Conquering the Linux network stack: decapsulating packets with eBPF at 6Mpps+
summary
Describes moving customer traffic to VXLAN and the tools used to analyze that traffic, then evaluates options for VXLAN decapsulation and explains why eBPF was chosen. Covers the specific tools and implementation used to perform VXLAN decapsulation with eBPF and gives performance results and deployment conclusions (targeting multi-million packets-per-second throughput).
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:10VLAN issues in cloud growth
- 5:48Moving to VXLAN
- 6:51NetFlow statistics loss
- 9:18Finding eBPF solution
- 11:02What is eBPF
- 13:53Packet in network stack
- 17:26Kernel packet handling
- 21:02eBPF program development
- 26:47Packet decapsulation
- 28:32Results and optimization
- 31:20Performance tips
- 32:23Conclusion
- 32:44Q&A
related talks
37mEfficient and Flexible Virtual Machine Networking Through eBPF
30mA Beginner's Guide to eBPF Programming for Networking
32meBPF: BPF kernel infrastructure
32mDemystifying eBPF - eBPF Firewall from scratch
14mZero-overhead Container Networking with eBPF and Netkit
05mLightning Talk: Back to the Future with eBPF
35mAn Introduction to eBPF - Superpowers for Networking, Observability & Security
21mHow to Use XDP and eBPF to Accelerate IPSec Throughput by 400%