Load testing of aviobook application in AWS

Customer need

Aviovision was already active in the European market, focusing on cargo flights with a volume of around 30,000 flights per week. With their planned rollout to the US market, this number would double to 60,000 flights per week.

This raised one key question for Aviovision: can the platform handle a doubling of the volume? Can the rollout to the US market happen without issues for their own application? They needed a performance test set capable of monitoring how the platform behaves under increased load.

There was also a need for automated performance tests. Aviovision was not used to running regular load tests on their platform, which meant performance bottlenecks in new features were identified late.

At the same time, this assignment created an opportunity for a historical catch-up of the application. The Aviobook application had already been live for two years, making it the ideal moment to uncover and address any potential future issues in the platform.

Refleqt solution + deliverables

Refleqt’s solution consisted of several elements:

Technical analysis

Based on a technical analysis, we mapped the API landscape of the app and the admin website. This gave us clear insight into the network traffic and allowed us to understand the Aviovision application in detail. With this detailed view of the platform, Refleqt was able to quickly design the right performance test set. After the analysis, Refleqt shared targeted recommendations and potential optimizations with the development team.

Performance testing

To set up the performance tests, we used the OpenAPI documentation together with an open-source Scala template. Combined with Swagger Codegen, we generated a Scala library as an artifact. This library was imported into Gatling to run the required load tests. A load generator was then configured on AWS using Docker.

For logging and monitoring during the performance tests, Gatling was connected to a Grafana dashboard via InfluxDB. This made it possible to visualize real-time results during load testing so that potential issues or errors could be identified quickly by the development team. The dashboard was used to monitor the health of the platform and showed, in real time:

- Number of hits per minute

- Response times

- Error percentages

Execution

The performance tests were executed iteratively and in continuous collaboration with the Aviobook technical team. This way, the root causes of negative test results could always be investigated.

Tools & technologies

- Technical analysis: Charles (proxy server)

- Performance testing: Gatling, Scala, Maven, OpenAPI, Swagger Codegen, Scala Build Tool

- Execution & monitoring: Grafana, InfluxDB, Kibana, AWS, Docker

Result

Aviovision responded very positively to Refleqt’s work. The technical analysis revealed several bottlenecks in the platform, which the Aviovision development team then addressed and resolved.

Thanks to the strong results from the performance testing, the Aviobook app is now ready for increased traffic. The platform can handle a doubling of the load without expected issues.

Refleqt also fully trained Aviovision’s QA team through a technical handover, enabling them to continue using and extending the performance test setup after the engagement.