Programmable Data Plane Intelligence: Advances, Opportunities, and Challenges

Abstract

Recently, Programmable Data Planes (PDPs) have brought new opportunities to unprecedented innovations in network protocols and architectures. Among them, programmable data plane intelligence, including achieving networking for Machine Learning (ML) and ML for networking with PDP, has attracted great interest from researchers. Motivated by this research trend, this article presents a survey of recent advances in the use of PDP in terms of machine learning. We first introduce PDP intelligence and the benefits of assisting ML with PDP. Then, the difficulties of realizing it are shown, which are caused by PDP's limited capabilities. After that, we present recent advances to cope with the difficulties, where each work is analyzed from multiple perspectives. Finally, we show the opportunities and challenges for further research.