Empower Programmable Pipeline for Advanced Stateful Packet Processing
Talkat 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) · 2024-04-16..04-18·
summary
Proposes RAPID: a pipeline extension that adds a low-cost side ring for fast backward state writes and uses speculative execution to avoid stalling, plus an enhanced P4 for programming. Builds an FPGA prototype and an ASIC emulator; shows 2–3× throughput and 1–2× latency improvement over blocking pipelines, with at most ~5% ASIC area/power overhead and low FPGA resource use. Demonstrates several stateful functions running correctly under these constraints.
AI-written summary · about
chapters (9)
- 0:00Stateful network functions intro
- 2:30Complex stateful packet challenges
- 4:39Traditional architecture decline
- 6:47Efficient stateful network solutions
- 8:16Speculative execution for stateful
- 10:05RAPID architecture design
- 12:13RAPID hardware prototype
- 13:18RAPID performance improvements
- 14:00Conclusion and future
related talks
15mEnabling In-situ Programmability in Network Data Plane: From Architecture to Language
15mWhen P4 Meets Run-to-completion Architecture
16mRuntime Programmable Switches
16mState-Compute Replication: Parallelizing High-Speed Stateful Packet Processing
15mIsolation Mechanisms for High-Speed Packet-Processing Pipelines
26mFaster Software Packet Processing on FPGA NICs with eBPF Program Warping
11mPacket Manipulator Processor A RISC V VLIW Core For Networking Applications
28mSafe, Modular Packet Pipeline Programming
