The verification of concurrent finite-state systems is confronted in practice with the state explosion problem (prohibitive size of the underlying state space), which occurs for realistic systems containing many parallel processes and complex data structures. Various techniques for fighting against state explosion have been proposed, such as on-the-fly verification, partial order reduction, and distributed verification. However, practical experience has shown that none of these techniques alone is always sufficient to handle large-scale systems. In this thesis, we propose a combination of these approaches in order to scale up their capabilities. Our approach is based upon Boolean Equation Systems (BESs), which provide an elegant intermediate representation for verification problems defined on Labeled Transition Systems, such as equivalence checking, tau-confluence reduction, model checking of alternation-free mu-calculus, and test-case generation. We propose DSolve and MB-DSolve, two new algorithms for distributed on-the-fly resolution of BESs (containing one, resp. several equation blocks), and employ them as computing engines for four on-the-fly verification tools developed within the CADP toolbox using the OPEN/CAESAR environment: the BISIMULATOR equivalence checker, the TAU_CONFLUENCE reductor, the EVALUATOR model checker, and the EXTRACTOR test-case generator. Experimental measures performed on clusters of machines show quasi-linear speedups and a good scalability of the distributed versions of these tools w.r.t. their sequential counterparts.
