Autonomic mobile agent-based parallel computing for distributed systems

Parallel computing is the simultaneous use of multiple computing resources to solve a computational problem. Parallel computing can be performed in shared and distributed memory environments. Because computers connected through local area networks (LANs) are less expensive and highly scalable, there is a growing interest in using distributed networked computers for parallel computing. Message Passing Interface (MPI) is a de facto industry standard that defines various functions for developing high-performance portable parallel programs in a distributed memory-based computing environment.;The conventional procedure to start an MPI parallel computation among distributed computers lacks flexibility because a user is involved in every step of the procedure and the parallel computation can only be performed with precompiled platform-dependent binary executables. The lack of flexibility does not give a conventional MPI parallel computing environment sufficient autonomicity to deal with unanticipated situations during runtime. Mobile agents can be dynamically created at runtime and migrate to destination machines to fulfill operations on the destination machines in an autonomous manner to handle unforeseen events. The autonomy and mobility of mobile agents make mobile agent paradigm a proper approach for developing an autonomic MPI-based parallel computing system.;This research is the first attempt to integrate the MPI model with the IEEE Foundation for Intelligent Physical Agents (FIPA) standard mobile agent technology for distributed parallel computing. The developed autonomic mobile agent-based parallel computing system is compliant with the MPI and C standards for parallel computations and the IEEE FIPA standard for agent interoperability. The system can achieve rapid deployment and ii execution of mobile agents over heterogeneous platforms for MPI parallel computing, and self-manage an MPI computation with respect to self-configuration, self-optimization, self-healing, and self-protection for the ongoing computation.;Autonomic properties of the developed system are validated through different examples. Different applications associated with the system including resource constrained mobile robot-based parallel image processing for target tracking, dynamic runtime alteration of behavioral algorithms for a mobile robot, and Web-based parallel computing over heterogeneous platforms for rapid prototyping and teaching are presented to demonstrate the developed system has a great potential to be applied to a variety of distributed parallel applications.