Performance overhead of the Elastic Distribution of OpenTelemetry Python
Elastic Cloud Serverless projects are autoscaled environments, fully managed by Elastic and available on Elastic Cloud. Availability This functionality is generally available and ready for production usage. The Elastic Stack includes Elastic's core products such as Elasticsearch, Kibana, Logstash, and Beats. Availability This functionality is generally available and ready for production usage. Unless stated otherwise on the page, this functionality is available when your Elastic Stack is deployed on Elastic Cloud Hosted, Elastic Cloud Enterprise, Elastic Cloud on Kubernetes, and self-managed environments. This documentation corresponds to the latest patch available for each minor version. If you're not using the latest patch, check the release notes for changes. The Elastic Distribution of OpenTelemetry (EDOT) Python SDK collects performance metrics, traces, and logs in OpenTelemetry format, and sends them to Elastic Observability. Availability This functionality is generally available and ready for production usage. This documentation corresponds to the latest patch available for each minor version. If you're not using the latest patch, check the release notes for changes.
Serverless Observability
Generally available
Stack
Generally available
EDOT Python
Generally available
This page explains the performance considerations when instrumenting Python applications with the Elastic Distribution of OpenTelemetry SDK, including impact analysis and mitigation techniques.
While designed to have minimal performance overhead, the EDOT Python agent, like any instrumentation agent, executes within the application process and thus has a small influence on the application performance.
This performance overhead depends on the application's technical architecture, its configuration and environment, and the load. These factors are not easy to reproduce on their own, and all applications are different, so it is not possible to provide a simple answer.
The following numbers are only provided as indicators, and you should not attempt to extrapolate them. Use them as a framework to evaluate and measure the overhead on your applications.
The following table compares the response times of a sample web application without an agent, with Elastic APM Python Agent and with EDOT Python Agent in two situations: without data loaded and serialized to measure the minimal overhead of agents and with some data loaded and then serialized to provide a more common scenario.
| No agent | EDOT Python agent | Elastic APM Python agent | |
|---|---|---|---|
| No data: Time taken for tests | 1.277 s | 2.215 s | 2.313 s |
| Sample data: Time taken for tests | 4.546 s | 6.401 s | 6.159 s |