On-line Traffic Scheduling optimization in IEEE 802.1Qch based Time-Sensitive Networks

Abstract

Time-sensitive networking (TSN) enables deterministic transmission on standard Ethernet with the time-aware shaping enhancements. The time-aware shaper usually requires complex schedules to control the input and output of packet queues of switches, and is applied in TSN systems with static time-sensitive traffic. However, for scenarios with dynamic deterministic transmission requirements, like industrial Internet of Things (IoT), such shaper may no longer be applicable as the heavy recomputation and reload overhead of scheduling configurations for each switch. To solve this problem, IEEE 802.1Qch recently proposed a Cyclic Queuing and Forwarding (CQF) based shaper that does not require complicated configurations. But currently no scheduling algorithm can be directly applied to the CQF based shaper with dynamic time-sensitive traffic. To make CQF practical in the dynamic TSN application scenarios, new scheduling algorithms are urgently required.In this paper, we propose an on-line flow injection time scheduling algorithm, namely Fits, to fully explore the deterministic transmission ability of IEEE 802.1 Qch based TSN networks. It incrementally generates traffic schedules for new time-sensitive flows by adjusting the sending time on the network adapter based on network resource utilization. There is no need to reconFigure previously scheduled flows. To evaluate this algorithm, it is compared with two counterparts Naive and Tabu under various TSN system settings. The experimental results show that the successfully scheduled time-sensitive flows of the proposed algorithm is 82.5% more and 29.7% less on average than Naive and Tabu respectively. And the computation time of Fits is two orders less than Tabu in a few seconds level, which is acceptable at system run time.