Gossip-Based Distributed Monitoring System

This project presents the design and implementation of a fully decentralized monitoring system based on gossip (epidemic) protocols. The system consists of multiple independent nodes that collaboratively maintain cluster membership and detect failures without relying on any centralized coordinator. Each node periodically exchanges partial state information with a subset of randomly selected peers. Through these probabilistic interactions, membership updates propagate across the system, enabling nodes to converge toward a consistent global view even in the presence of network delays, packet loss, and node crashes. Failure detection is achieved through heartbeat mechanisms combined with timeout-based state transitions (ALIVE, SUSPECT, DEAD),following a design inspired by the SWIM protocol. The system is designed to operate in asynchronous and unreliable environments while ensuring eventual consistency. An experimental extension, called Gossip Arena, is introduced to study information dissemination dynamics. Synthetic rumor events are injected into the system to evaluate the impact of different gossip strategies and parameters on dissemination speed and communication overhead. Experimental results show that the system remains robust under moderate network impairments and highlight the trade-offs between convergence speed, communication cost, and accuracy of failure detection.