Performance Hunting
Christian and Jens took part in CodeCamp 2025 and set out on the hunt for performance improvements. Instead of rifles, they were armed with two laptops and used the metaphor of hunting season. 🦌🍁🍂
To get a good overview, we installed the Jaeger-Tracing tool, from our “ raised hide.” This allowed us to detect all sneaky performance gaps. We supplemented our the core with numerous of spans, that collected and recorded all requests. The clear Jaeger tool helped us identify many potential improvements after only a short time. 🕶
However, implementing optimization proved to be tricky because each optimization had further effects on the engine's behavior. For example: Parallelizing threads was very beneficial for us. But here we also had to pay attention to the order of the parallel threads, as some of them were dependent on each other.
a Screnshot from a Engine StartUp in Jaeger-Tracing.
Our successes
- We have created a basis for identifying bottlenecks early on in the future. This can be useful for the further development of the platform, as it quickly provides an overview of where bottlenecks are hidden.
- We were already able to achieve a 10-15% performance improvement only after the CodeCamp.
-Further bottlenecks were identified, which we will continue to improve in the near future.
Teamwork
But the most important thing, however, was the great teamwork. We had a lot of fun, complemented each other well, and thanks to Christian's extensive know-how, I personally benefited enormously and learned a lot.
Cheers and "Weidmannsheil"
Christian & Jens
