Pavel Moravec

DiVinE: a tool for distributed verification

We present a tool for cluster-based LTL model-checking and reachability analysis. The tool incorporates several novel distributed-memory algorithms and provides a unique interface to use them. We describe the basic structure of the tool, discuss the main architecture decisions made, and briefly explain how the tool can be used.

Accepting Predecessors Are Better than Back Edges in Distributed LTL Model-Checking

We present a new distributed-memory algorithm for enumerative LTL model-checking that is designed to be run on a cluster of workstations communicating via MPI. The detection of accepting cycles is based on computing maximal accepting predecessors and the subsequent decomposition of the graph into independent predecessor subgraphs induced by maximal accepting predecessors. Several optimizations of the basic algorithm are presented and the influence of the ordering on the algorithm performance is discussed. Experimental implementation of the algorithm shows promising results.

Distributed Partial Order Reduction of State Spaces

State space explosion is a fundamental obstacle in formal verification of concurrent systems. Several techniques for combating this problem have emerged in the past few years, among which the two we are interested in are: partial order reduction and distributed memory state exploration. While the first one tries to reduce the problem to a smaller one, the other one tries to extend the computational power to solve the same problem. In this paper, we consider a combination of these two approaches and propose a distributed memory algorithm for partial order reduction.

Parallel algorithms for finding SCCs in implicitly given graphs

We examine existing parallel algorithms for detection of strongly connected components and discuss their applicability to the case when the graph to be decomposed is given implicitly. In particular, we list individual techniques that parallel algorithms for SCC detection are assembled from and show how to assemble a new more efficient algorithm for solving the problem. In the paper we also report on a preliminary experimental study we did to evaluate the new algorithm.

Effective verification of systems with a dynamic number of components

In the paper, we present a novel approach to verification of dynamic component-based systems, the systems that can have a changing number of components over their life-time. We focus our attention on systems with a stable part (called provider) and a number of dynamic components of one type (called clients) because dynamic systems can be often decomposed into segments like this. Our method for verification of such systems is based on determining a number k of dynamic components, such that if a system is proved correct for any number lower than k, it is consequently correct for an arbitrarily large number of dynamic components. The paper aims not only in proving the propositions that state this, it concentrates also on bounding the set of dynamic systems and verifiable properties in a way, that k is relatively small and thus practically interesting. In addition to this, we present an algorithm for computing k.

Complementarity of Error Detection Techniques

We study explicit techniques for detection of safety errors, e.g., depth-first search, directed search, random walk, and bitstate hashing. We argue that it is not important to find the best technique, but to find a set of complementary techniques. To this end, we choose nine diverse error detection techniques and perform experiments over a large set of models. We compare speed of techniques, lengths of reported counterexamples, and also achieved model coverage. The results show that the studied set of techniques is indeed complementary in several ways.

Formal verification of systems with an unlimited number of components

In many real component-based systems and patterns of component interaction, there can be identified a stable part (such as control component, server, instance handler) and a number of uniform components of the same type (users, clients, instances). Such systems, the so-called control-user systems, are often modelled using an infinite set of finite models of particular components, parameterised by the number of uniform components in the system. However, if the maximal number of components is not known, this results in infinite-state models, which cannot be directly verified with effective (finite-state) techniques, like model checking. In this case, more involved techniques have to be employed. A verification technique for checking linear temporal logic (LTL)-like interaction properties on control-user systems with unlimited number of components using finite-state verification is proposed. The method is based on computing a cutoff on the number of uniform components (users), such that if the system is proved to be correct for every number of user components up to the cutoff, it is guaranteed to be correct for any larger number of components. The authors define the cutoff, prove that it guarantees the required property, introduce heuristics for computing the cutoff and demonstrate the overall technique on a number of previously published models.

On combining partial order reduction with fairness assumptions

We present a new approach to combine partial order reduction with fairness in the context of LTL model checking. For this purpose, we define several behaviour classes representing typical fairness assumptions and examine how various reduction techniques affect these classes. In particular, we consider both reductions preserving all behaviours and reductions preserving only some behaviours.