Parallel Symbolic Observation Graph

• Published in: 2017 IEEE International Symposium on Parallel and Distributed Processing with Applications and 2017 IEEE International Conference on Ubiquitous Computing and Communications (ISPA/IUCC) pp. 770 - 777
• Main Authors: Ouni, Hiba, Klai, Kais, Abid, Chiheb Ameur, Zouari, Belhassen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-12-2017
• Subjects: Aggregates; Boolean functions; Data structures; distributed memory; Heuristic algorithms; Load management; Model checking; multi threading; parallel model checking; Partitioning algorithms; symbolic reachability
• DOI: 10.1109/ISPA/IUCC.2017.00118

Model checking is a powerful technique for verifying and analyzing complex systems in many application fields. The analysis process of complex and concurrent systems often requires large computation resources which represents a real challenge. Even with simple configurations, the well-known state explosion problem is faced as the generated state space of such systems grows exponentially with the number of the system components. Numerous methods and techniques have been developed to overcome this problem including parallel and distributed-memory processing. In this paper, we aim at improving the performances of the so called Symbolic Observation Graph (SOG) construction by using parallelization techniques. A SOG is a hybrid structure where the transitions of a system are divided into observed and unobserved ones. The nodes of this graph are then defined as sets of states linked with unobserved transitions (and encoded symbolically with a BDD) and edges are labeled with observed transitions only (and are explicitly represented). We propose two parallel algorithms to build the SOG. The first algorithm is dedicated for shared memory architectures, and is based on the distribution of the SOG construction on several threads using a dynamic load balancing scheme. The second algorithm is proposed for distributed memory architectures, and distributes the SOG construction on processes using a static load balancing scheme. These two algorithms are implemented and their performances are studied and compared to each other and to the sequential construction of the SOG.