One of the main challenges in observability is that when you want to observe a specific service or monitor your service, developers have to adopt a vendor-specific agent or instrument applications using vendor-specific instrumentation APIs. Open source OpenTelemetry is trying to address this issue by commoditizing observability, so the tooling around instrumentation, collection and exporting of telemetry data is not locked to a single vendor.

In this episode, I sat down with Juraci Paixão Kröhling, Principal Engineer at Grafana Labs, to talk about the evolution of OpenTelemetry, the importance of tail sampling and the need for a balance between data collection and storage costs. Kröhling also touched upon the adoption and commercial support of OpenTelemetry, its challenges, and future developments, including improved metrics support, stable semantic conventions, and the role of AI in observability.

Observability solutions, OpenTelemetry, and vendor lock-in

• Observability has become a critical focus in today’s tech landscape, particularly as various projects and vendors continue to emerge.
• One significant challenge faced with observability is vendor lock-in. If you want to monitor a specific service, you are required to use vendor-specific agents or instrumentation APIs. Kröhling explains moving away from a particular vendor often necessitates re-instrumenting applications, redeploying them, and even re-architecting deployment pipelines.
• To address these challenges, OpenTelemetry has emerged as a key solution.
• The primary goal of OpenTelemetry is to commoditize observability by standardizing the tooling around instrumentation, collection, and exporting of telemetry data.
• OpenTelemetry evolved from two complementary projects, OpenTracing and OpenCensus, to provide a unified standard for distributed tracing, metrics, and logs.

OpenTelemetry adoption, challenges, and commercial support for Grafana labs

• Cost is a major challenge in observability, with companies paying for unused data.
• Kröhling talks about challenges and use cases for OpenTelemetry, including car telemetry data and restaurant chain data collection.
• Kröhling discusses OpenTelemetry support in Grafana cloud, including new features and commercial angle.

OpenTelemetry roadmap, LLMs, and AI observability

• Kröhling discusses OpenTelemetry’s roadmap for 2024, including stabilizing semantic conventions and improving logging support.
• OpenTelemetry aims to provide log bridges for each language it supports, enabling native OTLP logging for applications.
• Kröhling explains how observability is crucial for LLMs/generative AI, to understand how they make decisions.
• Kröhling goes on to add that AI can help generate telemetry data for research purposes, and more.

Guest: Juraci Paixão Kröhling (LinkedIn)
Company: Grafana Labs (Twitter)
Show: Let’s Talk