Jesper D. Brouer

Me and ⁦@JesperBrouer⁩ wrote a blog post on the kernel’s BPF LPM trie implementation and cover some of the performance gotchas with the current code. blog.cloudflare.com/a-deep-dive-in…

Awesome blogpost/talk + tool by @bristot : Linux scheduling latency debug and analysis bristot.me/linux-scheduli… Let's us look into the "blackbox" of Linux sched latency!😃

Fascinating discussion about raw sockets aka ETH_P_ALL @mail.gmail.com/T/" target="_blank" rel="nofollow noopener">lore.kernel.org/netdev/CANP3RG… The theory is that any socket AF_PACKET ETH_P_ALL socket also catches TX packets, working against TX zerocopy.

New year, new spikes - XDP with Mellanox 100G NICs does pretty good since we took it in prod🙂

Netflix/Intel working together to find low-level PMC-based perf wins in the cloud (by my colleagues Vadim and Harshad) netflixtechblog.com/seeing-through…

@JesperBrouer Thanks for telling us about all the cool stuff eBPF can be used for 👍🏻 #DrivingIT22

@RedHat funny feature to why you should use eBPF to unlock the potential of your Linux kernel 😉🙌🏼 #DrivingIT22

@toke_dk is telling us about decreasing latency #DrivingIT22

Next up at #P99CONF: @TalawahTech uses an HTTP benchmark to compare the performance of the #Linux kernel networking stack with userspace networking doing kernel bypass via DPDK. Grab your seat here: bit.ly/3q3Qqif #ScyllaDB

I'm very happy to introduce FastNetMon Community 1.2.3. This release includes a completely new network packet parser and XDP support. You may find the full list of changes here: github.com/pavel-odintsov… Installation: fastnetmon.com/install/

Switching to Cilium Layer 4 Load Balancer XDP allowed @seznam_cz to drastically reduce their CPU consumption and increase performance Learn how in the video from @eBPFsummit youtube.com/watch?v=xwjZF3…

To achieve optimal speed and size of eBPF programs, use the latest version of the eBPF instruction set. By default llvm sticks to v1 which is non-optimal for most cases. pass -mcpu=probe to llc to auto-select latest instruction set supported by your kernel pchaigno.github.io/bpf/2021/10/20…

@omsandov Idea: Identify if memory address comes from a page_pool

My favorite new helper is identify_address(), which can tell you where an arbitrary address came from. Currently, it identifies symbols (e.g., global variables, function pointers) and slab objects, but we'd love to add more types of addresses in the future (give me ideas!)

drgn 0.0.21 is out now! This release adds a bunch of new helpers (especially for memory management) and fixes some important bugs. See the release notes here: github.com/osandov/drgn/r….

This Friday @lizrice and @dave_tucker will be exploring bpfd which is a framework for running BPF tracers with rules on Linux as a daemon that is container aware eBPF & Cilium Office Hours Friday, 14th October - 9am ET / 3pm CEST Livestream: youtube.com/watch?v=KskHG7…

One of the most fun videos about #bufferbloat #openwrt and #sqm I've *ever seen*. I am a huge fan of the right to repair also. Yay! @toke_dk @dsearls @doctorow @JimGettys @jlivingood

After a long wait, the first XDP based DDoS-Filter has landed in Frankfurt - throughout the week, all in! :)

Wondering how eBPF is changing edge computing? Come listen to @toke_dk from @RedHat explain how

Almost 2 days ago, we went live with our steaming new XDP Anti-DDoS approach at our partner @globedatacenter - nice to see it in action eating 85mpps, after months of hard development work 😃

Xdp hints...