This repository contains a set of benchmarks to assess the performance of the OpenTelemetry instrumentation library against the OpenTracing library.
List of current tests:
| Test | No Instrumentation | Jaeger Trace Exporter | Otlp Trace Exporter |
|---|---|---|---|
| Throughput | ✓ | ✓ | ✓ |
| Sample Time | ✓ | ✓ | ✓ |
Each benchmark contains common packages to the same purpose:
opentelemetry-benchmark-spring-cloud
│
│
└───io.opentelemetry.benchmark
│ │
│ └───config
│ │ ## Classes to initialize the different tracers
│ │
│ └───main
│ │ ## The class to be executed by JMH
│ │
│ └───petclinic
│ │ ## Depending on the test, there is one package which contains the example application to be executed by JMH.
│ │ In this example for opentelemetry for spring cloud, the official petclinic application is used for the tests.
│ │
│ Benchmark*.java
│ ## The different JMH benchmark tests are located in this package
│ │
Some tests additionally use the OpenTelemetry javaagent for automatic instrumentation.
This properties can be overrided by system properties:
benchmark.warmup.iterations=5
benchmark.test.iterations=5
benchmark.test.forks=1
benchmark.test.threads=1
benchmark.global.testclassregexpattern=.*Benchmark.*
benchmark.global.resultfileprefix=jmh-
Run the following command (inside on a specific project)
mvn clean install
java -jar target/benchmarks.jar
The tests can also be executed on a bare metal server using an automated script. Kindly see runner for more details on running the project.
Performance tests are tricky and there are additional cases that we wish to cover. Feedback will be highly appreciated.
We use JMH-visualizer to present the benchmark results. Thanks jzillmann for this amazing tool!.
The results of different scenarios are located in each project: