Originally published on Talk::Overflow
In Talk::Overflow #22, I've rounded up some fantastic talks from PromCon EU 2025.
Here are three highlights:
Julius Volz on Why I Recommend Native Prometheus Instrumentation over OpenTelemetry: Native Prometheus remains superior for most Kubernetes environments, with OTel SDKs introducing performance overhead and cognitive load that may not solve your actual reliability problems.
Jesus Vazquez, Alan Protasio, Michael Hoffmann on Beyond TSDB: Unlocking Prometheus with Parquet for Modern Scale: Shifting to Apache Parquet as a long-term storage format enables massive analytical queries at petabyte scale without crashing your TSDB or blowing your budget.
Bryan Boreham on CPU Hardware Counters: Surfacing hardware-level metrics like cache misses and branch mispredictions into Prometheus finds performance bottlenecks that application-level profiling completely misses.
The conference marked a pragmatic turn—away from complexity, back to performance-first engineering. Worth a read if you're managing observability at scale.
https://talkoverflow.substack.com/p/talkoverflow-22-promcon-eu-2025
Top comments (0)