A free tool to test the load on your sites and APIs

When setting up a classic website, it's often possible to get an idea of the user load it can handle by cross-referencing information about the hosting, the web server, and the technologies used.

But in certain cases, particularly for APIs that need to process a certain amount of data or perform heavy computations, it's difficult to get an estimate of the sustainable load without running a test.

A stress test therefore involves continuously sending requests to a site (or an API) until the breaking point where it's no longer able to respond.

In some cases it's possible to run a test with a simple homemade script, but if the requests need to contain substantial payloads, chances are that your machine won't be able to keep up the pace and therefore will never reach the breaking point.

That's why I went looking for a basic tool, quick to pick up, and above all one that wouldn't cost me an arm and a leg to run a few load tests on my APIs, and I found loader.io

The advantages

I was lucky enough to stumble upon a tool that suited me without having to try out hundreds of them, but that also means I don't have much of a basis for comparison.

Nevertheless, I was able to identify a few specific features of loader.io that I particularly appreciate and that made me want to keep using it more regularly, here they are:

• A free plan with no limit on the number of tests (but a limit of 10,000 clients for 1 minute per test)
• The tool is powered by the company SendGrid, a marketing and transactional email sending platform with enough server power for a tool like this one
• Requests are sent immediately, there's no queue for tests to start.

How it works

As I mentioned, the tool is very basic, both in its use and in its interface. The first step is to confirm that you own the domain on which you're going to run your tests.

Validation helps avoid any risk of the platform being hijacked to create a DoS attack!

To do this, Loader gives you the option of either DNS validation, or validation by creating a specific route on your server. I chose this second option, since I run my API directly using the URL provided by Heroku, for which I don't have access to the DNS.

Validation takes just a few minutes, and from that point on you'll be able to create your first test!

Once the test is created, all that's left is to configure it with your url, the HTTP options, the request body, etc...

But most importantly the desired number of clients and test duration, limited in the free version but more than enough to test your API's "bottlenecks".

A "bottleneck", or "goulot d'étranglement" in French, is a point in the software that can slow down (or even stop) the program's execution, often due to too long a processing time for the information sent.

Once your test is configured, all you have to do is run it and watch the results live thanks to the various graphs and visualizations of the API's responses.

When it's finished, you'll have access to a more detailed breakdown of the statistics such as the error rate, the average response payload, the average response time, etc...

One noteworthy little detail: it's possible to set an error rate limit beyond which the test will be aborted to avoid needlessly overloading your server, likely due to an issue in the code or in the test configuration.