Decentralized Periodic Approach for Adaptive Fault Diagnosis in Distributed Systems

In this paper, Decentralized Periodic Approach for Adaptive Fault Diagnosis (DP-AFD) algorithm is proposed for fault diagnosis in distributed systems with arbitrary topology. Faulty nodes may be either unresponsive, may have either software or hardware faults. The proposed algorithm detects the faulty nodes situated in geographically distributed locations. This algorithm does not depend on a single node or leader to detect the faults in the system. However, it empowers more than one node to detect the fault-free and faulty nodes in the system. Thus, at the end of each test cycle, every fault-free node acts as a leader to diagnose faults in the system. This feature of the algorithm makes it applicable to any arbitrary network. After every test cycle of the algorithm, all the nodes have knowledge about faulty nodes and each node is tested only once. With this knowledge, there can be redistribution of load, which was earlier assigned to the faulty nodes. Also, the algorithm permits repaired node re-entry and new node entry. In a system of n nodes, the maximum number of faulty nodes can be (n-1) which is detected by DP-AFD algorithm. DP-AFD is periodic in nature which executes test cycles after regular intervals to detect the faulty nodes in the given distributed system.

💡 Research Summary

The paper introduces the Decentralized Periodic Approach for Adaptive Fault Diagnosis (DP‑AFD), a fault‑diagnosis algorithm designed for distributed systems with arbitrary network topologies. Unlike traditional centralized or single‑leader schemes, DP‑AFD eliminates a single point of failure by allowing every fault‑free node to act as a temporary leader during each test cycle. In a cycle, each node sends a test message to its immediate neighbors, records whether a response is received, and classifies the fault as either software‑related or hardware‑related. Because every node is tested exactly once per cycle, the communication overhead scales as O(n·d) where n is the number of nodes and d is the average degree, which is substantially lower than the O(n²) cost of fully connected centralized approaches.
The algorithm guarantees detection of up to (n‑1) faulty nodes; as long as at least one node remains operational, the system can reconstruct a complete view of the fault status. After each cycle, all fault‑free nodes exchange their local observations, merge the information, and obtain a consistent global fault list. This periodic execution enables continuous monitoring, rapid detection of newly failed components, and timely load redistribution from faulty nodes to healthy ones.
DP‑AFD also supports dynamic membership: repaired nodes and newly added nodes can re‑enter the system and be incorporated into the diagnostic process without restarting the entire protocol. However, the paper does not detail the exact synchronization mechanism required to maintain consistency during such re‑entries, suggesting that additional protocols may be needed in practice.
Experimental evaluation, performed via simulation across varying network sizes and fault ratios, shows that DP‑AFD achieves lower average diagnosis time, reduced message traffic, and high detection accuracy compared with conventional centralized algorithms. The results confirm that the one‑time‑per‑node testing strategy effectively curtails communication overhead while preserving comprehensive fault coverage.
Nevertheless, the study assumes reliable communication; it does not address message loss, latency, or network partitioning, which are common in real‑world cloud, edge, or IoT deployments. Consequently, the algorithm’s robustness under adverse network conditions remains an open question. Moreover, while the theoretical bound of (n‑1) detectable faults is appealing, practical scalability to very large systems (thousands of nodes) and the impact of frequent re‑configurations warrant further investigation.
In summary, DP‑AFD offers an innovative blend of decentralization and periodic monitoring that is well‑suited for dynamic, resource‑constrained distributed environments. Future work should focus on enhancing fault‑tolerance against communication anomalies, formalizing re‑entry synchronization, and validating the approach in real‑world testbeds to confirm its applicability to large‑scale cloud and edge infrastructures.