Three Layered Hierarchical Fault Tolerance Protocol for Mobile Agent System

A Mobile Agent (MA) is autonomous and identifiable software process that travel through a network of heterogeneous machine and act autonomously on behalf of user. Improving the survivability of MA in presence of various faults is the major issue concerns with implementation of MA. During its life cycle, a MA can fail due to some uncaught exception, or due to the failure of the MAS, or its components or the host machine. The MA may also be lost on its way or blocked due to link failure. Since failure occurs at different places due to different reasons, specialized approaches should be used to tolerate different kinds of faults. This paper presents a brief introduction of Hierarchical Fault Tolerance Protocol (HFTP) for Mobile Agents. The proposed protocol is hierarchical in nature, which works at three levels. Based on the experience gained from prior work, our approach is to use the concept of fault masking without replication at one level. This ensures that failure is not visible to the MA through grouping of hosts within a network. At another level rear guard based fault detection and recovery based approach has been used. A thread based approach has been used to detect faults at the lowest level. In this way, the protocol tolerates various kinds of faults and takes the advantage of both centralized and distributed approaches. HFTP can tolerate host failure, system failure as well as link failure by grouping the hosts within a network and rear guard based migration of MA in the global network. A well known modeling tool Color Petri Net (CPN) has been used for architectural model of HFTP. Simulation results have been used to check the performance of HFTP in presence of various faults.