Dynamic Traffic Prediction of SDN Under Attention Mechanism of Spatiotemporal Graph

doi:10.3778/j.issn.1002-8331.2401-0119

Abstract

Abstract: Addressing the issue of static spatio-temporal dependency in traffic prediction for SDN (software-defined networking) traffic engineering, this paper proposes a dynamic network traffic prediction method, AGCNGRU (attention mechanism for GCNGRU model), which integrates graph convolutional neural network (GCN) and gated recurrent unit (GRU) with an attention mechanism. By leveraging GCN to capture the spatial dependency of traffic between nodes in the network and GRU to capture the temporal dependency of traffic flowing through various nodes in the network, this method designs the weight of each hidden state through a temporal attention mechanism to adjust the importance of traffic information at different time points. Meanwhile, a data-driven spatial attention mechanism is employed to dynamically and adaptively adjust the Laplace matrix, enabling the dynamic extraction of spatio-temporal correlations from network information data and ultimately achieving accurate dynamic traffic prediction. Experiments conducted on the GEANT dataset show that the proposed method reduces the mean squared error by 24.8% compared to GCNGRU and by 66.4% compared to GRU. Furthermore, when compared with traditional routing algorithms such as OSPF and DDPG, the network performance is improved by 24% over OSPF and 8.1% over DDPG at 90% traffic load intensity, further demonstrating the practicality and effectiveness of the AGCNGRU algorithm in accurately predicting network traffic.

Key words: software-defined networking (SDN), traffic prediction, graph convolution neural network, gating recursive unit

摘要： 针对SDN流量工程中流量预测基于静态时空依赖的问题，提出了一种基于注意力机制的图卷积神经网络（GCN）与门控递归单元（GRU）集成的动态网络流量预测方法——AGCNGRU（attention mechanism for GCNGRU model）。借助GCN捕获网络中节点之间的流量空间依赖性和GRU捕获流量经过网络中各节点的时间依赖性，通过时间注意力机制设计每个隐藏状态的权重，以调整时间点流量信息的重要性，同时通过数据驱动空间注意力机制动态自适应调整Laplace矩阵，实现动态提取网络信息数据时空相关性，最终完成动态流量精准预测。在GEANT的数据集上的实验表明，所提出的方法在均方误差方面比GCNGRU减少24.8%，比GRU减少66.4%，并通过与传统路由算法OSPF、DDPG算法比较，在90%的流量负载强度下，网络性能比OSPF提升了24%，比DDPG提升了8.1%，进一步说明了AGCNGRU算法网络流量准确预测带来的时效性和有效性。

关键词: 软件定义网络（SDN）, 流量预测, 图卷积神经网络, 门控递归单元

LYU Guanghong, WANG Kun. Dynamic Traffic Prediction of SDN Under Attention Mechanism of Spatiotemporal Graph[J]. Computer Engineering and Applications, 2025, 61(8): 267-273.

吕光宏, 王坤. 时空图注意力机制下的SDN网络动态流量预测[J]. 计算机工程与应用, 2025, 61(8): 267-273.

References

[1] ABBASI M R, GULERIA A, DEVI M S. Traffic engineering in software defined networks: a survey[J]. Journal of Telecommunications and Information Technology, 2016, 4: 3-14.
[2] 杨洋, 吕光宏, 赵会, 等. 深度学习在软件定义网络研究中的应用综述[J]. 软件学报, 2020, 31(7): 2184-2204.
YANG Y, LYU G H, ZHAO H, et al. Survey on deep learning applications in software defined networking research[J]. Journal of Software, 2020, 31(7): 2184-2204.
[3] MYKOLA B, KRYVINSKA N, BESHLEY H, et al. Traffic engineering and QoS/QoE supporting techniques for emerging service-oriented software-defined network[J]. Journal of Communications and Networks, 2024, 26(1): 99-114.
[4] LANER M, SVOBODA P, RUPP M, et al. Parsimonious fitting of long-range dependent network traffic using ARMA models[J]. IEEE Communications Letters, 2013, 17(12): 2368-2371.
[5] ZHAO X C, DU D J, ZHANG Y. Prediction of SDN heterogeneous network traffic based on improved LSTM with self-attention mechanism[C]//Proceedings of the 8th International Conference on Power and Renewable Energy, 2023: 2016-2021.
[6] GAO K H, LI D, CHEN L, et al. Incorporating intra-flow dependencies and inter-flow correlations for traffic matrix prediction[C]//Proceedings of the IEEE/ACM 28th International Symposium on Quality of Service, 2020: 1-10.
[7] JOSE S V, ALMASAN P, GAIMES-FERRIOL M, et al. Graph neural networks for communication networks: context, use cases and opportunities[J]. IEEE Network, 2023, 37(2): 146-153.
[8] YANG L, GU X X, SHI H F, et al. A noval satellite network traffic prediction method based on GCN-GRU[C]//Proceedings of the International Conference on Wireless Communications and Signal Processing, 2020: 718-723.
[9] WANG X T, SUN Y B, WANG X, et al. Traffic matrix prediction in SDN based on spatial-temporal residual graph convolutional network[C]//Proceedings of the 35th Chinese Control and Decision Conference, 2023: 3382-3387.
[10] 赵文竹, 袁冠, 张艳梅, 等. 多视角融合的时空动态图卷积网络城市交通流量预测[J]. 软件学报, 2024, 35(4): 2204-2225.
ZHAO W Z, YUAN G, ZHANG Y M, et al. Multi-view fused spatial-temporal dynamic graph convolutional network for urban traffic flow prediction[J]. Journal of Software, 2024, 35(4): 2204-2225.
[11] XU W, LIU J J, YAN J W, et al. Dynamic spatiotemporal graph wavelet network for traffic flow prediction[J]. IEEE Internet of Things Journal, 2024, 11(5): 8019-8029.
[12] XIONG Y J, WANG H J. Spatio-temporal contextual conditions causality and spread delay-aware modeling for traffic flow prediction[J]. IEEE Access, 2024, 12: 21250-2126.
[13] HUANG S Y, ZHANG H, WANG X F, et al. Fine-grained spatio-temporal distribution prediction of mobile content delivery in 5G ultra-dense networks[J]. IEEE Transactions on Mobile Computing, 2024, 23(1): 469-482.
[14] NAN H, LI R D, ZHU X Y, et al. Spatio-temporal identity multi-graph convolutional network for traffic prediction in the metaverse[J]. IEEE Journal on Selected Areas in Communications, 2024, 42(3): 669-679.
[15] GUO K, HU Y L, QIAN Z, et al. Optimized graph convolution recurrent neural network for traffic prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(2): 1138- 1149.
[16] VASWANI A, SHAZEER N, PARMAR N. et al. Attention is all you need[C]//Proceedings of the 31st Conference on Neural Information Processing Systems, 2017: 6000-6010.
[17] GUO K, HU Y L, QIAN Z, et al. Dynamic graph convolution network for traffic forecasting based on latent network of laplace matrix estimation[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(2): 1009-1018.
[18] UHLIG S, QUOITIN B, LEPROPRE J, et al. Providing public intradomain traffic matrices to the research community[J]. ACM SIGCOMM Computer Communication Review, 2006, 36(1): 83-86.
[19] HOU Y N, LIU L F, WEI Q, et al. A novel DDPG method with prioritized experience replay[C]//Proceedings of the 2017 IEEE International Conference on Systems, Man, and Cybernetics, 2017: 316-321.