Run-time monitoring provides an effective means for aiding in the difficult tasks of developing, testing and maintaining distributed computations. However, monitoring activities introduce intrusion, which can potentially alter the behavior of the computation being monitored. This project is investigating techniques to accommodate the effects of monitoring intrusion such that a computation will exhibit the same behavior while being monitored as it would have exhibited in the absence of monitoring. This project focuses on a variety of approaches, which encompass processor scheduling, communication protocols, time maintenance and process synchronization. The techniques developed will be incorporated into a system for the on- line monitoring of distributed computations, for experimentation and evaluation. This project is conducted jointly by Madalene Spezialetti of Trinity College, CT and Rajiv Gupta of University of Pittsburgh under grant CCR 9704350.
