Sliding network coding (SNC) has emerged as a promising solution for ultra-reliable and low-latency communication (URLLC) scenarios. The performance of SNC, particularly in terms of developing the encoding matrix and decoding strategy, is heavily influenced by the order h of the underlying Galois field,GF(2h). In this paper, we investigate high-order SNC, whereh> 1, to enhance transmission efficiency by employing a Vandermonde-based encoding matrix that ensures linear independence among coded packets. To maximize decoding efficiency, we design a structure-aware decoding strategy (SA-DS), which not only dynamically exploits the relationships between successfully decoded (SD) packets and the currently decoded (CD) packet, but also utilizes the first-packet deterministic decoding (FPDD) property of the Vandermonde matrix. Additionally, we develop a Markov chain-based performance analysis framework in terms of retransmission probability, packet error rate, and expected decoding delay. Numerical results demonstrate that in the evaluated settings, the proposed scheme outperforms several traditional schemes in the moderate-erasure region. In the low-erasure region, its advantage becomes particularly pronounced (reaching one to three orders of magnitude in both PER and retransmission probability while maintaining a comparable decoding delay), making it particularly suitable for URLLC applications.

Video stream has become the largest part of the Internet traffic, and video Quality of Experience (QoE) is an important factor affecting the overall QoE of Internet users. Accurate measurement of video$\mathbf{Q o E}$from the perspective of traffic can help network operators to better plan network resources and improve user retention. To accurately measure video QoE through traffic, multiple factors need to be considered, with stalling being a significant one. Mean Opinion Score (MOS) is an indicator that quantifies video QoE. Traditionally, labeling MOS for videos requires subjective quality assessment experiments, which is time-consuming and labor-intensive. Second, the statistical features used by traditional machine learning methods are difficult to represent the stalling events in video steams, leading to inaccurate MOS measurements. Therefore, a number of past studies only estimate the indirect metrics that affect MOS, such as resolution and bitrate. In this paper, we use the parametric model ITU-T P. 1203 to automatically label MOS for videos. And we propose a two-stage approach to accurately measure MOS from video traffic. First, we introduce a stalling detection algorithm to identify stalling events from HTTP Adaptive Streaming (HAS) traffic. Then, we use these stalling events to produce secondary features and combine them with other statistical video stream features to establish a machine learning model for direct MOS measuring. Experiments show that our proposed method reduces the Root Mean Squared Error (RMSE) of MOS estimation from 0.455 to 0.406 compared to the conventional method.