Oct 2026· Conference on Optics and Communication Technology· Vol 14385, pp. 143851F - 143851F-7· 0 citations· 14 references
Engineering
TL;DR
This paper proposes an anomaly traffic prediction and resource pre-scheduling mechanism based on the Informer model, offering a low-cost, high-precision solution for deterministic low-latency smart campus networks.
Abstract
The digital transformation of education introduces high-concurrency events (e.g., course registration, online exams) that cause sudden traffic surges in campus optical networks, leading to severe packet loss and delay jitter. Traditional static bandwidth allocation or reactive dynamic bandwidth allocation (DBA) fails to anticipate these bursts. This paper proposes an anomaly traffic prediction and resource pre-scheduling mechanism based on the Informer model. The model fuses historical traffic data with academic calendar events as prior knowledge, using ProbSparse self-attention to capture long-range dependencies and a generative decoder for multi-step prediction. Based on the predicted traffic trends, an SDN/ONU cooperative framework pre-allocates optical bandwidth before congestion occurs. Experiments on one-year real campus network data show that Informer achieves a MAPE of 8.64% (6.5% during peak events), reduces packet loss by 52%, and maintains end-to-end delay below 5 ms. The proposed mechanism requires no hardware upgrades, offering a low-cost, high-precision solution for deterministic low-latency smart campus networks.
The growth of distributed training models, with parameters now reaching the billion-scale, has shifted the system bottleneck from computation to communication. While Remote Direct Memory Access (RDMA) is widely deployed to improve network performance by circumventing the kernel mechanism, the synchronization-computatio...
Pingping Dong, Liying Chen, Xuan Yao et al.· IEEE Transactions on Network...· 0 citations
Artificial-intelligence data centers (AIDCs) can provide spatiotemporal load flexibility by migrating training workloads in response to regional electricity prices and renewable-energy (RE) availability. However, transfer-blind scheduling may overestimate migration benefits by overlooking transfer time or prevent train...
Xiu-Chi Tong, Zi-Jian Li, Xiao-Yi Fan et al.· 0 citations
In modern video streaming systems, clients typically make independent bitrate decisions without full knowledge of network conditions, often leading to inefficient resource usage and unstable quality. Network-assisted adaptive streaming is rapidly gaining importance in Network and Service Management (NSM), as operators...
Mustafa Türk, M. Sayıt, A. Begen et al.· IEEE Transactions on Network...· 0 citations
Irregular workloads, such as those stemming from Graph Neural Networks (GNNs) and Mixture-of-Experts (MoE) models, frequently induce severe incast congestion in large-scale parallel environments. Existing fabric-level congestion control mechanisms rely on reactive, link-level rate limiting that degrades overall through...
Noel Gomillion, Lance G. Fletcher, T. Steil et al.· 0 citations
Deploying trillion-parameter large language models across metropolitan environments is required to sustain real-time inference. Urban power constraints, however, prohibit monolithic GPU clusters, forcing the integration of distributed supernodes into a citywide compute fabric. Over 100-km distances, optical propagation...
Liang Guo, Ji-Zhuang Zhao, Wei Quan et al.· IEEE Transactions on Network...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.