CSS-LSTM: An Intelligent Caching Strategy in NDN Clusters

In-network caching is a key aspect of Named Data Networking (NDN) due to its significant impact on network performance and data delivery efficiency. As content demand, particularly for multimedia, continues to grow, deciding what to cache, where to store it, and when to remove it has become increasingly complex. This complexity highlights the need for more intelligent caching strategies in NDN. To address this challenge, we propose an intelligent caching strategy for NDN clusters called CSS-LSTM. This approach introduces a new data structure within the cluster's main router, called the Content Store Station (CSS), it controls the caching and eviction of material according to its age. Furthermore, a Long Short-Term Memory (LSTM) model is trained to optimize caching decisions, determining whether specific content should be stored and identifying the most suitable location within the store station. Experimental results show that the LSTM model achieves 90% accuracy in caching decisions. The CSS-LSTM strategy was compared with other approaches, specifically LCE, Random, and NECS, across two different scenarios using the cache hit ratio metric, demonstrating superior performance.