Between Packets and Predictions: Analyzing the Overhead of In-Network Machine Learning in Programmable Switches
Resumo
Recent advances in programmable data planes have sparked significant interest in executing machine learning inference directly within network devices. Prior work demonstrates the feasibility of this paradigm across diverse applications, ranging from Network Security to Quality of Service. However, the network-level performance overhead and effects in perceived quality remain largely unexplored. This paper assesses the impact of executing machine learning inference inside network devices on key metrics such as throughput, queueing delay, and application quality. Through controlled experiments covering web traffic, bulk file transfer, and video streaming, the performance of machine-learning-enabled pipelines is contrasted with standard forwarding. Specific attention is given to adaptive video delivery, where MPEG-DASH metrics indicate that while adaptive mechanisms preserve playback stability, the integration of additional processing features leads to a saturation of quality gains and a marked increase in tail latency. The findings provide valuable insights into the systemic costs of embedding intelligence into smartswitch data planes.
Referências
Akem, A. T.-J., Fraysse, G., and Fiore, M. (2025). Real-time encrypted traffic classification in programmable networks with p4 and machine learning. International Journal of Network Management, 35(1):e2320.
Akem, A. T.-J., Gucciardo, M., and Fiore, M. (2023). Flowrest: Practical flow-level inference in programmable switches with random forests. In IEEE INFOCOM 2023-IEEE Conference on Computer Communications, pages 1–10. IEEE.
Beuttenmuller, D., Valério, M., Silva, C., da Silva, I., Jr., P. D. M., and de Almeida, L. (2025). A new wave: Exploring new load pattern models for experimentation in computer networks. In Anais Estendidos do XLIII Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos, pages 11–22, Porto Alegre, RS, Brasil. SBC.
Bosshart, P., Daly, D., Gibb, G., Izzard, M., McKeown, N., Rexford, J., Schlesinger, C., Talayco, D., Vahdat, A., Varghese, G., and Walker, D. (2014). P4: Programming protocol-independent packet processors. SIGCOMM Comput. Commun. Rev., 44(3):87–95.
Boutaba, R., Salahuddin, M. A., Limam, N., Ayoubi, S., Shahriar, N., Estrada-Solano, F., and Caicedo, O. M. (2018). A comprehensive survey on machine learning for networking: evolution, applications and research opportunities. Journal of Internet Services and Applications, 9(1):1–99.
Fatur Lessa, J. V., Marques, J. A., Parizotto, R., and Gaspary, L. P. (2025). In-network inference with neuralp4: Auto-generating full-fledged nn models for pdps. In NOMS 2025-2025 IEEE Network Operations and Management Symposium, pages 1–10.
Musumeci, F., Fidanci, A. C., Paolucci, F., Cugini, F., and Tornatore, M. (2022). Machine-learning-enabled ddos attacks detection in p4 programmable networks. Journal of Network and Systems Management, 30(1):21.
Nguyen, H. N., Nguyen, M.-D., and Montes de Oca, E. (2024). A framework for in-network inference using p4. In Proceedings of the 19th International Conference on Availability, Reliability and Security, pages 1–6.
Shirmarz, A., Bragatto, M. N., Verdi, F. L., Singh, S. K., Rothenberg, C. E., Patra, G., and Pongrácz, G. (2025). In-network ar/cg traffic classification entirely deployed in the programmable data plane: Unlocking rtp features and l4s integration. In 2025 IEEE 11th International Conference on Network Softwarization (NetSoft), pages 477–485.
Xiong, Z. and Zilberman, N. (2019). Do switches dream of machine learning? toward in-network classification. In Proceedings of the 18th ACM Workshop on Hot Topics in Networks, HotNets ’19, page 25–33, New York, NY, USA. Association for Computing Machinery.
Zhang, K., Samaan, N., and Karmouch, A. (2024). A machine learning-based toolbox for p4 programmable data-planes. IEEE Transactions on Network and Service Management, 21(4):4450–4465.
Zhang, K., Zheng, C., Samaan, N., Karmouch, A., and Zilberman, N. (2025a). Design, implementation, and deployment of multi-task neural networks in programmable data-planes. IEEE Transactions on Network and Service Management, pages 1–1.
Zhang, K., Zheng, C., Samaan, N., Karmouch, A., and Zilberman, N. (2025b). Muta: Enabling multi-task neural network inference in programmable data-planes. In 2025 IEEE 26th International Conference on High Performance Switching and Routing (HPSR), pages 1–6.
Zheng, C., Hong, X., Ding, D., Vargaftik, S., Ben-Itzhak, Y., and Zilberman, N. (2024a). In-network machine learning using programmable network devices: A survey. IEEE Communications Surveys & Tutorials, 26(2):1171–1200.
Zheng, C., Xiong, Z., Bui, T. T., Kaupmees, S., Bensoussane, R., Bernabeu, A., Vargaftik, S., Ben-Itzhak, Y., and Zilberman, N. (2024b). Iisy: Hybrid in-network classification using programmable switches. IEEE/ACM Transactions on Networking, 32(3):2555–2570.
