Mobile agent frameworks have attracted a lot of attention in recent years, seen as counterparts of static distributed object frameworks but allowing also for object or agent mobility. A lot of research is currently being carried out trying to assess their applicability to network management and control environments. In this paper, we present our experiences of using a mobile agent framework to design and realize a performance management system which exhibits "constrained mobility", constrained in the sense that performance monitoring agents are sent to execute within network elements and stay there until their task is accomplished. We present the architecture, design and implementation of such a system, compare and contrast it to static object approaches and present a detailed performance comparison to a similar Java-RMI based implementation, trying to assess the overhead of mobile agent solutions.

No abstract

No abstract

No abstract

During the recent years of research on mobile agents, significant effort has been directed towards the identification of models of agent mobility suitable for network management applications. Also, a lot of research work is currently being carried out trying to provide an assessment of mobile agent frameworks used to build agent-based network management systems. In this paper we clarify three different models of agent mobility, present a mobile agent-based performance monitoring system that exhibits the "constrained mobility" model, and discuss its practical use for dynamically programming network elements. The implementation of this system is presented and compared with static object approaches. Furthermore we provide a performance evaluation of the mobile agent based system as it compares with Java-RMI and CORBA distributed frameworks, in order to assess the advantages, along with the overheads, of agent solutions.