Programmable Switch Aided Content Popularity Prediction and Caching Strategy

Abstract

Content distribution is the most critical task for the current Internet, (e.g., the estimated video traffic will reach 82 percent of the Internet traffic by 2022). With the fast increase of load of the network, traditional host-centric based network paradigm (i.e., TCP/IP) faces great challenges in terms of efficiency, security, and privacy. To solve the problems confronting the current Internet, the Information-Centric Network (ICN) becomes a promising solution, where the focal point is identified content rather than specific host addresses. This paradigm brings many benefits, e.g., network traffic reduction, low retrieval latency. Besides, benefiting from the advance of programmable network hardware, the operator can reconfigure the network hardware' behavior, thus providing hardware support to describe the ICN instances. However, ICN also poses new challenges to cache management. The cache redundancy and unequal resource allocation will seriously affect the performance of the network. In this paper, we propose a distributed variational Bayes aided content popularity prediction algorithm. The extensive and indepth simulations are performed to evaluate our proposed algorithm in comparison to the other state-of-the-art schemes.