introtext and fulltext.id field, the speed significantly increased.| Labels |
Added:
No Code Attached Yet
|
||
I just unpacked your backup on my localhost. I know it's not a web server with limitations, but I just wanted to share my results:
Homepage: 225ms average loading time
Back-end articles page with 20 items: 1.10s average loading time
Is it possible that anything else on your configuration may be slowing it down?
I found your test component, so I ran it to add a few more articles. I let it add just over 50 articles. Interestingly enough, the homepage loading time went up to ~2.5s. I deleted them, but the homepage is still slow now.
With a sufficiently powerful PC, for example, with a Core i5 12th generation, M.2 SSD, and 16GB of RAM, the performance will be satisfactory after several refreshes (F5). This implies that you would need to invest in a VPS costing at least $40 per month to achieve relatively smooth performance with 20,000 articles. However, I don't consider this to be an ideal solution.
And this is the result on my live site, hawkhost vps 8Gb ram, 4 core (8GB LAX package).
What happens if you make a backup of your live site and try on localhost? Is the loading time the same?
What I'm trying to figure out here is why did it load well for me when I downloaded your backup and why doesn't it now.
In my case, the three queries that take up most of the loading time are three consecutive queries to the #__content with a lot of columns on BaseDatabaseModel.php:164. They take 900ms, 770ms and 770ms.
What happens if you make a backup of your live site and try on localhost? Is the loading time the same? What I'm trying to figure out here is why did it load well for me when I downloaded your backup and why doesn't it now.
This might because mysql cache your query result on ram. But sometimes it does, sometimes not.
And at my local, the backend article list gets even more time to load, it takes 8s.
In my case, the three queries that take up most of the loading time are three consecutive queries to the
#__contentwith a lot of columns onBaseDatabaseModel.php:164. They take 900ms, 770ms and 770ms.
You can comment out these code on file administrator\components\com_content\src\Model\ArticlesModel.php to get better speed.

This might because mysql cache your query result on ram. But sometimes it does, sometimes not. And at my local, the backend article list gets even more time to load, it takes 8s to load.
Yup, that was it. After restarting MySQL, I have a homepage with 225ms loading time.
I added some articles using your test component, again. Homepage > 2.5s. Restart MySQL server > 225ms.
You can comment out these code on file
administrator\components\com_content\src\Model\ArticlesModel.phpto get better speed.
Commenting the lines you mentioned drops the back-end loading time from 1.25s to 450ms and does not seem to affect the functionality otherwise.
The content: search functionality still works, and a simple search for introtext and fulltext in the back-end reveals no further uses of these columns in the list view, as far as I can tell. I would suggest you make a PR so that others can check as well.
@mavrosxristoforos those code was added since 4.0. I think we have to check carefully before making any PR
That is correct.
I'm thinking now that despite the fact that the articles listing back-end doesn't use these two columns directly in the list view, any extension that loads articles using the ArticlesModel will probably require them.
| Labels |
Added:
bug
|
||
| Labels |
Added:
Performance
|
||
Unfortunately the backup from 3 years ago is not accessible anymore, so I can't look up what you did there. The question is how you generated those articles. I assume you didn't create 20k articles by hand for testing... If the code simply writes to the content table and doesn't do anything else (populate the assets table, populate the workflow table) then Joomla will in fact run slower because it tries to find those assets for each article separately.
While 20k articles isn't a small site, I also can't confirm performance issues in that size-range either. I'm running a site with similar size and it runs just fine. Maybe you can report back on this (even if it is 3 years later) and we can investigate further. Otherwise I would close this issue.
The
fulltextcolumn had once been added with PR #28384 .