Knowledge-based distributed system architecture
Anna Hác · ACM SIGMETRICS Performance Evaluation Review · 1989
This paper introduces the learning system, the expert system and an information broadcasting protocol for designing and managing distributed systems. A knowledge-based system can be implemented as a part of operating system software to make decisions about process transfer and message routing in a hierarchical network. A knowledge-based system uses dynamic information about the state of processors and applications in the local and wide arca network. This information consists of processors' and applications' queue lengths, and it is broadcast to directly connected processors. The expert system uses broadcast information to make decisions about process transfer and message routing, considering processor availability and system security. The expert system causes processors' queue lengths to become balaced on each network hierachy level. The number of process transfers is calculated and depends on network partitioning and the threshold values used by the expert system. The convergence of the algorithms for the knowledge-based system is proven. Performance of the proposed system is evaluated analytically using the elapsed time of process transfer or message transfer and the waiting time to begin transfer.