Network bandwidth reservation method combining machine learning and linear programming

Abstract

Network bandwidth reservation is an expected service under software-defined network environment, where users directly reserve network resources in an on-demand manner. To provide extensive bandwidth reservation services, instantaneous response to user requests, and high user request acceptance ratio are required. In this study, we propose a novel bandwidth reservation method to meet these two requirements by combining machine learning (ML) and linear programming (LP), for unpredictable bandwidth demands in which the use time is indicated strictly. In the proposed method, a user request is instantaneously judged through ML, and network resource allocation, including traffic routing, is optimally determined through LP. We demonstrate that the proposed method provides a suboptimal acceptance ratio with a difference of less than 1% compared to the optimal solution, and an instantaneous response of less than 0.1ms under a general computation environment.