A variable length gating approach for queue scheduling in deterministic networks

Abstract

The emergence of deterministic network technologies in B5G/6G networks such as Time Sensitive Network (TSN) and Deterministic Network (DetNet) overcomes the drawbacks of traditional Ethernet BE (Best Effort) service and provides bounded deterministic forwarding services such as low delay and high reliability for the time-sensitive critical services. However, the current gating queue scheduling method employed by the Deterministic Internet Protocol (DIP) is a timed polling multi-queue scheduling technology, and the gating period of each gate is fixed with limited flexibility. When the phase characteristics of services are not matched due to the nonadjustable gating period, extra queuing delay and low network resource utilization are likely to occur. Therefore, in this paper, a variable length gating approach for queue scheduling is proposed, which fully considers the phase characteristics of the packets. Based on the statistics of packet arrival rules, an observation window mechanism is set to adaptively adjust the gating period on a per-window basis. Numerical results are provided to demonstrate the proposed approach can further reduce the average delay of packet queuing of the traditional approach, which can also improve the utilization of network resources and the flexibility of queue scheduling.