Thread

Hi everyone,

I know this question has been asked in many different ways before, but I have never found a satisfactory answer.

All sites I develop using EE are slow to load. I know if you set the scripts executed in tag to show, it comes back with quite a normal if not better than average speed, but what is it between executing the script and displaying the page that causes the delay?

Could it be the way that I am building the site? The structure?

Up until now it has just been something that I was aware of and it hadn’t had too much impact on me as such, but today a site that I have built was rejected for funding, one of the reasons being that it was unacceptably slow and unuseable on dialup speeds.

I have built it using a total css based layout, almost all images are background images using css, and yet it still loads slow. Is there any way to improve this?

Thanks

Patrick

There could be so many things. I know, one of the things we do for speed tests is to check to see how long it takes the page to load with and without EE Queries. So if I think the archive page of a site is running slow. I’ll keep the EE page, and then just collect the HTML from the generated page, and place that in both an EE template and a standard html file to check and see how long it loads.

The pure html file allows you to check to see if EE is the problem. So if the pure HTML file is a lot faster than the HTML in a EE Template w/o ee code, then obviously part of the problem is time to communicate with the MySQL database.  If the pure HTML file isn’t that much faster than the HTML in EE; but both are a lot faster than a template with EE Code, than I assume my use of EE Code is bad or not optimized.  If even the pure HTML code is slow, then I start doing server, code and browser tests. You didn’t state if all your EE sites are hosted by the same ISP, but its amazing what speed differences there can be across ISPs.  A slow ISP can really ruin presentation and makes a huge difference if you have a lot of dial-up users.

The EE page render time, I believe just determines the length of time the queries took, it doesn’t really measure page load speed. I’ve seen pages render in .328 seconds from EE that take upwards of 3 seconds to display on screen. That just means your EE code isn’t the problem, it doesn’t mean you don’t have problems elsewhere. There are a number of tools that will go through your site or page by page to analyze download speeds. I occasionally use http://www.websiteoptimization.com/services/analyze/ I’m not a big fan of how it reports, but I do like that it breaks out sizes of different types of content such as HTML vs Javascript vs Multimedia vs images.  It also gives good speed estimates for different configurations such as 56k vs 44k or even 14.k modem speeds.

That’s a pretty wide open question, could be a lot of things. Do you have a link to the site?

Even though it’s all CSS doesn’t necessarily mean it’ll be fast…would need to see the site to tell.
If the times EE are showing for processing are acceptable/normal, then it’s more than likely not EE, the slowdown takes place after EE’s done it’s thing.
Could be the network. Have you tried a traceroute to see what that looks like

Have you already optimized the templates? I assume you have, but just thought I’d ask…

yep, all optimised… link is here

I do use different hosts, and I find it does vary, but still overall opinion is that it is slower than other sites.

Sorry it is such a wide open question, it is just something that seems to cover all the sites I have done, and hasn’t been too much of an issue until now.

Thanks allgood2 for the detailed response, I will have a look at doing some of that to see if it is of use.

Thanks

Patrick

Took about 10 to 15 seconds to load twice. I found that on one of my sites that utilizes the EE calendar which was looking up content from multiple weblogs (and looking up thousands of records) was slowing the load page time for me. But even so, it still loads pretty damn fast even when uncached.

And your site on refresh isn’t loading all that much faster for me. I’d look at your network host, and then whether you are caching your site effectively, and if you have any abnormally super inefficient EE code that you might have custom created.

If EE was as slow as your website loading, these forums would be empty.

I don’t dispute that it isn’t ‘EE’ as such, as the forums do load normally, as do a few other EE sites I have seen. It just seems that there are a lot of people who experience the same as me, so was wondering if it was possible to pin-point where the error was occuring.

From my experience there are several things that can cause a site to slow-down:

1) A cache directory that doesn’t have correct permissions.
2) Spiders and bots that are absolutely out of control.
3) The wrong infrastructure for a dynamic site - non-optimized hardware and server setups.
4) Templates - custom plugins, extensions, sql queries, nesting. 

You can find out if it’s your templates.  Load the default template into another template group calling your weblog and visit it - is it fast?

The only way we can help you if it is the templates is by seeing them, along with the debugging - SQL queries and template debugger output.

The URL is helpful, here’s what I can tell you from just loading the page:  I loaded the page twice, and my Safari Activity log read that it lost network connection for two items, one being statcounter.js which is stored remotely. Also the xinha script took awhile to load, it’s also pretty darn large—over 10 times the size of anything else on the page except the logo.  There was a calendar item that didn’t load (a plus.gif), I’m not even certain if it should be on the page.  If you’re not allowing trackbacks and comments you should get rid of the code for that, it’s just extraneous html.  Also your multimedia file is large, is that a flash file? That file all by itself accounts for about 8 seconds of page load. 

Removing trackback data would almost reduce your HTML code in half, but that doesn’t seem to be as much as an issue as flash, javascript, and images. I also ran a traceroute, and 33% of the package data was lost going to the domain. I was zooming until I hit optus.net.au than slow crash. That may just be the ISP isn’t returning proper response. Some ISPs block pings and traceroutes.  That was all about 5 min of inspections. I ran a traceroute, a browser activity report and a web page speed report, as well as looked at the site code, so there’s a lot that can be looked at.

wow that is detailed… is there a similar thing to the activity log for firefox or opera? I am thinking of removing xinha, and replacing with TinyMCEditor, as xinha isnt compatible with opera. I will make the other amendments, and see how much of a difference that makes.

thanks heaps for the analysis, i’d love to know about the log thing, as it would be extremely useful to have this information to hand regarding other websites.

Thanks again

Patrick

Ok, I have removed statcounter, the popup calendar which is where i think you were getting that gif from, removed flash image, and removed xinha. Would it be possible to check now to see how it compares?

Thanks

Patrick

Well except for traceroute/ possible dns issues, your site loads much faster for me now.

I’ve attached a speed test. Unfortunately, I didn’t save the first on, but on the first test your T1 speeds was almost 6 seconds to load, and your 14.4k was suicidal, over 120 seconds. Comparatively, your T1 speed is down to .98 seconds, and 14.4K down to 28 seconds. Much better. Also you got rid of 3-4 red warning, in the recommendations section.

No errors in the Safari Activity monitor, that’s good. I’m not certain if Firefox has an equivalent. I’d guess yes, just because one of the PHP developers I work with is always telling me Firefox is better for coding than Safari.  But I like using Safari, I keep the Debug menu active, and use the Activity log fairly frequently.

[Okay] It must be the after midnight monsters, because this is the my third attempt to post a response. Amazingly each response is so much less in detail than before, but the gist is the same.  I’m just using the fast reply, because the normal post reply is bugging me. I sent the speed test results as a PM.

i found the magic button in firefox 😊 thanks for testing again, much appreciated. Thanks to everyone for your suggestions and help, they have all made it clearer to me how I can streamline my approach to building using EE.

Thanks

Patrick

Sounds like it was determined that most (if not all) problems were outside of EE…correct?

seems that way 😊 Sorry if it made it sound like I was ‘blaming’ EE… I was merely saying that the only time I had experienced the slowdown was with EE sites, and there are so many other people on the forums who have the same issue. So wanted to pinpoint where it was. It appears that the above suggestions have resolved the speed issues, at least on paper. I have yet to test the site on dialup. In any case, it may not be enough to save the funding crisis for this site, but it will certainly help with future sites.

So thanks again for all replies, I hope that this thread helps others too 😊

Patrick

I just want to address this:

here are so many other people on the forums who have the same issue.

First - site slowdowns are not that prevelant on the forums. I can say that because *I* am prevelant on the forums and they simply do not come up as often as implied by your statement.

In every one of the cases where it has come up the problem has been tracked down to badly coded templates - EE is flexible and powerful, that doesn’t mean you should use *every feature* multiple times in one template! A spider/bot attack, or a hosting situation, often in a shared environment, where there is some large amount of over-selling going on.

And there are other cases - where people are calling external services (Flickr used to be a huge cause of this) and those calls are slowing down the system - as you experienced here.

Ultimately, these forums are for technical support and so we get people here that are in need of assistance.  It is *never* good to look at support forums as an indicator of if a problem is common.  For instance - I have a sick cat, and I go to support systems for help with her - if you read those forums, you would think 90% of cats are very sick and that they’re a bad pet choice.  I’m sure that we all know that is not true.

The same thing happens here.  We have thousands of customers that will never come near the forums because they simply have no need of support.  There is an even higher number that simply lurk, finding their answers and never posting a single question.

So while yes, you can look for posts about speed problems, and quite a few threads - that is not indicative of it being a common ExpressionEngine problem - it is not.  ExpressionEngine itself has no inherent speed problems, it is extraordinarly fast.

Patrick, I do not mean to soapbox on your thread, and I do apologize - but I wanted to address that for future viewers of this thread.