Stimulating proactive fault detection in distributed systems
Namita Naik · RIT Scholar Works (Rochester Institute of Technology) · 2007
Fault Tolerance is an important issue considered when developing a reliable Distributed System. Reactive fault systems are designed to redistribute the current process on to other machines when failure occurs. In contrast to the conventional method of reactive recovery, an emerging concept in the field of fault tolerance is a proactive approach. This approach exploits pre fault symptoms and initiates fault recovery henceforth. This project is to implement a proactive fault prediction simulator for a distributed system. This will include developing a language for simulation, which allows the user to define a distributed system. The language is further used to develop an environment that integrates two fault prediction algorithms, Wilcoxon s Rank-Sum and DFT (Dispersion Frame Technique). Both these algorithms are presented as alternatives for SMART (Self Monitoring Analysis and Reporting Technology) in this project. The project also includes a comparison metrics for these prediction algorithms in terms of prediction precision and prediction accuracy.