Over-optimisation slows your site

Ah, that is way out of my league!! I imagine for some companies it is money well spent.
Yeah from memory, the most I've helped a paying client save is ~US$144K per year after optimization :)
For those who are interested, @MattW has a rather nice pageinsights score:
again focusing the the grading score which is lab data derived and desktop

If folks reading this thread have a few take away summary points from all this is:
  1. Focus on mobile scores as google is going mobile first. Google mobile speed synthetic/lab testers are testing on 3G or 4G speed profiles on Nexus 5 level mobile devices. So bear that in mind as it may not be as relevant to you if 99% of your mobile visitors are on LTE speeds and newer higher end mobile devices
  2. Focus on Google Pagespeed Insight field/origin and Google Search Console Speed report data which is real world
  3. Do not focus on Google Pagespeed Insights's single grading score based on lab data but focus on individual metrics from reported lab data. You can have 2 sites where one has a higher grading score but slower specific individual metric which results in slower overall experience for visitors than the lower grading scored site and vice versa. This is due to lab data metrics having uneven weighting used to calculate the single grading score.
  4. Page speed optimization isn't just front end. If you do not look at backend and server level optimizations, then you're only doing half the effort as front end caching only helps for guest visitors and/or repeat visitors and won't help as much for logged in members and/or single non-repeat visitors. That's why my paid consults involve holistic approach for both backend/server and front end optimisations.
 
Page speed optimization isn't just front end. If you do not look at backend and server level optimizations, then you're only doing half the effort as front end caching only helps for guest visitors and/or repeat visitors and won't help as much for logged in members and/or single non-repeat visitors. That's why my paid consults involve holistic approach for both backend/server and front end optimisations.
In the past when I've spoken about front end and back end caching, there seemed to be a bit of confusion over which was which. So let me have a stab at it!

Front-end:
  • caching css
  • browser cache http header expires
  • CDN
  • litespeed cache

Back-end:
  • mysql tuning
  • litespeed or nginx setup
  • Redis installation
  • opcache installation

This is my mad-hatters guess.
 
What else have you done so far?
Also, you might like https://instant.page/
Hey, that is a really clever thing you just showed me @m1ne! Do you use that on your sites?
It's a cute gimmick designed to create an illusion of faster page loads but in fact it does the opposite.

I tried this on a smaller WordPress site - it made page load speeds noticeably worse in page speed tests, moreso with the mobile version than the desktop version for some reason.

There is another similar scheme available as a plugin. Users report the same thing with that plugin.
 
I tried this on a smaller WordPress site - it made page load speeds noticeably worse in page speed tests, moreso with the mobile version than the desktop version for some reason.
Synthetic page speed tests are not real world as they usually test 1st non-repeat visits. So unless you are specifically devising tests of repeat visits or subsequent navigation visits with hyperlink hovering, you won't see benefits for synthetic tests with instant.page. While real world visitors will see benefits with repeat visits or subsequent navigation visits.
 
I'm still figuring that out myself, TBH. One of the sites where I provide tech support was recently migrated to a Knownhost VPS and we added Litespeed on the recommendation of @JonathanW. Among other things, Cloudflare wasn't managing multiple hits from China any more and then I ran into an annoying issue where I couldn't log in to the Cloudflare dashboard to make changes; we canceled Cloudflare.

If you sign in to the Litespeed Dashboard you will see that there are popups next to every setting to explain what is does and the recommended settings. It seems that by default many options are left at "Not set" and if you're unsure it's probably best to leave it that way. Our biggest concern was not site speed per se but having 2000+ "guests" from China and Taiwan online at the same time.

But ask your host about how you can improve performance with Litespeed. We don't have LSCache installed (yet anyway) but I understand that can also be optimized.




You can get some more hints at



Do you know if we just can disable NGINX and install Litespeed and then XENFORO will run without any additional modification?

I have not used litespeed before.
 
Do you know if we just can disable NGINX and install Litespeed and then XENFORO will run without any additional modification?

I have not used litespeed before.
Sorry. I've never used NGINX. But I can say that installing Litespeed (with Apache) did not require any configuration changes to Xenforo.
 
Sorry. I've never used NGINX. But I can say that installing Litespeed (with Apache) did not require any configuration changes to Xenforo.
So actually, in my case, I would just disable NGINX and switch on LieSpeed so everything would work?

Anyone around here with bride knowledge of the technical part that can confirm this is all to do?
 
So actually, in my case, I would just disable NGINX and switch on LieSpeed so everything would work?

Anyone around here with bride knowledge of the technical part that can confirm this is all to do?
It should be. Litespeed will read the .htaccess (since nginx doesn't) for anything Xenforo requires. If you have anything you've added to your nginx conf you'll need to duplicate that in .htaccess using the correct format. But otherwise, out of the box it should be fine.

That's provided you've done the required configuration in litespeed for your vhost, listener(s) and certs. But that's not Xenforo specific.
 
Top Bottom