| Elaborating a correct design of a concurrent system is extremely difficult. In part this is due to the infinite number of possible system runtime behaviors that result from the concurrent (or pseudo-concurrent) execution of interacting processes or threads. Data consistency, deadlock, starvation and fairness issues are the most well-known problems encountered in concurrent systems. Accurate concurrent system design verification approaches require an expensive system runtime behavior analysis and consequently result in prohibitively high development costs (i.e. for testing).;In the second part of the thesis we show how a concurrency-enriched specification can be transformed into a system application-level consistent design. The approach starts with concurrency-aware OCL-based operation schemas that describe all system functionality using pre-, rely-, and post-conditions. These schemas are then mapped to Rely diagrams. Based on the rely diagrams, sequence diagrams describing the concurrent system design are elaborated. The approach uses locks to ensure consistency and deadlock freedom. We then further show how these locks can be used to enforce certain fairness policies.;The usefulness of our approach is demonstrated by applying it to the design of a non-trivial case study: an online auction system. We first illustrate an original online auction system design that is built according to user requirements elicitation and analysis using the object-oriented development process Fondue; then we check the original design for deadlocks using our first approach. We subsequently modify the original design based on the results.;In this thesis we try to address this problem by presenting several static approaches that can help the developer of a concurrent system during the design phase. In the first part of the thesis we present an approach that can analyze an existing concurrent system design to detect potential deadlock situations. This is done by mapping object interaction diagrams such as sequence diagrams to System Synchronization Hasse diagrams, which are then analyzed to detect deadlock cycles. Since the approach is static, it is overly pessimistic, meaning that it is possible that the algorithm detects a deadlock that, in reality, cannot occur. On the other hand, if the algorithm cannot detect any deadlocks, the developer can be sure that the design is deadlock-free. |
