. These costs are mainly caused by adapting components to bridge interoperability errors between unfitting components. One way to lower the costs of component deployment is to support component adaptation by tools, i.e., for interoperability checks of (semi-)automated adaptor generation. This automation of component adaptation requires a deep understanding of component interoperability errors. In particular, one has to differentiate between different classes of interoperability errors, as different errors require different adaptors for resolving. Therefore, the presented classification of component interoperability errors supports the automation of component adaptation by aiding automated interoperability problem detection and semi-automated adaptor generation. The experience gained from already implemented solutions for a specific class of interoperability errors provides hints for the solution of similar problems of the same class.

Component Composition with Parametric Contracts

Abstract: We discuss compositionality in terms of (a) component interoperability and contractual use of components, (b) component adaptation and (c) prediction of properties of composite components. In particular, we present parametric component contracts as a framework treating the above mentioned facets of compositionality in a unified way. Parametric contracts compute component interfaces in dependency of context properties, such as available external services or the profile how the component will be used by its clients. Under well-specified conditions, parametric contracts yield interfaces offering interoperability to the component context (as they are component-specifically generated). Therefore, parametric contracts can be considered as adaptation mechanism, adapting a components providesor requires-interface depending on connected components. If non-functional properties are specified in a component provides interface, parametric contracts compute these nonfunctional properties in dependency of the environment.

Towards a Generic Framework for Evaluating Component-Based Software Architectures

Abstract: The evaluation of software architectures is crucial to ensure that the design of software systems meets the requirements We present a generic methodical framework that enables the evaluation of component-based software architectures It allows to determine system characteristics on the basis of the characteristics of its constituent components Basic prerequisites are discussed and an overview of different architectural views is given, which can be utilised for the evaluation process On this basis, we outline the general process of evaluating software architectures and provide a taxonomy of existing evaluation methods To illustrate the evaluation of software architectures in practice, we present some of the methods in detail