Load Testing

Load testing can be done against your instance at any time by running the included load-test.js script, which uses k6 to submit requests with the given parameters and provide you a report with the results at the end.

Ideally, you will want to also want to have a connection to your server while you are running these tests so that you can monitor them for CPU load (using a tool like htop) to see what combination of parameters is the limit your instance can handle. For example, your server might start to consistently see 100% CPU usage and return failed request responses with slow response times when you get past 200 users for the search page URL, but the same 200 users might be fine for the home page URL. This can help you focus your efforts on what pages or areas of your infrastructure to optimize and know what limits you can expect of your servers.

The script itself has instructions for using it, but as a quick example, you could quickly run these tests locally within a Docker container:

# Queries the catalog home page with 100 users over a 1 minute duration
docker run --rm -v /path/to/load_test.js:/load_test.js grafana/k6 run -u 100 -d 1m /load_test.js --env URL="https://catalog.lib.msu.edu"

Benchmarking

Another tool you can use if you don't want to test your site over a period of time is the Apache benchmarking tool which is built into this VuFind image and is part of most Ubuntu installations.

# Request the URL 100 times, doing 10 concurrent requests
ab -l -n 100 -c 10 "https://catalog.lib.msu.edu/"

The advantage with this output is it is much simpler, so it can be quicker to get a high level picture of performance gains when comparing the before-and-after of a code change.