Managing fault tolerance information in multi-agents based distributed systems

초록

In a fault tolerant system using rollback-recovery protocols, the performance of the system is degraded because of the increment of saved fault tolerance information. To avoid degrading its performance, we propose novel multi-agents based garbage-collection technique that deletes useless fault tolerance information. We define and design a garbage-collection agent for garbage-collection of fault tolerance information, a information agent for management of fault tolerant information, and a facilitator agent for communication between agents. And we propose the garbage-collection algorithm(GCA) using these agents. Our rollback recovery method is based on independent checkpointing protocol and sender based pessimistic message logging protocol. To prove the correctness of the garbage-collection algorithm, we introduce failure injection during operation and compare the domain knowledge of the proposed system using GCA with the domain knowledge of another system without GCA. © Springer-Verlag 2003.