Specification and Verification of a Dynamic Reconfiguration Protocol for Agent-based Applications

Abstract

INTRODUCTIONAs computing resources become decentralized, the development of distributed applications receives increasing attention from the software engineering community. These applications are often complex and must satisfy strong reliability and availability constraints. To avoid stopping an entire distributed application for maintenance operations (e.g., repair, upgrade, etc.), it is essential to provide mechanisms allowing distributed applications to be reconfigured at run-time. Such mechanisms should ensure a proper functioning of the application regardless of run-time changes (e.g., creation or deletion of agents, replacement of agents, migration of agents across execution sites, modification of communication routes, etc). Moreover, these mechanisms should not induce heavy penalties on applications during maintenance operations. * This author was partially supported by CONACYT-SFERE and UAM Iztapalapa, Mexico. AbstractDynamic reconfiguration increases the availability of distributed applications by allowing them to evolve at run-time. This paper deals with the formal specification and model-checking verification of a dynamic reconfiguration protocol used in industrial agent-based applications. Starting from a reference implementation in JAVA, we produced a specification of the protocol using the Formal Description Technique LOTOS. We also specified a set of temporal logic formulas characterizing the correct behaviour of each protocol primitive. Finally, we studied various finite state configurations of the protocol, on which we verified these requirements using the CADP protocol engineering tool set.

Using Temporal Logic for Dynamic Reconfigurations of Components

“…Related work Dynamic reconfiguration of distributed applications is an active research topic [1,2,7,18] motivated by practical applications like those modelled in Fractal [10]. In the context of dynamic reconfigurations, ArchJava [3] gives means to reconfigure Java architectures, and the ArchJava language guarantees communication integrity at run-time.…”

An experience report on the verification of autonomic protocols in the cloud

“…In [25,28], the authors show how to formally analyze behavioral models of components using LTSA (Labeled Transition System Analyzer). Another related work is [13], where the authors verify some temporal properties using model checking techniques on a dynamic reconfiguration protocol used in agent-based applications. The main difference here is that we focus on autonomic protocols, which makes their development and specification/verification much more complicated.…”

A Generic Formal Framework for Constructing Agent Interaction Protocols

“…• Language Lotos [31,11,9]. All agents are modelled as Lotos processes which are synchronized on some gates and execute concurrently.…”

“…We provide here a generic formal computational framework that can be reused to generate multiple AIP specifications across many domains. This framework is based on the formal specification language Lotos [7] which is an ideal choice for the description, validation and verification of AIP [31,11,9] considering the following facts:…”