Andreas Strikos, GitHub
As we are in an AI era where demand for computation increases and traffic/load increases rapidly, performance and efficiency are more important that ever. This talk presents a systematic exploration of how CPU utilization impacts system performance, grounded in a controlled experiment. We designed and executed a series of tests on a production-grade workload to directly measure CPU utilization on latency and throughput. The talk culminates in the derivation and validation of a practical formula ("the golden formula") that connects CPU utilization, request latency, and system capacity. Attendees will leave with concrete, actionable insights for capacity planning and CPU-bound workload optimization, based on real data and reproducible methods.
This presentation will be based on previous public blog post https://github.blog/engineering/architecture-optimization/breaking-down-cpu-speed-how-utilization-impacts-performance/
Andreas Strikos is a seasoned software and platform engineer with almost 20 years of diverse experience. His professional journey has taken him from the heart of non-profit organizations to the dynamic landscapes of large corporations, where he has seamlessly navigated through various engineering roles. Currently he is part of the Performance Engineering team at GitHub.
author = {Andreas Strikos},
title = {{CPU} Utilization: The Hidden Cost of Running Hot},
year = {2025},
address = {Dublin},
publisher = {USENIX Association},
month = oct
}