Brendan Gregg

Makes things fast. Expert in computer performance and eBPF. OpenAI, ex-Netflix. Created flame graphs. Aussie. Opinions my own. [email protected]

[Brendan Gregg @brendangregg]

Announcing my next book: Systems Performance: Enterprise and the Cloud, 2nd Edition (2020) https://t.co/x9AwlJ1Rdb

[Brendan Gregg @brendangregg]

Huge news for Doctor Who fans: DMP 1 & 3 found! https://t.co/jOHuQMSyC5

[Greg Brockman @gdb]

excited to work with Brendan, have been a big fan of his work for many years!

[Brendan Gregg @brendangregg]

New blog post: Why I joined OpenAI https://t.co/xwcPrOA5SR

[Brendan Gregg @brendangregg]

New post: Leaving Intel https://t.co/rkcGw8VdyU

[Brendan Gregg @brendangregg]

new post: When to Hire a Computer Performance Engineering Team (2025) part 1 of 2 https://t.co/BYP9ry06C4

[Brendan Gregg @brendangregg]

Had a great time at #srenext Tokyo

[P山 @pyama86]

今日もいい日だった #srenext

[Brendan Gregg @brendangregg]

@cperciva I guess I'm lucky; I've also been sick & WFH-ok many times, 10 in 9 years is just "days off sick."

[Brendan Gregg @brendangregg]

new blog post: 3 Years of Extremely Remote Work https://t.co/4fD0rjjnjw

[Brendan Gregg @brendangregg]

New blog post: Doom GPU Flame Graphs, a fun demo of AI/GPU Flame Graphs and FlameScope, now open source https://t.co/mIxKWxCyRA

[Brendan Gregg @brendangregg]

My website is back online https://t.co/1cRpD64hyP after migrating to a different provider (and https works again)

[Brendan Gregg @brendangregg]

My website is online as http at https://t.co/1cRpD64hyP, but I just broke the https links, sorry. If you delete the "s" then the links should still work.

[Brendan Gregg @brendangregg]

AI Flame Graphs: Showing what's actually running on the HW and how we got there. Uses Intel EU stall profiling and eBPF. https://t.co/zIkkBG5nly

[Brendan Gregg @brendangregg]

WSJ covering the eBPF "super revolutionary technology" https://t.co/nRQnsWmTAF

[Brendan Gregg @brendangregg]

@msherman1970 I mean that it might seem that we've accomplished nothing if we move a crash from one body of software (a security agent) to another (eBPF code). It's a point people make as well: "oh, but eBPF has had panics too." Yes it has, but it's still an overall improvement.

[Brendan Gregg @brendangregg]

New post: No More Blue Fridays: How eBPF is already being adopted to prevent kernel crashes https://t.co/pufMvRc4UB

[Brendan Gregg @brendangregg]

@senolcolaks Thanks, I was wondering why it wasn't already on my list then saw it's provided by the systemd package, so maybe I've never had to add it

[Brendan Gregg @brendangregg]

New blog post: Linux crisis tools https://t.co/uTAXexoQCM

[Brendan Gregg @brendangregg]

@CarlMastrangelo Yes by "perf trace" and the bpf tracers, which use ring buffers. Current strace is dangerous for production use (ok as a last resort) https://t.co/uCNxPsMHvf

[Brendan Gregg @brendangregg]

@indygreg Some have switched already (golang) others have options. It'd be nice for it to be default, but it's less of a problem than the OS, because (A) profiling becomes obviously broken and developers go hunting for the fix, and (B) easier to add runtime options than OS changes.

[Brendan Gregg @brendangregg]

New blog post: The Return of the Frame Pointers (Fedora, Ubuntu) https://t.co/7JPatBZpIf

[Brendan Gregg @brendangregg]

@iiirogers Your post was super helpful, thanks! https://t.co/clCDAgw04P