Making content caching policies 'smart' using the deepcache framework

Abstract

In this paper, we present Deepcache a novel Framework for content caching, which can significantly boost cache performance. Our Framework is based on powerful deep recurrent neural network models. It comprises of two main components: i) Object Characteristics Predictor, which builds upon deep LSTM Encoder-Decoder model to predict the future characteristics of an object (such as object popularity) - to the best of our knowledge, we are the first to propose LSTM Encoder-Decoder model for content caching; ii) a caching policy component, which accounts for predicted information of objects to make smart caching decisions. In our thorough experiments, we show that applying Deepcache Framework to existing cache policies, such as LRU and k-LRU, significantly boosts the number of cache hits.

An Incremental Learning Based Edge Caching System: From Modeling to Evaluation

“…At the same time, some studies utilize some complicated learning methods to manage edge caching content by using offline learning models. Narayanan et al [8] use an offline-trained LSTM Encoder-Decoder model to forecast the popularities of objects and to prefetch popular objects. The method needs a large training set to train a deep learning model.…”

“…Different from learning methods used in [8], [9], the incremental learning algorithm continuously predicts input instances and incrementally updates the predicting model after receiving new instance. The model is originally proposed by a small amount of training data, and make predictions for the coming instance.…”

“…Then, instance error weights w t i are updated according to Equation (8) and are normalized to F t i to evaluate the competence of base classifiers. Misclassified instances give a higher weight for better learning to base classifiers than those instances correctly classified.…”

“…However, we observe that some of these learning approaches [8], [9], especially those offline learning models, may have the following problems for edge caching. Firstly, these methods periodically discard the old model since the model may no longer be suitable for current environment.…”

Tile-Based Edge Caching for 360° Live Video Streaming

“…Our system uses Long Short-Term Memory (LSTM) networks, which are efficient for time series forecasting [15], to decide which tiles of the 360 o videos should be cached at the edge caches and in what quality. LSTM networks have been previously used for predicting video popularity [16], [17] of traditional videos in cache networks, but they have not been used for optimizing the cached content in cache networks for the case of 360 o videos we examine in this paper. Through the use of the LSTM networks, the popularity of the 360 o videos and tiles for the next Group of Pictures (GOP) is predicted with only a small error.…”