An adaptive channel number tuning mechanism on parallel transfer with UDT

Abstract

Utilizing available bandwidth is a big challenge with advance of high bandwidth network infrastructures and devices. Especially, parallel protocol which creates multi-connections aims to maximize data transfer rate in a HBDP(High Bandwidth Delay Product) network. It is a critical issue to decide the number of optimum channels according to network environment variations. In this paper, we deal with an adaptive channel number tuning mechanism to decide the number of optimum channels on parallel transfer with UDT(UDP based Data Transfer protocol). Data transfer distance and network congestion are prominent network variants. They can be generally represented into RTT(Round Trip Time). We firstly conduct a theoretical analysis to investigate the impact of data transfer distance on the number of parallel channels, and present it as an equation. Then, we discuss about an adaptive channel number tuning algorithm to decide the number of optimal channels which may reflect the network status. The experimental results show our proposed method really rigorous in HBDP network.