Deep Learning-based Predictive Caching in the Edge of a Network
Saidur Rahman, Md. Golam Rabiul Alam, Md. Mahbubur Rahman · 2020
Edge computing can allow seamless online content streaming for users because edge devices are usually placed closer to the users than the cloud data storage. As a result, users can be provided with the content faster than the time it would have required if the content was provided directly from the cloud server. However, there is a finite storage space in the cache of an edge device and since there has been a rapid growth in the amount of web content generated daily and also a growth in the number of users, therefore, smarter decisions need to be made to cache contents effectively so that maximum users can enjoy seamless web content streaming. In this paper, we propose a Deep Learning-based caching strategy to store content in the edge network. Our proposed method uses Recurrent Neural Network model using Long Short Term Memory (LSTM) cells trained on the MovieLens 20M dataset. The end result is a robust learning algorithm capable of predicting the popularity of a web content. We then propose an algorithm that uses the results of the Deep Learning model to make decisions about which contents to store in the cache of the Edge device at the Network Edge.