PROCESSOR-BASED PERIODIC TESTING AND REPAIR IN DISTRIBUTED SYSTEMS USING SOFTWARE AGENTS Liviu Miclea * , SzilÆrd Enyedi * , Gavril Toderean ** , Dimitris Gizopoulos *** , Yiorgos Xenoulis *** *,** Technical University of Cluj-Napoca, Romania * Department of Automation ** Department of Telecommunications Baritiu Str., 26-28, 400027, Cluj-Napoca, Romania Tel: +40-264-401427, Fax: +40-264-594469 Email: {Liviu.Miclea, Szilard.Enyedi}@aut.utcluj.ro Gavril.Toderean@com.utcluj.ro *** University of Piraeus Department of Informatics 80 Karaoli & Dimitriou Street, 18534 Piraeus, Greece Tel: +30 10 4142372 Fax: +30 10 4142328 Email: {dgizop, gxen}@unipi.gr) Abstract: This paper proposes a distributed use of software-based self-testing, where intelligent agents are responsible for the transfer of software routines to the distributed processors, which in turn will be able to execute the routines and test/repair the corresponding subsystem. This distributed strategy is flexible, re-usable and re- programmable. Keywords: Intelligent agent, distributed BIST and BISR, processor testing. 1. INTRODUCTION Plain BIST and BISR are not well suited for the testing, diagnosis and repair of heterogeneous, distributed and geographically scattered systems, such as nationwide telecommunications or energy distribution systems. A simplified view of such a distributed system is presented in figure 1. Decentralization of test and repair greatly reduces the communicational overhead and increases the flexibility and reliability of the testing system itself. The multiagent approach is only natural to such a problem, as multiagent societies are naturally heterogeneous, decentralized and distributed. An agent is, as implemented here, a piece of software capable of independent existence within an environment provided for it, which is able to communicate with entities similar to it, to unaidedly accomplish the work assigned to it and also to travel between geographically separated locations in its environment. The agents communication capabilities and mobility lead to the concept of multiagent society, which is here a distributed collection of interacting, mobile agents, residing in different parts of the multiagent