Optimizing Content Delivery Networks for Enhanced Performance and Energy Efficiency

Content Delivery Networks (CDNs) are fundamental to modern Internet content distribution, enabling high-speed delivery of web pages, videos, and other online resources. Optimizing caching strategies within CDNs is crucial for improving network efficiency and ensuring a seamless user experience. This research explores the efficacy of traditional caching algorithms such as LFU, LRU, RAND, and FIFO, in comparison to machine learning algorithms including Linear Regression (LR), Gradient Boosting (GB), K-Nearest Neighbours (kNN), Decision Tree (DT), AdaBoost, Random Forest (RF), LightGBM, and XGBoost. The findings demonstrate that machine learning techniques can significantly enhance CDN performance by identifying optimal caching locations and adapting to traffic patterns. Notably, the XGBoost model achieved a cache hit rate of Hyperparameters used for different 76.31 %, a latency of 0.98 seconds, and a throughput of $18,796.16$ requests per second, surpassing traditional methods in hit rate and overall efficiency. The study underscores the need for predictive and adaptive caching approaches in modern traffic management, highlighting the benefits of improved caching on user experience, operational costs, and environmental impact.