In the context of the internet of things (IoT), backbone network traffic prediction is essential for intelligent resource allocation, dynamic traffic engineering, and network management. However, network traffic often exhibits strong nonlinearity, complex spatio-temporal dependencies, and significant sensitivity to network topology. Existing methods still suffer from limitations in long-sequence modeling, implicit topology learning, and effective spatio-temporal feature fusion. To address these challenges, this paper proposes ATGFormer, a spatio-temporal network traffic prediction model based on an adaptive topology gating mechanism. Specifically, the proposed model employs multi-scale pyramid encoding, adaptive graph convolution, and a topology-gated unit to capture both short-term and long-term temporal patterns, learn implicit topological dependencies, and dynamically fuse spatial and temporal features. A hierarchical decoder is further designed to aggregate multi-scale predictions for improved forecasting performance. Experimental results on the real-world Abilene and GEANT dataset demonstrate that the proposed model outperforms mainstream baselines in terms of prediction accuracy, long-sequence stability, and robustness. These results indicate that ATGFormer can provide accurate and robust traffic prediction for intelligent network management in Smart IoT environments.
Zhi-Peng Wang, Wei Wei, Yang Yu et al.· International Conferences on...· 0 citations
With the explosive development of LLM-empowered agent technology, LLM inference performance has become more important than training. Cloud computing is a popular deployment approach, where performance prediction is vital for instance selection and QoS assurance. However, prediction is challenging due to GPU hardware heterogeneity, Transformer operator variations, and dynamic inference configurations. Virtualization and other features vary across clouds, further increasing prediction difficulty. Existing methods suffer from low accuracy and poor generalization. To tackle these issues, we propose Dispeller, a prediction model for GPU-accelerated cloud environments with three feature sets: 1) basic GPU hardware feature with 7 dimensions; 2) operator-level GPU performance feature with 4 dimensions; 3) inference configuration feature with 4 dimensions. We conduct experiments on public cloud GPUs and collect a real-world dataset of 10,112 samples. Random Forest is adopted to learn the nonlinear mapping between features and performance. Experimental results show that Dispeller achieves high prediction accuracy with TPS $\mathrm{R}^{{2}} = 0.951$ on seen GPUs and 0.989 on unseen GPUs, demonstrating strong cross-GPU generalization. An ablation study confirms that inference configuration features contribute 87.3% of the predictive power. Dispeller is therefore able to recommend cloud resources and optimize LLM deployment costs.
Huan Zhou, Zhi-Peng Wang, Meng-Juan Li et al.· Fall Joint Computer Conferen...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
