Enhanced Cache Replacement Strategy for Named Data Networking: A Modified LeCaR Algorithm with Eviction History Tracking
Alesandora Emanuella Pinto, Sendy Saputra, Ridha Muldina Negara, Rohmat Tulloh, Nana Rachmana Syambas, Naufal Hanan Lutfianto · 2025
Efficient content retrieval in Named Data Networking (NDN) relies on optimized caching mechanisms. Traditional cache replacement policies, such as Least Recently Used (LRU) and Least Frequently Used (LFU), struggle to adapt to dynamic workloads. The Learning Cache Replacement (LeCaR) algorithm was introduced to address these challenges using regret minimization between LRU and LFU. However, LeCaR suffers from overfitting in specific workloads and lacks a truly adaptive reinforcement learning framework. This study proposes a Modified LeCaR Algorithm, incorporating eviction history tracking to enhance cache replacement decisions. The proposed approach is evaluated using three datasets: Synthetic, FIU, and Medisyn, representing diverse workload conditions. Experimental results demonstrate that the modified LeCaR consistently outperforms traditional caching policies in small cache configurations while maintaining competitive performance against LRU in larger cache scenarios. These findings highlight the importance of adaptive caching strategies in improving NDN efficiency, particularly in dynamic network environments.