Why such a wide difference in results?

[deking]

I am trying to identify the cause of my sites slow loading time. First of all, why such a wide range of results in the different online testers? My results range from 51.5s on UpTrends.com to 2.8s on Site-Perf.com. I get a 14.51s on WebsiteOptimization, 39s on WebToolHub, and 31.95s on WebPageTest.org
This site seems to offer the most information, so I joined this forum. Can I get a recommendation to fix my problem? Is it in the coding? …or the shared hosting (Hostgator)? My site is here… www.naffs-kingdom-cat-shop.com

[jarrod1937]

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
http://www.webpagetest.org/result/100624_d4e6bb383d5ed734ad7cc5e0f5dfb08d/

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

[p83822]

@jarrod1937

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
WebPageTest Test - Running web page performance and optimization tests...

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

If I may suggest:

I’d recommend running the YSlow and Page Speed FireBug plugins before testing real world locations. By the time you arrive at a personal view on the strengths and weaknesses of both tools (both are very good though) you will have made some quick wins, have a view on what you want to fix medium term and will have a better view on the way problems manifest when looking at WebPageTest results.

Good luck.[hr]

@p83822

@jarrod1937

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
WebPageTest Test - Running web page performance and optimization tests...

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

If I may suggest:

I’d recommend running the YSlow and Page Speed FireBug plugins before testing real world locations. By the time you arrive at a personal view on the strengths and weaknesses of both tools (both are very good though) you will have made some quick wins, have a view on what you want to fix medium term and will have a better view on the way problems manifest when looking at WebPageTest results.

Good luck.

Just re-read my reply and I’m worried that it sounds patronising and stand-offish (quite the opposite of what was intended.) It’s midnight here and I’m in the office chasing down a server side bottleneck in our big rig, so apologies if I didn’t get the tone right.

BR

Neil

[deking]

@jarrod1937

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
WebPageTest Test - Running web page performance and optimization tests...

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

Many thanks, jarrod1937, for your response. Actually I took only the fastest (DSL) reading from each of the testing sites for comparison purposes. I contacted Hostgator about enabling “Keep Alive” and learned that it’s not available with my shared hosting package…I guess I’m now faced with with upgrading to their VPS hosting or changing companies. I am using the latest CRE Loaded Pro B2B shopping cart which contains a Gzip feature…but, it obviously works only on the php files. I did manage to improve my test score on “Cache Static Content” by adding a code snippet I found in this forum to my .htaccess file. I still need to address the other problems…“Compress text”, “Combine JS & CSS” and ETags". I have no clue where to begin and would appreciate any direction you might offer.

Tom

[jarrod1937]

I’d recommend switching hosts then. Keep alive should be one of those basic things they should either already be doing or something they shouldn’t mind changing. It is up to you if you want to go the vps route, but if you do, you’ll definitely be given better control to help speed up your site.

[WebKing]

Also, double check if your host [existing or new] doesn’t already Gzipping your data. So, you are not doubling it with your cart’s default settings.
Combining JS files or combing CSS into one isn’t that tough. Cutting down css for master page/index page, is another way to speed up the process.
ATB

[sbglobal]

@p83822

@jarrod1937

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
WebPageTest Test - Running web page performance and optimization tests...

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

If I may suggest:

I’d recommend running the YSlow and Page Speed FireBug plugins before testing real world locations. By the time you arrive at a personal view on the strengths and weaknesses of both tools (both are very good though) you will have made some quick wins, have a view on what you want to fix medium term and will have a better view on the way problems manifest when looking at WebPageTest results.

Good luck.[hr]

@p83822

@jarrod1937

The variations in the other test sites could be from their location from your server. But more than likely, due to the wide range of times, they’re probably testing at different internet speeds. The default test is dulles va at 1.5 Mb dsl, the other sites are probably using faster connection speeds which can be misleading unless your site is visited only by the elite of the internet
Looking at the results of your site:
WebPageTest Test - Running web page performance and optimization tests...

The first, and biggest thing you need to do is make sure keep-alive is enabled and have it set at a reasonable time. Otherwise you’ll see, as your site currently is, each request requires a new connection to be made, which can greatly delay a page. Your repeat view shows you don’t have any of the correct caching headers set, causing even the repeat view to be just as slow as your first view. Proper caching headers will prevent the browser from having to send a conditional request and the server from sending a 304 (not modified) response. You also have 9 .css files loading, if you can combine these into as few as possible css files, preferably one. Along with this you also have 7 requests that are, for whatever reason, going over an ssl connection. Ssl connections are quite expensive and should be avoided if possible. You should also look into css sprites, counting just by eye i can see 12 requests that can be eliminated.

If I may suggest:

I’d recommend running the YSlow and Page Speed FireBug plugins before testing real world locations. By the time you arrive at a personal view on the strengths and weaknesses of both tools (both are very good though) you will have made some quick wins, have a view on what you want to fix medium term and will have a better view on the way problems manifest when looking at WebPageTest results.

Good luck.

Just re-read my reply and I’m worried that it sounds patronising and stand-offish (quite the opposite of what was intended.) It’s midnight here and I’m in the office chasing down a server side bottleneck in our big rig, so apologies if I didn’t get the tone right.

BR

Neil

Thanks for telling about these some information

[pmeenan]

My recommendation would be to get off of host gator (not pay them more for their VPS) and make it exceedingly clear to them that you’re leaving because of the horrible performance penalty they are introducing by not supporting keep-alives. There’s no good reason for it to be disabled (yes, there is a good reason for THEM to want it disabled but just because they are oversubscribed on their capacity). You’re not the first person I’ve talked to that has had this problem with them and it’s inexcusable.

If you have too much content to move you can look at using a CDN for your static content. If you only serve the base page from hostgator then the persistent connections won’t matter as much. Cloudflare has a free option and MaxCDN is pretty cheap, otherwise most will probably cost more than your hosting.