Google Webmaster Tools Site performance

[tpearl5]

I'm curious what other vbulletin users are seeing for this new labs feature.

Here's what mine says:

On average, pages in your site take 14.1 seconds to load (updated on Dec 18, 2009). This is slower than 97% of sites.

Yuck! And this is with two very powerful servers (8 cores each), and offloading a lot of static content to CDN's. I do have a lot of hacks and custom stuff going on, so I'm sure that doesn't help.

What is yours?

[imported_silkroad]

We have the same problem..... strange. Is it related to SimpleCDN?

On average, pages in your site take 10.3 seconds to load (updated on Dec 18, 2009). This is slower than 93% of sites. The chart below shows how your site's average page load time has changed over the last few months. For your reference, it also shows the 20th percentile value across all sites, separating slow and fast load times.

However, our site loads much faster, so we are confused as well :D

[tpearl5]

I'm mainly using Amazon s3 & cloudfront, so I don't think so. The data comes from users using the google toolbar, explained here: http://www.google.com/support/webmasters/bin/answer.py?answer=158541&hl=en

[imported_silkroad]

I'm mainly using Amazon s3 & cloudfront, so I don't think so. The data comes from users using the google toolbar, explained here: http://www.google.com/support/webmasters/bin/answer.py?answer=158541&hl=en

Ah.... Interesting... good to know is is not a problem specific to SimpleCDN.

When we check using other tools, the load time is normally less than 500ms, so I don't know why Google reports such an odd performance/load number.

Any ideas?

[tpearl5]

I'm guessing it's because not all users are connecting to our sites at broadband speed - some could be farther or closer to the server. Relatively 10 seconds is better than my 14 though.

[imported_silkroad]

I'm guessing it's because not all users are connecting to our sites at broadband speed - some could be farther or closer to the server. Relatively 10 seconds is better than my 14 though.

You think Google Webmaster Performance is measure from their Search Engine referrals and not directly by GoogleBot?

I am not sure I would agree with that, but I don't yet have any factual basis to offer any counter suggestions. However, I don't think Google measures performance based on limitations of user clients. That would not be accurate at all.

[petchat]

I get 2.2 seconds on average, on a middle of the road VPS, with a few hacks installed :).

[tpearl5]

You think Google Webmaster Performance is measure from their Search Engine referrals and not directly by GoogleBot?

I am not sure I would agree with that, but I don't yet have any factual basis to offer any counter suggestions. However, I don't think Google measures performance based on limitations of user clients. That would not be accurate at all.
Well, it specifically states that the data comes from google toolbar...

Page load time is the total time from the moment the user clicks on a link to your page until the time the entire page is loaded and displayed in a browser. It is collected directly from users who have installed the Google Toolbar and have enabled the optional PageRank feature. (read further down in 'if data is missing or seems inaccurate)

I see what you mean, but after I've thought about it, it is a good way to measure, on average, what your users are experiencing. If more of your users are having a hard time loading your site then perhaps the servers are too far away from the users or a full CDN solution should be considered. For example, I'm sure if I had an influx of users from India that average load time would go up since my main servers are located in the US.

Likewise, if a low number of people using the google toolbar go to your site, and those people happen to have broadband connections, then you're not really getting accurate data - especially if your audience doesn't like toolbars. Yeh, I'm sure the tool could be more intuitive. I guess that's why they put these lines in there:
Data may also not be available for your site if not enough users (with Google Toolbar and the PageRank feature turned on) have visited your site's pages during the time period displayed.... If your site is small, or doesn't attract a lot of traffic, the results may be slightly skewed.

[imported_silkroad]

Well, it specifically states that the data comes from google toolbar...

Oh, I see.. it says:

It shows you the average page load time for pages in your site, the trend over the last few months, and some suggestions on how to make the pages load faster. Page load time is the total time from the moment the user clicks on a link to your page until the time the entire page is loaded and displayed in a browser. It is collected directly from users who have installed the Google Toolbar and have enabled the optional PageRank feature

You are right! If you have users on slow links, you will see very slow results. If you have users on very fast links who don't have the Toolbar installed, their results will not count.

This is a terrible measure of performance!

--------------- Added 1261473677 at 1261473677 ---------------

I get 2.2 seconds on average, on a middle of the road VPS, with a few hacks installed :).

It has very little to do with your server, LOL.

We have a super dedicated server with 16GB of memory and plenty of power and our server is very fast.

The issue is where your users are who have installed Google Toolbar. The further away from your server and the slower their link, the worse performance.

In other words, this Google Labs feature is not very useful at all on sites with global traffic.

[Princeton]

In other words, this Google Labs feature is not very useful at all on sites with global traffic.it doesn't matter where your visitors are coming from - a high # is an indicator that your site needs tweaking

it's a useful tool but don't rely on it - use it as an indicator of how well your changes are working

this tool averages the amount of time that the entire page takes to load to the end-user ... this includes html, images, scripts, javascript execution, css, etc ... your server may output the html in 0.9sec but by the time the page loads completely to the end-user that 0.9sec may turn into 12 sec

unlike Googlebots this tool parses all pages - this includes pages that are disallowed via robot.tx and other pages that are not normally indexed by Googlebots

for example, reply.php, newrepley.php, etc, etc are pages that are not normally accessed by Googlebots

however, the toolbar will count these pages in your overall site performance
(because of the javascript involved in displaying the wysiwyg editor these pages render slow which will increase your site performance average)

another thing that may affect how fast your page load is your page design - to learn more about this read a few of my articles (found in sig)

if you are currently getting a high number - it means you need to work on your pages...
Google has already stated that it may use web page speed for indexing pages.

[BSMedia]

I'm curious what other vbulletin users are seeing for this new labs feature.

Here's what mine says:

On average, pages in your site take 14.1 seconds to load (updated on Dec 18, 2009). This is slower than 97% of sites.

Yuck! And this is with two very powerful servers (8 cores each), and offloading a lot of static content to CDN's. I do have a lot of hacks and custom stuff going on, so I'm sure that doesn't help.

What is yours?

You can easily improve that with some vBulletin 4 .htaccess optimization (http://davidmchenry.info/2009/12/18/vbulletin-4-optimized-htaccess/) and it sounds like you've already got Parallel hostnames to serve images to Optimize vBulletin (http://davidmchenry.info/2009/12/18/how-to-serve-parallel-hostnames-for-vbulletin-optimize-vbulletin-4/) working for you.

But what version of vBulletin are you using btw

[imported_silkroad]

It means nothing.

Our pages load lighting fast from any "normal" location on the Internet.

The numbers from the Google Performance Tools are a joke and are simple meaningless.

PS: If you want to measure performance use cURL. For example our pages load in less than 100 ms with cURL tests, sometimes in 20ms. The Google Performance Tool currently shows the same page loading in 9.1 seconds. Furthermore, Google Crawl Status show pages loading at less than 400 ms.

As I said, the Google Labs Performance Tool is pretty much useless.

[Andreas]

The numbers might not be 100% accurate, but they are far from being meaningless.

You need to understand that the raw time to deliver the HTML for a page is only a (usually pretty small) part of the total page loading time.

So even if you are delivering the page in 400 ms it might very well take 9 seconds to get it loaded completely.

[imported_silkroad]

We understand the numbers. We deliver over 4 Million Page Views a month and run many stats on our pages.

The numbers are pretty much meaningless. Our pages do not load in 9 seconds for anyone unless they are in on a very slow link in their home country.

Then are not even close to accurate.

You cannot judge the speed of a server based on the performance of an unknown client in locations that might be on dialup or very slow links.

Sorry, but to argue otherwise shows a lack of understanding of basic networking.

Client side performance does not equate to service side performance.

In addition, for large global sites, most smaller and less developed countries have very slow access back to the US or even global delivery nodes.

The only way the numbers would have any meaning would to provide the country and perhaps also the measured bandwidth stats.

[Andreas]

Well, I think I do have some basic understandig on networking (our forums are generating ~ 70 million Page Impressions/Month) ;)

Feel free to post your URL and i'll post my loading times (6 MBit DSL @ Home, 16 MBit DSL @ Work)

Client side performance does not equate to service side performance.
You are absolutely right - the Google Webmaster Tools Site Performance is not measuring you raw server throughput/response time but end-user performance (which should be < 2 seconds for the complete page, otherwise you will loose visitors).

[imported_silkroad]

Well, I think I do have some basic understandig on networking (our forums are generating ~ 70 million Page Impressions/Month) ;)

Feel free to post your URL and i'll post my loading times (6 MBit DSL @ Home, 16 MBit DSL @ Work) :-)


You are absolutely right - the Google Webmaster Tools Site Performance is not measuring you raw server throughput/response time but end-user performance (which should be < 2 seconds for the complete page, otherwise you will loose visitors).

Andreas,

I don't need to post the URL. We have many hosts in the US and EU and we measure very easily from high speed datacenters on the Internet backbone, which is faster than your edge node. Our load times are very fast. We know what we are doing, I promise you. We are not vBulletin theorists, we serve real pages and over 4M a month to over 2 M users :-)

Also, you are misleading in your blanket statement:

.... end-user performance (which should be < 2 seconds for the complete page, otherwise you will loose visitors).You can have the faster server in the world. If you end user is in a slow link in India or Malaysia there is nothing you can do to speed up the process. We deliver static content with a CDN, having used both SimpleCDN and CloudFront/S3 with global nodes in the US, EU and Asia, so we know what we are talking about.

It matters zero if you have the fastest servers and the closest CDN nodes if your end user is on a slow link, slow part of the world, etc. Users in a slow part of the world are used to slow load speeds because they cannot do any better than that. Yours statement assumes the user has fast access every day and is used to fast access. That does not apply to much of the world outside of developed countries like the US, developed EU countries, Japan, Singapore, S. Korea, etc.

I find it funny (but it's OK, I'll try to be patient) that are you making generic statements as if you are talking to a novice, but nevermind. Let's move the level of conversation up to a more professional level where you understand you are talking to someone who has a busy server (but not busy like yours!), distributed architecture, and can easily use many nodes to measure from, etc. We know the Google Performance stats are meaningless. Our users and moderators measure load times for us and post!

In addition, I can easily run Firebug and their performance tools when I am on travel to third world countries, and do so nearly every day, and I am currently on a slowish link on holiday in Asia. We have over 20 moderators around the world and they post very quickly if the load is slow..... LOL

I am telling you with 100% confidence, not theory, that the Google Toolbar performance tool is very inaccurate and has very little useful meaning. You can argue with me until we both grow weary of the discussion; but you will not change the facts and figures of real-world global network operations.

Your arguments only hold true if the user base is in developed countries with high speed Internet access. The real world is much bigger than that :D and so are many forum user bases.

[BSMedia]

I am telling you with 100% confidence, not theory, that the Google Toolbar performance tool is very inaccurate and has very little useful meaning. You can argue with me until we both grow weary of the discussion; but you will not change the facts and figures of real-world global network operations.


But if those numbers are what Google is reporting to you about your site, and those numbers are what Google is going to use to factor in page rankings, then they are far from meaningless for users who love SE traffic and get the majority of it from Google.

[imported_silkroad]

But if those numbers are what Google is reporting to you about your site, and those numbers are what Google is going to use to factor in page rankings, then they are far from meaningless for users who love SE traffic and get the majority of it from Google.

No, sorry... you are wrong.

Those numbers in Google Labs are not used by Google in PageRank factoring.

Where on Earth did you get that idea?

Google PR comes from GoogleBot and standard GoogleBot post-processing not from Google Toolbar reporting... this is common knowledge, BTW. You can easily learn about PR here on Wikipedia and elsewhere; http://en.wikipedia.org/wiki/PageRank

I am a bit surprised by your post... it is completely wrong.:confused:

PageRank in its simplest sense (as per Sergey and Larry's thesis at Stanford) is the relative value of any single page on the internet in relation to every page on the internet. Toolbar PageRank is the pagerank score for any page (on a scale from 1-10) that is shown by google to the public in the google toolbar... For many years, there has been some question about the true relationship between real-time pagerank and toolbar pagerank, because at best, toolbar pagerank only shows the pagerank of a page as of the most recent toolbar pagerank export, which traditionally, google only does once every 3-4 months (give or take)... This most recent toolbar pagerank export/update, my site rose to a PR6, meaning that it's real pagerank the day/week before the update was undoubtedly PR6, however it showed in the toolbar as a PR5, because that's what it had the last update, months earlier...

In other words, Google Toolbar only reports (and not very well), PageRank... it is not used to calculate PR or any other Google metric that determines how your page is indexed.

[Andreas]

If you end user is in a slow link in India or Malaysia there is nothing you can do to speed up the process.
Of course you can (eg. design the site so it does still work fine with limited bandwidth).

We know the Google Performance stats are meaningless. Our users and moderators measure load times for us and post!
That's your opinion and I am not going to argue with you here as it seems rather pointless.

To sum it up: Speed does matter (http://videos.webpronews.com/2009/11/13/matt-cutts-interview/).
And IMHO the Google Webmaster Tools are far from being meaningless.

[tpearl5]

anyone else try this out recently?

[Andreas]

Those numbers in Google Labs are not used by Google in PageRank factoring.
Really?
Then Google must by lying :)

http://googlewebmastercentral.blogspot.com/2010/04/using-site-speed-in-web-search-ranking.html