The developers have instrumented the applications with OpenTelemetry — great! But that doesn't mean you're ready to roll it out in production yet. What do you need to keep in mind for your instrumentation infrastructure?

* Quick OpenTelemetry overview.
* Tradeoffs between the three architectures you use with OTel (depending on your vendor): vendor exporter vs OTel Collector vs OTel protocol support
* Sampling, including head vs tail based, and how to keep it representative and / or useful.

Target Audience: Architects, Developers
Prerequisites: Software development knowledge and some monitoring experience is also helpful
Level: Advanced

Extended Abstract:
The developers have instrumented the applications with OpenTelemetry — great! But that doesn't mean you're ready to roll it out in production. What do you need to keep in mind for your instrumentation infrastructure? OpenTelemetry is THE emerging standard for monitoring and creating more observable systems. Starting with an overview, this talk then dives into two areas that make a big difference in cost, flexibility, and insights:

Firstly, the three possible integration architectures with vendor exporter, OpenTelemetry Collector, and native protocol support; and why a combination of Collector and native protocol are the most common choice today.
Secondly, sampling or how to get the big picture from a subset of the data (and cost). Here the tradeoffs evolve around head- vs tail-based sampling and how to keep the collected data representative and / or useful. It generally comes down to simpler and cheaper with the head-based approach, while the tail-based one is potentially more useful with higher overhead. And now it's time to roll it out in production!