Latte is a simple fixed-throughput latency testing framework that embraces open system model. It is based on k6.
Download docker-compose.yml in the releases tab. The docker-compose version should be >= 1.28.
$ docker-compose --profile worker up
$ docker-compose --profile controller up
When the server is up, register the worker URL (http://<worker addr>:8081) and user info.
-
Configuration
A worker should be configured to run a test. Register or update the worker in the
SETTINGStab (skip if configured properly during installation). -
Writing a test script
Latte uses k6 internally, so you need to write k6 scripts for testing. The scripts are maintained in the controller and are accessible through git (please check
SCRIPTStab on the web console). You can checkscripts/DemoTest.jsfrom the initial git repository to see how the script is written. Please refer to the k6 scripting guide for more usage. -
Running a test
After writing a test script, you can run the test in the
RUNtab. The output of the test, which is the k6 standard output, is displayed at theRUNtab. When the testing is done, the latency distribution of the test is recorded and provided through theHISTORYtab. For performance reasons, only a single test can be run at a time.
Controller node
├── latte-controller-gui: Frontend server
├── latte-controller-api: Backend server
└── latte-controller-git: HTTP git apache server
Worker node
└── latte-worker
Latte is composed of two nodes: controller and worker.
The system state is managed in the controller.
latte-controller-api maintains the history of previous tests, forwards requests to the worker and returns responses back to the frontend server.
It also caches the responses and they are streamed to any users that open the RUN tab.
Worker is where the actual test is run. The standard output of the test is streamed back to the controller node. Currently only a single worker can be attached to controller.