Incorporating Fault Tolerance in Distributed Agent Based Systems by Simulating Bio-computing Model of Stress Pathways

Bio-computing model of ‘Distributed Multiple Intelligent Agents Systems’ (BDMIAS) models agents as genes, a cooperating group of agents as operons ─ commonly regulated groups of genes, and the complex task as a set of interacting pathways such that the pathways involve multiple cooperating operons. The agents (or groups of agents) interact with each other using message passing and pattern based bindings that may reconfigure agent’s function temporarily. In this paper, a technique has been described for incorporating fault tolerance in BDMIAS. The scheme is based upon simulating BDMIAS, exploiting the modeling of biological stress pathways, integration of fault avoidance, and distributed fault recovery of the crashed agents. Stress pathways are latent pathways in biological system that gets triggered very quickly, regulate the complex biological system by temporarily regulating or inactivating the undesirable pathways, and are essential to avoid catastrophic failures. Pattern based interaction between messages and agents allow multiple agents to react concurrently in response to single condition change represented by a message broadcast. The fault avoidance exploits the integration of the intelligent processing rate control using message based loop feedback and temporary reconfiguration that alters the data flow between functional modules within an agent, and may alter. The fault recovery exploits the concept of semi passive shadow agents – one on the local machine and other on the remote machine, dynamic polling of machines, logically time stamped messages to avoid message losses, and distributed archiving of volatile part of agent state on distributed machines. Various algorithms have been described.