Providing Applications with Mobile Agent Technology Paulo Marques, Paulo Simões, Luís Silva, Fernando Boavida, João Silva CISUC, University of Coimbra, Portugal Dep. Eng. Informática, Polo II – U. Coimbra 3030 Coimbra, Portugal {pmarques, psimoes, luis, boavida, jgabriel}@dei.uc.pt Abstract. Over the last couple of years we have been working on the development of mobile agents systems and its application to the areas of telecommunications and network management. This work path produced positive results: a competitive mobile agent platform was built, the run-time benefits of mobile agents were proved, and our industrial partners have developed practical applications that are being integrated into commercial products. However, despite the positive results, we feel that mobile agent technology is still not ready to enter the path of mainstream software development. In our perspective, one of the main reasons for this situation arises from the traditional approach to mobile agent technology. This approach, based on the familiar concept of the mobile- agent distributed platform as an extension of the operating system, focuses too much on the mobile agents and associated issues (mobility, agent lifecycle, security, coordination, etc.) and provides poor support for the development of applications where mobile agents are just one of several available technologies. Learning from past experience, we are now working on a new approach where the focus is brought back to the applications and mobile agents become just one the tools available to develop distributed systems. This provides a much lighter framework for application-based mobile agent systems. This paper presents the lessons learned from our previous project and discusses the new concept we are developing: application-centric mobile agent systems. Keywords: Mobile Agents, Component Technology, Distributed Computing, Network Management I. INTRODUCTION Over the last few years we have observed to the proliferation of available mobile agent (MA) systems, which currently reach the impressive number of seventy-two known platforms [1]. This is a lot more than available RPC [2] or CORBA [3] implementations. Nevertheless, this technology is still far from the mainstream programmer and, oddly as it seems, there are now much more MA platforms than MA-based applications. In the last couple of years we were involved in the JAMES project [4]. This project was held in consortium with Siemens Portugal SA and Siemens AG, and the objective was to develop a new MA platform specifically tuned and customized for the area of telecommunications and network management (NM). Our industrial partners used this platform to produce a few MA-based applications that are now being integrated into commercial products. These applications use mobile agents to perform management tasks (accounting, performance management, system monitoring and detailed user profiling) that deal with very large amounts of data, distributed over the nodes of GSM networks. With this project we have learned that MA- technology, when appropriately used, provides significant competitive advantages to distributed management applications. However, we have also realized that the current status of MA technology is still not appropriate to take those advantages to mainstream application development. In order to overcome some of the limitations found, we are now working on a new approach that abandons the classic concept of MA platforms as extensions of the operating system. Instead, we are focusing on providing agent mobility within application boundaries, rather than within system boundaries. The objectives of this paper are twofold. First, we present a critic perspective on the status of current mobile agent technology. This perspective is directly influenced by the results gathered during the JAMES project but we feel it applies to most of the current mobile agent implementations. In the second part of the paper we describe the approach that we are currently undertaking to address the identified problems.