An interesting new open-source project born out of university research is Beeper to provide application-layer parsing within the in-kernel eBPF. With an HTTP parser implemented by Beeper in eBPF and providing an in-kernel HTTP cache, the throughput for web servers could be improved by 3x at high rates. Beeper and the notion of using in-kernel eBPF for application-layer parsing was presented earlier this month at the Linux Plumbers Conference. It's quite interesting and the videos from LPC2026 were recently posted. Below is the Beeper presentation. This work by researchers from ETH Zurrich as well as NYU and NVIDIA is quite promising and demonstrating yet another innovative use for eBPF. There is also the PDF slide deck for those short on time. The Beeper code for application-layer parsing of L7 protocols in eBPF is developed in Rust and available from GitHub. Use-cases include monitoring application layer traffic, caching, re--directions based on payload, and more. The presentation showed off as much as 3.4x throughput improvement by leveraging this eBPF approach. The hyper-fast-path project as an eBPF-based Hyper Rust HTTP library fast path is showing as much as 4.5x maximum throughput: It will be interesting to see what practical production implementations ultimately come out of this interesting research work.
Accelerating HTTP Web Servers With In-Kernel eBPF Deliver Up To 4.5x Throughput
Full Article
Original Source
Read the full article at Phoronix →KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.