Sirius: Composing Network Function Chains into P4-Capable Edge Gateways
Jiaqi Gao, Jiamin Cao, Yifan Li, Mengqi Liu, Ming Tang, Dennis Cai, Ennan Zhai
Abstract
Alibaba Cloud designs and deploys P4-capable gateway to accelerate the processing of the diverse business traffics in the edge cloud. Since the programmable ASIC in the gateway only accepts a monolithic, pipelined P4 program, the dozens network function chains for different business traffics have to be composed into one. This is non-trivial due to the contention between the complexity of network function chains and the limited resource in the programmable ASIC. In this paper, we present Sirius, a system that automates network function chain composition process. Sirius synthesizes tables to identify which business traffic the input packet belongs to, pipelines loops in the merged network function graph via recirculations, and partitions the graph between programmable ASIC and CPU when the required memory consumption exceeds the ASIC's capability. So far, Sirius has automated network function arrangement in hundreds of gateways, and has effectively decreased our programmers' workload by three orders of magnitude, from weeks to minutes.
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 1713244b-05f8-4e08-8bfe-25a12b9dff86Cited by top-tier papers4
- P4runpro: Enabling Runtime Programmability for RMT Programmable SwitchesYifan Yang, Lin He, Jiasheng Zhou, Xiaoyi Shi et al.SIGCOMM 2024 · 12 citations
- Towards Optimal Rack-scale μs-level CPU Scheduling through In-Network Workload ShapingXudong Liao, Han Tian, Xinchen Wan, Chaoliang Zeng et al.USENIX ATC 2025 · 1 citation
- SPLIDT: Partitioned Decision Trees for Scalable Stateful Inference at Line RateMurayyiam Parvez, Annus Zulfiqar, Roman Beltiukov, Shir Landau Feibish et al.NSDI 2026 · 1 citation
- Capybara: Dynamic Load Balancing with Microsecond-Scale TCP MigrationInho Choi, Nimish Wadekar, Guangda Sun, Raj Joshi et al.SIGCOMM 2026
Builds on8
- TEA: Enabling State-Intensive Network Functions on Programmable SwitchesDaehyeok Kim, Zaoxing Liu, Yibo Zhu, Changhoon Kim et al.SIGCOMM 2020 · 121 citations
- Sailfish: accelerating cloud-scale multi-tenant multi-service gateways with programmable switchesTian Pan, Nianbing Yu, Chenhao Jia, Jianwen Pi et al.SIGCOMM 2021 · 111 citations
- Flightplan: Dataplane Disaggregation and Placement for P4 ProgramsNik Sultana, John Sonchack, Hans Giesen, Isaac Pedisich et al.NSDI 2021 · 95 citations
- Composing Dataplane Programs with μP4Hardik Soni, Myriana Rifai, Praveen Kumar, Ryan Doenges et al.SIGCOMM 2020 · 54 citations
- Switch Code Generation Using Program SynthesisXiangyu Gao, Taegyun Kim, Michael D. Wong, Divya Raghunathan et al.SIGCOMM 2020 · 51 citations
Related papers
- Cetus: Releasing P4 Programmers from the Chore of Trial and Error CompilingYifan Li, Jiaqi Gao, Ennan Zhai, Mengqi Liu et al.NSDI 2022 · 22 citations
- LuoShen: A Hyper-Converged Programmable Gateway for Multi-Tenant Multi-Service Edge CloudsTian Pan, Kun Liu, Xionglie Wei, Yisong Qiao et al.NSDI 2024 · 27 citations
- OptimusPrime: Unleash Dataplane Programmability through a Transformable ArchitectureZhikang Chen, Yong Feng, Shuxin Liu, Haoyu Song et al.SIGCOMM 2024 · 6 citations
- Albatross: A Containerized Cloud Gateway Platform with FPGA-accelerated Packet-level Load BalancingJianyuan Lu, Shunmin Zhu, Jun Liang, Yuxiang Lin et al.SIGCOMM 2025 · 4 citations
- T4G: Trace-based P4 Program GenerationChenxing Ji, Timo Jugariu, Sebastijan Dumancic, Fernando KuipersINFOCOM 2026
