Offloading Network Traffic Classification to Hardware
Talkat Open Source Summit Europe + Embedded Linux Conference Europe 2019 · on Wednesday, 2019-10-30· at 15:15· Forum 3·
summary
Hardware classification is implemented end-to-end: header parsing in the MAC uses a TCAM-based parser to dissect packets at 10Gbps, then classifier tables (PPv2 example: 256 TCAM entries, 512 instructions, four engines; driver uses two) match 2/5-tuple and VLAN keys to drop, redirect, or steer to a receive queue; offload is configured from userspace via TC flower and ethtool (netfilter/flower unification is in progress). Limits: TCAM is scarce and shared across ports, and dropped packets can be invisible to the kernel counters.
AI-written summary · about
chapters (6)
- 0:00Start
- 6:22A Typical Network Packet Is Built
- 15:05Pros and Cons
- 16:43Cons
- 29:39Mac and Villian Filtering
- 34:21Hardware Counters
related talks
18mLeo: Online ML-based Traffic Classification at Multi-Terabit Line Rate
21mAccelTCP: Accelerating Network Applications with Stateful TCP Offloading
19mConntrack offload: why and how
34mStrategies for Rate Limiting Network Packet Ingress
30mNicPic: Scalable and Accurate End-Host Rate Limiting
19mEnabling Programmable Transport Protocols in High Speed NICs
53mWhat is eating my traffic? Packet thieves revealed
17mRearchitecting the TCP Stack for I/O-Offloaded Content Delivery