HTTP response time measurements

This data is provided by beebyte, see below for more information.



Data source: https://beebyte.wppt.se/data/




Data source: https://hetzner.wppt.se/data/




What is this?

This webpage displays results from HTTP response time measurements from identical WordPress websites deployed to different hosting providers. All test websites use Cloudflare to try to minimize the effects of network latency on the measurements.

But why?

We wanted to get an idea of response times for different Swedish webhosting providers. As we are collecting the data anyway, we might as well make it public.

The test webpage

The test website is a standard WordPress install with WooCommerce and a singel product. The website start page, which is the URL being tested, includes WooCommerce products.
For the tested websites we have used PHP 7.4 where available, or the nearest PHP version available if not.

How are the measurements performed

Each website is tested every 5 minutes. A HTTPS connection is established with the website using HTTP keepalive. The monitored page is fetched five times using a standard HTTP GET request and the time from when the request is sent until all data has been fetched is saved. An average response time value of the five requests is stored in a database.
Only the website index page is measured, no static data (images, css, js etc) are fetched for the page as there generelly isn't much difference in serving up static data.

So what is it you want to measure?

The goal is to measure how fast the tested webserver can return the HTML data for a WordPress (PHP) website. This includes PHP processing (CPU/disk), and DB access. No caching is used on the tested websites as that would basically make the test a network performance test when the goal is to measure webserver performance.
Once per hour, the past 12 values (from the past hour) are once again averaged and stored, these are the values that are displayed in the graphs.

In effect, the graphs display highly averaged response time data to try to give a reasonable picture of what each monitored URLs response times look like, without being to influenced by temporary highs or lows.

A note on reading the results

Differences in response time will be affected by network latency to some degree, and that will, in the real world, vary based on the visitors location and ISP. A difference of 5ms or less in the graphs can probably be discounted as network related.

Why are there two graphs?

To help highlight the impact of network related delays, the same test is performed from multiple locations. As the tested websites are all behind Cloudflare, this mostly displays the difference in network latency from different locations.

What webhosting packages have been used?

Standard webhosting packages from each provider have been used, trying to stay within the same price range as far as possible.

Why aren't you testing more providers?

Cost.

Who is performing this test?

This test is performed by beebyte. As a hosting provider we are obviously biased in a test like this, we have therefore tried to make the test as honest, transparent and reproducible as possible. As the tests are using publicy available resources, they can be reproduced by anyone that sets up accounts/websites with the tested providers.

What tools are used to perform the testing?

The tools used and the source for this website can be found on github.

Conclusion

Webpage response times are one important factor when selecting a hosting provider, but it is far from the only factor. Tooling, support, easy of use, price etc. are all factors that should be taken into account when selecting a hosting provider that fits your particular needs.