Mechanisms for reliability and crash recovery in distributed computing systems are being studied, particularly in distributed database systems. Tools are being developed for evaluating and comparing the solutions; and using these and other tools, solutions for several specific problems are classified and categorized. The research emphasizes algorithm implementation and experimental performance measurement on the new experimental network consisting of five computers with over one gigabyte of disk storage. The goal is to investigate novel algorithms for reliable operation, so the systems implemented are used to drive the research. That is, new algorithms are implemented and evaluated in order to understand their strengths and practical limitations. This, in turn, suggests variations and improvements on the algorithms that will lead in new directions. Even though the efforts center on system implementation, they continue to use simulation and analysis when appropriate.
