What is Telemetry?
For the purpose of measuring running computer software and systems, our instruments are virtual instruments. That is to say, code that measures other code. It sounds simple: read a measurement and send it to a remote location. In practice, to make that telemetry data useful in today’s cloud-native and ever more complex environments, there are huge logistical and technical hurdles to overcome. OpenTelemetry is heralding a revolution that will enable disruptions to the monitoring and observability landscape by solving these challenges in an open-source, vendor-agnostic way.
What is OpenTelemetry?
OpenTelemetry is a Cloud Native Computing Federation (CNCF) project started in 2019. The OpenTelemetry project covers collecting, processing, and transmitting telemetry data, but doesn’t handle storing, displaying, or using the data beyond collection and transmission. OpenTelemetry’s mission is “to enable effective observability by making high-quality, portable telemetry ubiquitous.”
There are three main parts to the OpenTelemetry project to make this a reality.
1. The OpenTelemetry specification
The OpenTelemetry Specification defines the cross-language data specifications including the API, SDKs, and transport protocols. These specs essentially form the new ‘lingua franca’ for Observability, allowing the full stack to speak the same language, gather data in the same format, and ultimately transport that data for processing. This is the crucial puzzle piece required in order to make telemetry truly ubiquitous.
Unified Metrics, Traces, and Logs
OpenTelemetry is defining unified specs for Metrics, Traces, and Logs. Unified is essential to unlocking the power of these separately generated measurements and events. OpenTelemetry defines standard conventions for attributes and resources associated with events. Applications and infrastructure stacks that use OpenTelemetry gain vendor agnostic distributed tracing and metric superpowers through the entire stack.
Understanding metrics, traces, and logs with this deep context connects everything.
2. OpenTelemetry APIs and SDKs
There is an API library and SDK library written natively in every major programming language which are the implementations of the OpenTelemetry specification. The API and SDK libraries have different purposes and are used by different developers and are thus deliberately separate libraries.
Historically, APM providers wrote agents that would dynamically inject code (instruments) into your application’s frameworks and libraries such as Rails, Redis, MongoDB, etc., in order to measure and collect telemetry data. Writing each APM provider’s own code for instrumenting frameworks and libraries was difficult, tedious, and delicate work, resulting in duplicated or wasted effort.
This instrumentation work is now being shared among the OpenTelemetry community instead of each APM provider. Now we can all contribute to a single instrumentation implementation for the benefit of everyone.
The long-term (and ideal) goal is to have the framework or library developers themselves use the OpenTelemetry API within their own code, making the instrumentation native to the framework or library. Instead of outside-in instrumentation injected into the frameworks and libraries, it will be baked in natively within the framework or library code.
If a library developer chooses to add native OpenTelemetry instrumentation, the only dependency required is the OpenTelemetry API. The API is designed such that even if the framework or library is instrumented with OpenTelemetry, it will not require the end-user application to make any changes if they don’t want to enable OpenTelemetry instrumentation. This avoids having “instrumented” and “non-instrumented” versions of frameworks and libraries. Finally, the library developers won’t know or care what the specific OpenTelemetry implementation or configuration is within the end-user application. This provides a strong stability guarantee to the library developers which should alleviate concerns about implementing OpenTelemetry within their library.
If you are an application developer, the SDK library is what you will be using to configure OpenTelemetry within your application. This is how you will select which libraries to collect telemetry from, how to handle and enrich that data, and how to export or transmit the data.
Exporters are transportation protocol specific libraries that integrate with the OpenTelemetry SDKs. These allow you to send the telemetry data to different backend systems that speak different protocols, such as Jaeger or Zipkin. OpenTelemetry also has its own protocol: OpenTelemetry Line Protocol, or OTLP for short. OTLP utilizes Protobuf to encode the payloads, and utilizes HTTP/S or gRPC for the transport protocol. Every language with an OpenTelemetry SDK also has an OTLP exporter.
3. OpenTelemetry Collector
OpenTelemetry Collector is an application written in “Go.” The GitHub readme does an excellent job of describing the collector:
The collector offers a vendor-agnostic implementation on how to receive, process and export telemetry data. In addition, it removes the need to run, operate and maintain multiple agents/collectors in order to support open-source telemetry data formats (e.g. Jaeger, Prometheus, etc.) sending to multiple open-source or commercial back-ends.
Usable: Reasonable default configuration, supports popular protocols, runs and collects out of the box.
Performant: Highly stable and performant under varying loads and configurations.
Observable: An exemplar of an observable service.
Extensible: Customizable without touching the core code.
Unified: Single codebase, deployable as an agent or collector with support for traces, metrics and logs.
The simplest way to export data is to a backend via an OTLP exporter is to a collector. The collector can then process, transform, and relay the data to another endpoint upstream.
OpenTelemetry is leveling the playing field with regard to collecting, processing, and transmitting telemetry data. This focuses the value prop of Monitoring and Observability providers onto what insights and value you can derive from unified data.
It’s a turning point in monitoring and observability, and the next few years will be exciting.