Ad posted on LinkedIn messages; what a terrible idea. The content generally is terrible with all the self proclaimed guru's, but this increases the level of spam even further.

The free PDF version is now available 🎉 github.com/dendibakh/perf…

New term: "local coordinated omission" - github.com/artilleryio/ar…

A load generator can only properly deal with coordinated omission, if the latency is determined based on the scheduled starting time and not the actual starting time.

LinkedIn: cesspool of professional opinions from people I'm not connected to and don't care for.

debugTheDebugger redd.it/1iquql5

A nice thing about working at Adaptive is that you get to see presentations long before the general public can see them.

In Java, when final will mean final, it will cause networking code to create litter. Because the Selector hack to replace the litter producing hashset by something that doesn't produce litter, won't work any longer. Netty and Aeron are some projects that apply this trick.

Panama uring Java This is an exploratory project that uses Java's new FFI mechanism to introduce io_uring to Java. github.com/dreamlike-ocea…

It has been more than a decade I have used aspect-oriented programming, but with ByteBuddy it is pretty sweet. Working on an Aeron async-profiler agent for latency profiling and an Aeron HdrHistogram agent for latency tracking over time.

Software engineers should not be concerned about AI replacing their jobs (for now) unless they are performing extremely trivial tasks. I have paid accounts for ChatGPT and Claude, and although it can be useful, AI is typically on massive amounts of LSD.

The more I work with AI (both ChatGPT and Claude) the less I believe engineers should be worried to be replaced unless they are doing extremely trivial work.

A nice opportunity to play with Zig and eBPF. github.com/tw4452852/zbpf

For the first time in history $AMD has reached a 50% market share in server CPUs It marks the end of an era for $INTC

Prompt engineer 1: it doesn't work 2: AI: I understand the problem now, let me fix it. 3: goto 1

Huge thanks to Peter (Performance Engineer from Adaptive) for spotting a coordinated omission issue in my benchmark. His feedback was spot on. (See Gil Tene's 'How NOT to Measure Latency' for a deep dive on this). My original test showed peak throughput. This CO-corrected

anotherToughDayAtWork redd.it/1n287ff

Why does installing perf under Ubuntu need to be such a huge pain.