Toy-IoT-Oriented data-driven CDN performance evaluation model with deep learning

Abstract

Abstract Content Delivery Network(CDN) is geographically distributed network of cache servers. It can deliver the Internet content based on the users’ geographic position and real-time quality of service(QoS). Now with the rapid development of the Internet of things, IoT also needs CDN acceleration. Therefore, CDN not only needs to serve people, but also needs to serve IoT, such as sensors,toys. Especially for toy computing, CDN need to sink further to support the uplink data transmission. Because the performance of CDN is crucial to the resource management of existing platform and the acceleration of data transmission, CDN providers use different models to exactly describe the performance of CDN. Traditional models are linear models or need to be adjusted manually. These normally existing drawbacks make the methods hardly to be applied stably. Recently, deep learning(DL) has made great breakthroughs in solving many problems. We use the RNN in deep learning to model the CDN performance. Our design can exactly capture the nonlinear relationship between high-dimensional machine data and the CDN performance. And it can also realize the correct prediction of the reach rate which is the CDN’ main performance evaluation index in our design.The experimental results have shown that our model is able to outperform state-of-the-arts models.