TechPrufer
Get started
Datadog
Datadog logoDatadog · Feature trail

Akka ForkJoinPool Optimization

1 post · Sep 2021

What this trail covers

Datadog is optimizing its Akka applications by analyzing and addressing performance bottlenecks related to the ForkJoinPool. This involves understanding how the pool dynamically manages threads and identifying scenarios where irregular task flow combined with high thread counts can lead to excessive CPU usage due to frequent thread parking and unparking. Solutions include reconfiguring dispatchers and limiting actor instances to stabilize load and reduce overhead.

All Datadog trails
  1. How we optimized our Akka application using Datadog’s Continuous Profiler | DatadogThis post details the discovery and resolution of a performance issue in an Akka application where the ForkJoinPool was consuming 20-30% of CPU. The root cause was identified as an irregular flow of tasks combined with a high maximum number of threads, leading to frequent `Unsafe.park()/unpark()` calls. The fix involved moving the `LatencyReportActor` from the default Akka dispatcher to a 'work' dispatcher with a stable task flow and a reduced number of threads, resulting in a 30% drop in overall CPU usage.