Flow-based Packet Processing Framework on DPDK and VPP
summary
Framework builds on DPDK and VPP and claims per‑flow state, ordered packets and feature sharing by assigning a persistent flow ID used across NIC offload and software path with RSS and RTD flow tables as fallback. Hardware flow offload accelerates classification, software mapping handles overflow, and TCP tracking watches 3‑way handshakes and sequence numbers. Supports bridge‑domain and VRF aware processing, policy routing and tunnel offload, maps one flow to many NICs, and reports up to 10,000 streams handled concurrently in hyper‑scan mode.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:21Traditional packet processing challenges
- 3:23Flow-based solution advantages
- 5:00Layer 7 flow identification
- 8:05Flow configuration mechanisms
- 10:03Software flow handling with RSS
- 12:36Framework integration with DPI
- 13:47TCP connection tracking
- 15:28Header scanner rules orchestration
- 22:31Performance and offloading summary
- 24:48Testing with open source tools
- 26:11Dynamic flow fallback
- 29:17Q&A
related talks
26mFlowBlaze: Stateful Packet Processing in Hardware
38mTelecom Packet Processing and Correlation Engine Using DPDK
16mBackpressure Flow Control
26mFreeFlow: Software-based Virtual RDMA Networking for Containerized Clouds
22mMonitoring 400G Traffic in DPDK Using FPGA-Based SmartNIC with RTE Flow
20mCan OVS-DPDK be further optimised?
51mAccelerate containers and VMs networking with vDPA
17mLightning Talk: eBPF based UPF (User Plane Function) for 5G
