Determine the number of users - Distributed Load Testing on AWS

Determine the number of users

The number of users a container can support for a test can be determined by gradually increasing the number of users, and monitoring performance in Amazon CloudWatch. Once you observe that CPU and memory performance are approaching their limits, you’ve reached the maximum number of users a container can support for that test in their default configuration (2 vCPU and 4 GB of memory). You can begin determining the concurrent user limits for your test by using the following example:

  1. Create a test with no more than 200 users.

  2. While the test runs, monitor the CPU and Memory using the CloudWatch console:

    1. From the left navigation pane, under Container Insights, select Performance Monitoring.

    2. On the Performance monitoring page, from the left drop down menu, select ECS Clusters.

    3. From the right drop down menu, select your Amazon Elastic Container Service (Amazon ECS) cluster.

  3. While monitoring, watch the CPU and Memory. If the CPU does not surpass 75% or the Memory does not surpass 85% (ignore one time peaks), you can run another test with a higher number of users.

Repeat steps 1-3 if the test did not exceed the resource limits. Optionally, the containers resources can be increased to allow for a higher number of concurrent users. However, this results in a higher cost. For details, refer to the Increase the container resources section of this guide.

Note

For accurate results, run only one test at a time when determining concurrent user limits. All tests use the same cluster and CloudWatch container insights aggregates the performance data based on cluster. This causes both tests to be reported to CloudWatch container insights simultaneously, which results in inaccurate resource utilization metrics for a single test.

For more information on calibrating users per engine, refer to Calibrating a Taurus Test in the BlazeMeter documentation.