Beiträge zur Anforderungsanalyse im Rahmen der Entwicklung betrieblicher Anwendungssysteme aus Sicht der Wirtschaftsinformatik

Web service compositions coordinate Web services of different enterprises. They are expected to constitute the foundation of service-oriented architectures, to improve business processes as well as to foster intraand inter-organizational integration. Especially in inter-organizational contexts, quality of service referring to non-functional requirements and conformance to functional requirements are becoming vital properties. With Web service compositions being asynchronous and distributed systems, the latter property – which is also called correctness – can be shown best by verification. This paper examines from a system-theoretic perspective how correctness can be operationalized for Web service compositions. It also proposes a requirements framework for service-oriented modeling techniques so that correctness can be shown by verification and Web service compositions can be modeled intuitively. In order to show the framework’s principle applicability, an example approach is analyzed with respect to the corresponding requirements. II Anforderungsanalyse für verteilte Systeme – am Beispiel Web Service-basierter Systeme (Beitrag: „Verification of Web Service compositions: An Operationalization of Correctness and a Framework for Service-oriented Modeling Techniques“) II-2 1 Motivation and object of research Web services and Web service compositions (WS compositions) are currently considered to be the most widespread possibility for implementing service-oriented architectures (Erl 2004; Newcomer and Lomow 2005). A (basic) Web service is a software system that exposes its functionality by means of a machine-processable interface consisting of several operations. It enables (asynchronous) message-based machine-to-machine interaction over a network (Booth et al. 2004). Web services are offered by internal IT departments or by external service providers. A WS composition is a Web service that realizes complex functionality by coordinating multiple (component) Web services in transactions (Alonso et al. 2004, p. 141). Analogous to (basic) Web services, WS compositions expose their functionality via interfaces. In this paper, WS compositions refer to the orchestration of Web services, not to their choreography (Dostal et al. 2005, p. 226). According to a survey of the Yankee Group, WS compositions raise high expectations (Kallus 2004). Among other things, they are expected to improve business processes and to foster the integration of existing e-business, CRM, SCM, and ERP initiatives. Already in 2004, more than 50 % of the US companies relied on Web services whereof 60 % estimated their impact on business-to-business integration (very) high. However, designing and running WS compositions is error-prone. Involved teams of software engineers and modelers usually stem from multiple companies and have different skills, experiences, or functional backgrounds. Web services are usually managed by different companies and may not have been developed for composition. Analogous to other distributed systems based on asynchronous communication, it is difficult to anticipate how WS compositions behave during execution and whether they conform to the functional requirements identified during requirements engineering. Errors, however, may violate service level agreements. This may cause losses or penalties and have negative impact on the reputation of the companies involved. Therefore, it is an important task to make sure that WS compositions conform to their functional requirements. A possibility of reliably showing conformance to functional requirements is verification. In this context, conformance to functional requirements is also called correctness. The II Anforderungsanalyse für verteilte Systeme – am Beispiel Web Service-basierter Systeme (Beitrag: „Verification of Web Service compositions: An Operationalization of Correctness and a Framework for Service-oriented Modeling Techniques“) II-3 idea of verification dates back to Floyd (1967) and Hoare (1969). In contrast to testing, for instance, where correctness can only be shown with respect to selected test data, verification aims at exhaustively proving correctness for all behavioral facets and inputs of a given program. During verification, the program is analyzed automatically. This requires the semantics of each program statement to be defined unambiguously. A program is said to be correct if its implementation (i. e. its program code or a corresponding formal model) is consistent with a corresponding specification of behavioral claims (i. e. functional requirements on the program’s behavior) (Balzert 1998, pp. 445-472). Despite the need for correctness, research focuses on non-functional requirements such as availability, scalability, capacity, etc. (Lee et al. 2003), which are already known from network research. Current approaches to (Web) service-oriented modeling do not (or hardly) cope with correctness and verification. This includes specification languages like the de facto standard WS-BPEL (Alves et al. 2007) – despite some attempts (e. g. Farahbod et al. 2005; Stahl 2005; Moser et al. 2007; Lohmann 2007) – and high-level approaches (Arjansani 2004; Bell 2008; Papazoglou and van den Heuvel 2006; Zimmermann et al. 2004). Current approaches to the verification of Web services (WS verification) require formal models, e. g. in terms of finite state automata or Petri nets, which are not necessarily intuitive for conceptual modelers. Such approaches also postulate a variety of formal claims on WS compositions so that it is not clear what correctness actually means. Against this background, we address the following research questions: How does correctness have to be operationalized so that it fits the peculiarities of WS compositions? How do (Web) service-oriented modeling techniques have to be shaped so that correctness can be shown by verification and WS compositions be modeled intuitively? This paper relies on a design-oriented, deductive, and argumentative research approach (Hevner et al. 2004; Wilde and Hess 2007). In section 2, we identify the research gap by compiling current approaches to WS verification and service-oriented modeling. Section 3 proposes a definition of correctness and a requirements framework as artifacts. Section II Anforderungsanalyse für verteilte Systeme – am Beispiel Web Service-basierter Systeme (Beitrag: „Verification of Web Service compositions: An Operationalization of Correctness and a Framework for Service-oriented Modeling Techniques“) II-4 4 aims at showing the framework’s basic applicability in the sense of a basic evaluation. Section 5 briefly summarizes the findings and points out further research.