Multi‐agent architecture approach for self‐healing systems: Run‐time recovery with case‐based reasoning

Abstract

Self-healing is an approach that maintains the health of the system with proper supervision of its functioning and emerges from any unacceptable state during the execution. The complexity of modern distributed systems and dynamic hike in terms of users' access has led to an increase in maintenance. In this paper, a self-healing architecture for services that exploit the autonomous capability of agent technology is proposed. The proposed mechanism is a multi-agent system that comprises different agents with different capabilities and roles. The planning agent, responsible for taking the right decision to revive the system from an unhealthy state to a healthy state, uses a case-based fault recovery mechanism at runtime. The architecture contains a persistent layer that maintains previously occurred failed cases. To determine the best suitable solution, the similarity between the detected fault and recorded failed cases is calculated. The case, having a maximum similarity index value is considered closest to the failure. Multiple recovery strategies like a replacement, restart, alternative resources are been utilized. Also, to validate the proposed architecture, an SOA-based application is used and performance-based evaluation metrics are analyzed.