DTE-SDN: A Dynamic Traffic Engineering Engine for Delay-Sensitive Transfer
Chuan Lin,Yuanguo Bi,Hai Zhao,Zheng Liu,Siyuan Jia,Jian Zhu
DOI: https://doi.org/10.1109/jiot.2018.2872439
IF: 10.6
2018-01-01
IEEE Internet of Things Journal
Abstract:With ever-rapid development of information and communication technologies, e.g., smart city, industrial Internet, Internet of Things, etc., enormous amounts of data are explosively generated and delivered to the computing center for further processing, which brings additional burden to both transfer components (e.g., the Internet) and data processing units. To efficiently schedule data transfer especially for delay-sensitive traffic, a scalable network architecture with intelligent traffic engineering (TE) policy is indispensable. In this paper, we propose a TE engine DTE-SDN by utilizing the software defined networking (SDN) technology, aiming to schedule the transfer of delay-sensitive traffic. Particularly, with OpenFlow, DTE-SDN captures an overall view of the network in real-time and can monitor the quality of service (QoS) metrics (e.g., throughput and delay) of each network link. In order to schedule the delay-sensitive transfer, a dynamic-scheduling scheme capable of multipath routing is proposed, which allows DTE-SDN to compute the near-optimal scheduling (e.g., path selection and flow distribution), based on the instantaneous QoS metrics. Especially, in the scheduling scheme, we propose a probabilistic-matching approach that aims to distribute the traffic among multiple end-to-end paths according to the computed flow distribution policy and can be deployed in SDN-enabled switches (e.g., Open vSwitch). The simulation results demonstrate that DTE-SDN is able to measure the throughput and delay with acceptable error range and can dramatically enhance the transfer efficiency than the traditional scheduling algorithms.