Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a detailed case study presentation from SREcon25 Americas where Nicolas Arroyo from Bloomberg LP investigates an unexpected variant of the "thundering herd problem" in the wild. Learn about this performance issue where multiple threads wake up simultaneously to handle the same event, causing resource waste through unnecessary context switches. Follow the systematic performance troubleshooting process, beginning with aggregated sampling, progressing through dynamic instrumentation and detailed sampling, and culminating with kernel mode sampling. Discover how the investigation revealed system calls hidden within common libraries that use absolute timers, causing thread synchronization and simultaneous thread awakening. This 35-minute technical talk provides valuable insights for SREs and systems engineers dealing with complex performance problems.
