Empirical Evaluation of Query Router Scalability in Sharded MongoDB

As distributed databases become the backbone of modern large-scale applications, the ability to scale horizontally is paramount. MongoDB, a leading NoSQL database, employs a sharded architecture to distribute data across multiple servers. While the scalability of the data layer (shards) is well-documented, the performance implications of the routing layer remain underexplored. This paper presents an empirical evaluation of router scalability in sharded MongoDB clusters. Using a controlled benchmark environment, router count is isolated as the primary experimental variable across multiple shard configurations and concurrency levels. Our results show that a single router saturates under moderate load, while scaling to a few routers yields substantial gains. However, beyond a threshold, additional routers provide negligible benefit, revealing a clear performance plateau. Our results further indicate that logical contention within single-shard deployments constrains throughput independently of router capacity. These findings offer practical (data-driven) insights useful for sizing sharded MongoDB architectures.