Enhancing Storage Efficiency in Large-Scale Internet of Things Networks: A Longitudinal Study of Least Recently Used Cache Replacement

Abstract

The continuous generation of telemetry data by Internet of Things deployments poses critical challenges for edge devices characterized by limited storage and constrained processing power. Frequent access to primary storage media increases latency and bandwidth utilization. While sophisticated predictive models exist, lightweight, low-overhead solutions are prioritized in practical, medium-scale IoT edge environments. This paper evaluates the efficacy of the Least Recently Used cache replacement strategy, serving as a foundational baseline within a distributed IoT context. Empirical validation is essential, as longitudinal performance in real-world distributed settings often diverges from idealized simulations. This investigation, involving 20 environmental sensors linked to five Raspberry Pi edge servers over three months, assessed caching efficacy using request-hit and data-volume-hit ratios. The LRU configuration achieved an average request-hit ratio of 82.73% and a data-volume-hit ratio of 90.90%. These results confirm that lightweight local caching provides a robust, practical storage-efficiency mechanism for medium-scale IoT systems, avoiding the overhead of complex synchronization or computationally intensive models. The findings underscore the utility of simple replacement heuristics in memory-constrained environments requiring minimized computational overhead.