Test Metrics
We conducted a load test to ensure our system can handle a large number of users, targeting a user base of 50k-100k at this stage. Currently, we have over 5k active users and we are planning to scale up to 50k+ users in the upcoming months.
We used the Locust load testing tool to simulate 100, 300, 500, and 800 concurrent users accessing our system simultaneously and continuously making requests. Our server configuration for this test was a small EC2 instance.
Our app does not have any features that require a large number of concurrent user connections. We have well-distributed traffic throughout the day based on our data from the last 6 months. We are also planning to implement a caching mechanism to reduce the load on the server, and add offline data support in our app to further reduce the server load.
In summary, we are confident that our system can handle the expected user base of 50k+ users with the current server configuration and planned optimizations.
Here are the results of the load test:
- Load test with 100 concurrent users:
We used the following configuration: 100 concurrent max users, and a hatch rate of 2 users per second.
Graphs:
- Load test with 300 concurrent users:
We used the following configuration: 300 max concurrent users, and a hatch rate of 6 users per second.
Graphs:
- Load test with 800 concurrent users:
We used the following configuration: 800 max concurrent users, and a hatch rate of 16 users per second.
Graphs:
With the above results, when we tested with 800 concurrent users, we started seeing failures in the requests and the response time increased significantly, which is not acceptable.
Conclusion
After conducting multiple tests, we found that our system can handle up to 300 concurrent users with the current server configuration. By analyzing our usage patterns from the past few months and our concurrent numbers with a 5k user base during peak hours, we are confident that our system can handle 50k+ users with the current server configuration and planned optimizations.