Dark Roast
k_sched_lat
A high-precision kernel latency monitoring module designed for tracking task wake-up and runqueue latencies across multi-socket CPU topologies.
I am a Linux Kernel Developer at IBM, based in Limerick, Ireland. I focus on the core Linux Kernel Scheduler—tuning thread placements, load-balancing pathways, and low-level performance in C and Rust.
Hello! I'm Euan, a systems programmer based in Limerick, Ireland. Currently, I work as a Linux Kernel Developer at IBM, focusing on the core Linux Kernel Scheduler (`kernel/sched/`).
Much like brewing the perfect shot of espresso, task scheduling on complex multi-core processors requires strict control, rigorous precision, and an appreciation for physical topology. I spend my days debugging lock-free queues, mapping scheduling domains across physical cache boundaries, and tuning load-balancing routines to maximize system throughput and minimize task wake-up latency.
My design and coding philosophy draws heavily from Zen principles and Japanese minimalism: eliminating unnecessary abstractions, designing clear pathways, and ensuring every single CPU instruction and memory access serves a clean, high-performance purpose.
When I'm not writing driver paths or reviewing scheduler patches, you can find me experimenting with pour-over coffee techniques, collecting craft beans from local roasters, or admiring clean, structured architectural layouts.
A high-precision kernel latency monitoring module designed for tracking task wake-up and runqueue latencies across multi-socket CPU topologies.
A high-performance, lock-free custom memory allocator written in Rust, leveraging zero-cost abstractions and local caching strategies.
An interactive monospaced visualization of CPU scheduler load-balancing migrations and runqueues, displaying active cores and task domains.
Investigating unexpected task migration latency across CPU sockets, tracing load-balancing routines, and analyzing cache penalties in nested domains.
Read ArticleHow the Linux scheduler models physical hardware structures, maps NUMA clusters, and makes cache-aware scheduling decisions.
Read ArticleAn exploration of writing safe low-level interfaces for hardware subsystems, testing safety models directly against traditional POSIX C APIs.
Read ArticleHave a low-level systems project in mind, want to discuss CPU scheduling topology, talk performance micro-optimizations, or simply wish to share your favorite pour-over recipe? I'm always open to discussing new opportunities and interesting ideas.