8

I am trying to improve the speed of a site - however when I go to Pingdom to test: http://tools.pingdom.com/fpt/

I get wildly different times even though I haven't made any changes. For example:

  • 8.92s
  • 3.68s
  • 11.40s

Also - I get the same issue with gtmetrix (http://gtmetrix.com/).

As a result I can't tell if the performance changes I have made make a difference.

So - how does one test for site speed?

  • The sites you're currently using to test are fine. It could be simply due to a poor hosting environment. If you're on a cheap host, then that might be the issue as they're usually cheap for a reason – Lodder Mar 12 '15 at 15:53
  • +1 I sometimes wonder about the accuracy of these sites too. @Lodder can you suggest any alternatives or perhaps a different approach to testing Joomla websites for speed? – jonboy Mar 12 '15 at 16:55
  • Thanks - the site is on siteground cloud and has 2 sites hosted on it: siteground.com/cloud-hosting.htm - so I'm hoping its not the host.I'll send them an email. – iamrobert Mar 12 '15 at 17:07
  • Ahh ok Siteground, they're brilliant so can't be a hosting related issue. Try using Webpage test and Google page Speed to test the speeds too, to see if any differences are shown. There are quite a few but I would only tend to use the more popular ones. If you're using any caching extensions/settings, try disabling them and run the tests again. The re-enable them and re-run the tests – Lodder Mar 12 '15 at 17:09
  • Siteground: "Please note that Pingdom is an external service and we can not specify why these tools return different results." – iamrobert Mar 12 '15 at 17:45
5

There are a few factors beyond your control such as the load on the server at the time and the available bandwidth between the test tool server and the target website at the time the test is run.

You can minimise the variability of the bandwidth by creating an account (a free account is fine) on GTMetrix and logging in so you can use the advanced settings and choose a test server location close to where your target website is hosted.

Pingdom may have a similar option.

You will probably still need to rerun the test a few times and take an average but this will likely yield a more consistent result.

  • Thanks - Neil - do you know if GTmetrix cache your site - so it will be faster next time? – iamrobert Mar 13 '15 at 16:51
  • Yes, subsequent tests do seem to run a little faster. – Neil Robertson Mar 13 '15 at 21:23
  • 1
    I emailed them: "We use a fresh profile on every test, so nothing gets cached on our side except for DNS. Be sure to look at your waterfall graphs (Timeline tab) to see where the time differences lie. It could be your server (slow SQL queries that are cached after the first run), or it could be an external widget that's loading inconsistently." – iamrobert Mar 15 '15 at 10:43
  • Thanks Neil - I think this is the best solution. From my tests - gtmetrix with an account seems to be the most consistent - giving me similar results each time. However - I will also check with my google analytics stats to see load time and pingdom. – iamrobert Mar 15 '15 at 10:46
4

Using third party tool like pingdom, gmetrix and even Google page speed it is obvious to get variation in result. There are many factors on that page speed depends.

For example: If broser cache is enabled on site then very first check with any of this service will take longer time and after other consecutive tries it may load faster as the page is already cached.

With my own site http://saur.in (under development)
First attempt on pingdom: 1.83s
Second attempt: 786ms

That is almost haft in time and more than double the page speed. The same way server cache plays its role too.

There are other factors like domain name resolve, dns connectivity current traffic on your site and many more.. so result may vary. While working with performance we have to consider average response time or page load time.

At last, one more thing, I should add here:

Create a set of results from 2-3 services and check performance after every 2-3 hours. Pingdom provides testing for every minute and its free... you can try that. Try if for 24 hrs and look average result of it.

4

There are some tools used in the industry for "stressing" websites. They are rather complex tools and NOT appropriate for small websites or shared host environments.

To do comparison tests, you need to have your own server (to be able to control the infrastructure), so that you can control the load. Testing a shared host does not really provide any comparable data.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.