Benchmarking Fault Tolerance in Hadoop MapReduce with Enhanced Data Replication
Rakesh Recharla · 2025
In distributed file systems (DFS), ensuring fault tolerance is critical for maintaining system robustness and reliability. This research focuses on evaluating fault-tolerant mechanisms in the Hadoop MapReduce framework, specifically examining the impact of increasing data replication on system performance. We conducted a comparative analysis of fault-tolerance improvements by modifying the replication factor from 2 to 3 within the Hadoop MapReduce framework, while maintaining the original system’s performance metrics. The study utilized the TestDFSIO benchmark to assess performance across three key measurements: duration, throughput, and average I/O rate. Our findings indicate that increasing the replication factor enhances fault tolerance without significantly affecting the overall performance, demonstrating that the Hadoop MapReduce framework can maintain high reliability and efficiency with higher data replication levels. This research contributes to understanding the trade-offs between fault tolerance and performance in distributed data processing environments.