Enabling In-situ Programmability in Network Data Plane: From Architecture to Language
Talkat 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22) · 2022-04-04..04-06·
summary
A new switch architecture and language extension enable changing protocol parsing, table layouts, and per-packet processing at runtime without stopping traffic: IPSA and rP4 compile incremental updates, use a template-based processor skeleton, a configurable interconnect and a disaggregated memory pool, and apply a big-bubble update to keep packets consistent. Implementations include the software switch ipbm and an FPGA prototype; FPGA costs are ~12.09% more LUTs and 9.69% more FFs, clocking 110.45 MHz, and pipeline downtime is only 0.34% of packets.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:29Development of Switching ASIC
- 0:58Motivations
- 2:18Programmability in PISA - Protocol (why not)
- 2:52Programmability in PISA - Processing Logic w
- 3:30Programmability in PISA - Memory
- 3:55IPSA vs. PISA (how we solve limitations of PISA)
- 5:01IPSA - Template-based Stage Processor why can
- 8:02IPSA - Big Bubble Update
- 9:22Workflow of rP4 (r - reconfigurable)
- 10:32Use Case
- 11:36Evaluation - Resources
- 12:53Evaluation - Update Performance
- 13:33Conclusion
related talks
01h 08mThe Network is the Computer: A Programming Language Perspective
47mDeep Programmability: A New Lens on Networking
20mContra: A Programmable System for Performance aware Routing
15mSynthesizing Runtime Programmable Switch Updates
16mA High-Speed Stateful Packet Processing Approach for Tbps Programmable Switches
11mPacket Manipulator Processor A RISC V VLIW Core For Networking Applications
14mBrain-on-Switch: Towards Advanced Intelligent Network Data Plane via NN-Driven Traffic...
14mEmpower Programmable Pipeline for Advanced Stateful Packet Processing
