Slow Load Speed

[tekguru]

Guys did you get the issues resolved here? Still looking to go ahead with the change but obviously want to avoid any file deletions.

I asked Invision support about this and they responded with:

Unfortunately, there is no "approved" method by IPS in order to do this but from what I read, you would not need to initiate a transfer.
 
>>>>>Theoretically, if you are aiming to do this for all your files, you would not need to create a new configuration and initiate a move. You would just change the "Custom URL" of your configuration to point to your sub-domain. As you sub-domain is pointing to your root and the folder already exists with the files in it. <<<<
 
Please ensure before you try this that you have a full backup (files and database) done, just as a precaution as there will be nothing we can do about deleted files.

 

If you note the part I've highlighted  it looks like there is no need to create a new storage at all?

It also makes me wonder as to what would be gains for myself as I was already using the database for storage (hmm at some point this has changed back to files). Weird.

[Rhett]

The location of the files is the same.( .../uploads/...) If you created a true cookie - free domain no problem.
This method changes the only way of loading, but not content & location files.

It seems all you are doing here is tricking the testing tool, it shouldn't make any performance difference by changing the sub/domain used here if they are being served from the same server in the same location. So actual performance here would be identical, but the test tool you are using seems to rate this higher, real world performance would be no change. Can you clarify? 

[Adlago]

It seems all you are doing here is tricking the testing tool, it shouldn't make any performance difference by changing the sub/domain used here if they are being served from the same server in the same location. So actual performance here would be identical, but the test tool you are using seems to rate this higher, real world performance would be no change. Can you clarify? 

Sub domain addresses the root directory of domain acts as a cookie-free domain.

For example, if your domain is site.com, resources are loaded from the "site.com/uploads". If you create a sub domain addresses in the root of the domain, for example "stat.sait.com" , then sub domain and the domain itself are equivalent in their work. One small difference is that "stat.site.com" loaded without cookies.
This improves the performance of the site.

The tool with which test all web servers test - All indicators are very good

You can do tests with my website you want a tool and you will see in this.

 

Here is an example with Google Development

This method I applied in many sites. Its perfect work.

 

[Rhett]

Hello, thanks for the clarification, however in testing actually performance (not a test site tool) differences there is no difference, you are only improving the results on the testing tool and not a real performance improvement.  So for those that want to see a higher test score, I guess this helps, however in real world usage, loading a cookie or not is not going to improve anything from what I'm seeing here. So from what I can see so far, this does nothing more than report a better score. 

 

 

[Adlago]

 So from what I can see so far, this does nothing more than report a better score.

 

This disagree.

Is such a simple efficiency is missed by teams developing Internet communication?

[Adlago]

I think that this topic is a confusion of concepts.

Have performance software, and performance site.

Performance software depends on your company. Performance of site depend on software, applicable ideas user of software, members liked the idea of your users, all combined in the global network.

Performance software improves work
Files - Server - Network
Performance site depends on Users - Site- Network.
This combination is administered by the site owners, and their benefit is what I've written here.
And it is not simply a  message to a test server and real-performance sait in network.

[tekguru]

For the moment I'm holding off just in case due to this bug. As fronm what has been suggested a new storage is not required (as that will trigger file moves) but just enough to add in and turn on the custom URL.

[Adlago]

For the moment I'm holding off just in case due to this bug. As fronm what has been suggested a new storage is not required (as that will trigger file moves) but just enough to add in and turn on the custom URL.

This bug has nothing to do with what I write.

[Xae]

It's worth noting that cookie-less domains and CDNs only affect the initial loading of static resources. On every subsequent request, they will all be retrieved from the cache, at which point the only factor affecting site speed is the TTLB, i.e. the time it takes for IPS4 to generate and send the response. Unless you run a site with a significant focus on static content, such as images and video, there's no need to serve it any way other than default. Yeah, the first visit may load a little faster but your regulars won't see a difference.

The score you get from site speed tools, like YSlow or PageSpeed, has no direct relation to perceived loading speed for the user. The tools are designed to tell you what can be optimised from a technical standpoint. In other words, the score on its own is relatively meaningless. It's there to provide an overall estimate for people who don't know how to interpret the data. You may have noticed that Google removed the score from the Chrome PageSpeed tool.

Anyway, aside from running off good hardware (like a Linode), one of the easiest and most effective ways of increasing performance is PHP caching, which isn't always enabled by default. Then there's the web server. Preferably, you'll want nginx, or Apache with the event MPM, paired with PHP-FPM. Apache with mod_php is abysmal in comparison but it's still used on many servers, especially for shared hosting.

[RevengeFNF]

The only test i care is the page load time. I prefer to have a score of 92/100 and a page load time of 0,4s, than a score of 99/100 and a page load time of 1s. Simple as that.

In terms of webserver, i tried Apache and Nginx standalone, and at the end i prefer to use both. Nginx as a reverse proxy to Apache. Now im waiting for PHP 7, that will give double or more the speed in terms of requests handling.

[tjk]

It's worth noting that cookie-less domains and CDNs only affect the initial loading of static resources. On every subsequent request, they will all be retrieved from the cache, at which point the only factor affecting site speed is the TTLB, i.e. the time it takes for IPS4 to generate and send the response. Unless you run a site with a significant focus on static content, such as images and video, there's no need to serve it any way other than default. Yeah, the first visit may load a little faster but your regulars won't see a difference.

The score you get from site speed tools, like YSlow or PageSpeed, has no direct relation to perceived loading speed for the user. The tools are designed to tell you what can be optimised from a technical standpoint. In other words, the score on its own is relatively meaningless. It's there to provide an overall estimate for people who don't know how to interpret the data. You may have noticed that Google removed the score from the Chrome PageSpeed tool.

Anyway, aside from running off good hardware (like a Linode), one of the easiest and most effective ways of increasing performance is PHP caching, which isn't always enabled by default. Then there's the web server. Preferably, you'll want nginx, or Apache with the event MPM, paired with PHP-FPM. Apache with mod_php is abysmal in comparison but it's still used on many servers, especially for shared hosting.

Xae nailed it.  Focus on page load times and your content. 

[tekguru]

 

Anyway, aside from running off good hardware (like a Linode), one of the easiest and most effective ways of increasing performance is PHP caching, which isn't always enabled by default. Then there's the web server. Preferably, you'll want nginx, or Apache with the event MPM, paired with PHP-FPM. Apache with mod_php is abysmal in comparison but it's still used on many servers, especially for shared hosting.

We're on a low resource shared hosted server and we're running Apache with mod_php. Could you give any tips as to what 'Apache with the event MPM, paired with PHP-FPM' is and what it may entail getting it configured? Anything we can do to improve efficiency and gain speed that is viable is worthwhile investigating.

[Xae]

We're on a low resource shared hosted server and we're running Apache with mod_php. Could you give any tips as to what 'Apache with the event MPM, paired with PHP-FPM' is and what it may entail getting it configured? Anything we can do to improve efficiency and gain speed that is viable is worthwhile investigating.

Unfortunately, shared hosting won't let you do that, since it involves making changes to the server software and configuration. It's a dead end, I'm afraid. If you were on a VPS, I could help you make the change.

I don't have any other hosting providers to recommend -- all my sites are self-managed on Linode -- but someone else here might know a good one. You don't have to manage your own server to use the configuration I suggested in my previous post, it's just that most shared hosting providers are notorious for running software that's been obsolete for years.

[tekguru]

We are on shared hosted machine but it is a VPS if that helps at all? I'm with ServInt on the Flex $55 package as you can see here.

[Xae]

We are on shared hosted machine but it is a VPS if that helps at all? I'm with ServInt on the Flex $55 package as you can see here.

Yes, that does help. I'd assumed you were talking about shared hosting because of the bit you quoted but if you're on a VPS, that means you've got control over the operating system. According to the link, you're on a managed plan, which means that ServInt should be willing to make these changes for you. I'd say it's worth asking them first to be sure.

Send them a message and let me know. Making the change manually isn't difficult but it sounds like these guys have got some custom setups going on which means they'll know them by heart.

[tekguru]

Okay so what exactly do I need to ask them? I've no familiarity with MPM or PHP-FPM.

[Xae]

Okay so what exactly do I need to ask them? I've no familiarity with MPM or PHP-FPM.

Tell them that you'd like to "switch to using the event MPM for Apache with PHP-FPM instead of mod_php". You could also mention it's for performance reasons. They'll know what you're talking about.

[tekguru]

Okay before I do that should I take the changes out of the .htacces file that were put in for mod_php? Wondering as I know if we do that we lose a heck of a lot of speed on site. Will the new combination more than make up for the speed difference?

 

[RevengeFNF]

Make sure you are on Apache 2.4 to use the Event MPM. In Apache 2.2 the Event MPM was experimental. In that case, use instead Worker MPM.

Note that Worker or Event are faster than Prefork if you have many concurrent requests to your server. If your server have few access's, prefork with mod_php is actually faster, because its embed in Apache. 

[tekguru]

Now that makes me wonder if we should stay where we are as we are a very low usage site with maybe 1-2 users and 10 bots on site at one time. So maybe stay as we are?

[Xae]

Note that Worker or Event are faster than Prefork if you have many concurrent requests to your server. If your server have few access's, prefork with mod_php is actually faster, because its embed in Apache. 

I'm highly surprised by this statement since I'm confident it's the exact opposite. Have you got benchmarks at hand? The IP loopback (or unix socket) overhead for PHP-FPM is miniscule, and the difference in processing time is due to FastCGI.

Now that makes me wonder if we should stay where we are as we are a very low usage site with maybe 1-2 users and 10 bots on site at one time. So maybe stay as we are?

If you've got very low activity and you're happy with the browsing speed, you don't need to change for the sake of it.

[RevengeFNF]

@Xae i never run bechmarks myself to prove this, but i have read many benchmarks about the many MPM's. The conclusion is that if you have a site with very few requests, mod_php is faster. If you have many concurrent requests, worker or event with php-fpm is faster because it uses threads.

PS: If you are on a VPS with low resources, you should consider using Nginx instead of Apache.

[tekguru]

If you've got very low activity and you're happy with the browsing speed, you don't need to change for the sake of it.

LOL not happy with it, but without buying larger resources we'll manage 

PS: If you are on a VPS with low resources, you should consider using Nginx instead of Apache.

If that so much more efficient? We'd lose mod_php though?

[RevengeFNF]

Nginx works with php-fpm and have the advantage of consume less ram. So, if your vps have low ram, its better to use Nginx.

In my case, i use Nginx as a reverse proxy to Apache, but i have a lot of Ram.

[tekguru]

Right but if php-fpm works well for busy sites and ours is not maybe not worth the effort of too much messing around to achieve little gain.