The Future of Next Generation Web: Juxtaposing Machine Learning and Deep Learning-Based Web Cache Replacement Models in Web Caching Systems

Massive data generated by connected smart devices, particularly in distributed computer networks, contributed to the large network traffic burden caused by the ever-increasing use of the Internet infrastructure and web caching systems. Such connected smart devices generate massive data, shared and synchronized together in real-time over the connected nodes using various emerging technologies through the Web. However, the performance of classical web caching methods usually degrades when caching web objects due to various factors. Therefore, this study presents a comprehensive review of machine learning and deep learning-based web cache replacement models that could be effectively used to improve the performance of web caching systems. The study revealed that random forest, artificial neural networks, support vector machine, LSTM and fuzzy approaches are among web cache replacement models that have been used to improve the performance of web caching systems. However, due to the sparse use of web object features, some models are typically unable to handle today’s unpredictable web caching demands. Therefore, to deal with rising web usage, high latency, increased user-perceived delays, web cache overload, network traffic congestion, increased web object size, real-time data sharing, and limited bandwidth size, various web object attributes should be included in the next-generation adaptive and robust machine and deep learning-based web cache replacement models. In addition to web object features such as recency, frequency, cost, hit rate, modification and expiration time, more features are required in developing adaptive web caching algorithms to improve the performance of web caching systems.