Creating a large data set

The first step of performance testing is getting a large enough data set on an environment similar to the production one. For OpenIMIS, we are doing:

• Generating data: the backend has a few commands like “generateinsurees” and “generateclaims” that will generate large amounts of random data with the help of the Faker library. This allows the use of localized data (names, phone numbers…)

• Retrieve an existing data set, preferably anonymized: this approach was used to investigate the performance with the database of Nepal

• Generate real-life traffic through the regular API: this is achieved by using Locust.io

Locust load testing

Locust.io is a open source load testing tool using Python. We created some helper methods for authentication and some service calls. From there, it is rather easy to create test scenarios.

Repository home: https://github.com/openimis/openimis-loadtest

The beginning of the script provides a few configuration parameters for the ready-made scenarios:

Open

Then come a few functions that encapsulate a single action and annotated with @task. These will be shown in the report with their performance metrics.

The on_start() method performs the necessary authentication and supports both hybrid mode and full modular with JWT. (This allows the testing of older versions)