This content is likely not relevant anymore. Try searching or browse recent questions.
Where is Google's PageSpeed Insights servers hosted? Is Africa prejudiced?
I note from the data centre map below that Google doesn't have a single data centre in Africa.
From doing various network tests from Cape Town, South Africa, I can attest that there is a minimum of 160 ms of the submarine cables to London. From there to NY and LA it just gets worse.
Although Google might have a cache in Johannesburg, South Africa, I'm highly sceptical if Google PageSpeed Insights does it tests from here.
The other glaring discrepancy is that when I compare PageSpeed Insight with GTMetrix I get exactly the same discrepancies that one would expect for distance based testing. So, for example:
The other glaring discrepancy is that when I compare PageSpeed Insight with GTMetrix I get exactly the same discrepancies that one would expect for distance based testing. So, for example:
- Maximum distance. 300+ ms. From GTMetrix Vancouver, around 3 to 4 seconds.
- Google PageSpeed Insights, ~5 seconds
- GTMetrix London, <2 seconds
- Google PageSpeed Insights, ~5 seconds
- GTMetrix London, <2 seconds
So we have +- 2, 4, and 6 seconds to works with.
This is also very evident on our GSC "Core Web Vitals". Even though our sites loads in less than 2 seconds from London, on PageSpeed Insights we never see anything below 5 or 6.
So the question is, in which DC is this service hosted? How does it compensate for Europe? Does it compensate for Africa, e.g. South Africa?
This is also very evident on our GSC "Core Web Vitals". Even though our sites loads in less than 2 seconds from London, on PageSpeed Insights we never see anything below 5 or 6.
So the question is, in which DC is this service hosted? How does it compensate for Europe? Does it compensate for Africa, e.g. South Africa?
Thank you for your expert replies.
Latest Update Latest Updates (0)
Recommended Answer Recommended Answers (1)
Recommended Answer
Relevant Answer
Hi Eugène van der Merwe
PSI does mostly seem to be us based. A good example of why it should only be part of your testing toolset, along perhaps with the potentially more useful real world user measurements.
Running lighthouse locally can also be a good addition to the arsenal too, that naturally overcomes the distance issues by running from your local machine.
But the ones that will matter as part of the page experience ranking signal are from CrUX data. This is real world data, as in the load times actual users of your sites experienced, so by it's nature this overcomes that distance issue, it's the actual experience site visitors are getting.
If you are interested in collecting rum data yourself, this is a handy library to help: https://github.com/GoogleChrome/web-vitals/
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Relevant Answer Relevant Answers (0)
Our automated system analyzes the replies to choose the one that's most likely to answer the question.
All Replies (6)
Recommended Answer
Relevant Answer
Hi Eugène van der Merwe
PSI does mostly seem to be us based. A good example of why it should only be part of your testing toolset, along perhaps with the potentially more useful real world user measurements.
Running lighthouse locally can also be a good addition to the arsenal too, that naturally overcomes the distance issues by running from your local machine.
But the ones that will matter as part of the page experience ranking signal are from CrUX data. This is real world data, as in the load times actual users of your sites experienced, so by it's nature this overcomes that distance issue, it's the actual experience site visitors are getting.
If you are interested in collecting rum data yourself, this is a handy library to help: https://github.com/GoogleChrome/web-vitals/
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Recommended Answer
Relevant Answer
Hey Eugène,
The PageSpeed Insights 'Lab Data' tests are done using Lighthouse from US servers as far as I know.
However, the data that Google uses/will use to determine performance (the 'Field Data' section you see in PageSpeed Insights), as well as the data you see in Search Console, is from actual users who are visiting your website (data which is stored in the Chrome User Experience Report).
As such, it doesn't really matter where the PSI test is actually run from, as that would only impact the 'lab data', not the data you see in Search Console, which is 'real world usage data' gathered from actual users.
Hope that makes sense.
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Recommended Answer
Relevant Answer
Ah sorry Dave, didn't see you there :)
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Recommended Answer
Relevant Answer
Nice to know we're in agreement Mihai 😀
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Recommended Answer
Relevant Answer
Thanks @Mihai A. and dwsmart,
I guess I feel a bit better now because I've learnt a few things. The first thing I learnt was this new term "CrUX' and somehow I got to https://web.dev/chrome-ux-report-data-studio-dashboard/ and https://crux.run/ which both seemed useful.
The second thing I learnt was about "field data" and the Chrome real-life experiences is being reported on. I have a ton of other questions because this is a highly complex technical field, but I'll start different threads for it.
In the meanwhile is there anything glaringly obvious slow or wrong reported in PSI? Although our site has been running for over two years, and gets a 1000 unique visitors a day, somehow no Field Data.
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
Recommended Answer
Relevant Answer
Your CLS score is pretty bad, that's entirely independent of speed, looks like it could be related to the slider.
There's plenty of stuff to to look at in the rest of the report. Remember, it's not just the speed stuff comes over those wires, but what happens in the browser too, so take a look at the render blocking scripts, unused CSS and other things that perhaps aren't a direct function of speed.
As for field data, not every site gets enough metrics collected to display them, it can often be the more heavy traffic sites. Which is why collecting your own real user measurements can be a good thing.
recommended this
marked this as an answer
Recommended based on info available
Our automated system analyzes replies to choose the one that's most likely to answer the question. If it seems to be helpful, we may eventually mark it as a Recommended Answer.
This question is locked and replying has been disabled.
Link to post
Delete post?
This will remove the reply from the Answers section.
Notifications are off
Your notifications are currently off and you won't receive subscription updates. To turn them on, go to Notifications preferences on your Profile page.
This reply is no longer available.
Badges
Some community members might have badges that indicate their identity or level of participation in a community.
