Net-P4ct: Enhanced WAN Bandwidth Fair Sharing Using P4 Programmable Switches
summary
Net-P4ct moves QoS control into P4 programmable switches and pairs lightweight NetAgent eBPF tags with a centralized controller that runs weighted max‑min fair allocation (water‑filling). It replaces host rate limits with DSCP remarking plus two enforcement modes (per‑packet meters and statistics‑bucket coloring with burst mitigation), steers traffic via VXLAN to P4 clusters, and reports production numbers: 200 Tbps backbone, ~1M host agents costing ~20K CPU cores avoided, <3% CPU on 16K flows vs ~15% for host throttling.
AI-written summary · about
chapters (17)
- 0:00Net-P4ct purpose
- 1:45Existing bandwidth challenges
- 4:34Conventional switches' limits
- 5:18NetPact P4 architecture
- 6:44Bandwidth request model
- 8:11Reserved vs non-reserved allocation
- 9:15Fairness model details
- 10:19Traditional rate limiting limits
- 11:23DSCP remarking policies
- 13:51Statistics enforcement & bursts
- 14:36Enforcement policy evaluation
- 15:39CPU overhead comparison
- 16:00Deployment insights
- 16:43Cluster scaling strategies
- 17:26Failure detection & fail-safe
- 18:29Portability & future plans
- 18:50Conclusion
related talks
17mHiP4-UPF: Towards High-Performance Comprehensive 5G User Plane Function on P4...
01h 08mThe Network is the Computer: A Programming Language Perspective
16mPowerTCP: Pushing the Performance Limits of Datacenter Networks
12mP4EC: Enabling Terabit Edge Computing in Enterprise 4G LTE
16mBackpressure Flow Control
15mEverything Matters in Programmable Packet Scheduling
15mWhen P4 Meets Run-to-completion Architecture
15mEnabling In-situ Programmability in Network Data Plane: From Architecture to Language
