Google Launches Trial of Page Speed Service

[engine]

Google Launches Trial of Page Speed Service [googlecode.blogspot.com]

Page Speed Service is an online service that automatically speeds up loading of your web pages. To use the service, you need to sign up and point your site’s DNS entry to Google. Page Speed Service fetches content from your servers, rewrites your pages by applying web performance best practices, and serves them to end users via Google's servers across the globe. Your users will continue to access your site just as they did before, only with faster load times. Now you don’t have to worry about concatenating CSS, compressing images, caching, gzipping resources or other web performance best practices.

In our testing we have seen speed improvements of 25% to 60% on several sites. But we know you care most about the numbers for your site, so check out how much Page Speed Service can speed up your site. If you’re encouraged by the results, please sign up. If not, be sure to check back later. We are diligently working on adding more improvements to the service.

At this time, Page Speed Service is being offered to a limited set of webmasters free of charge. Pricing will be competitive and details will be made available later.

[bears5122]

There are plenty of CDNs out there that have been doing this for years. They are much more experienced and have worked out the kinks. Despite all the privacy issues I think website owners should have with this, it's more about putting your site on an untested service from a company that tends to have their share of launch problems.

[Sgt_Kickaxe]

Did Google buy webpagetest.org?

I just spent some time testing the service, as it is now, and this is what I found...

- Google has some serious hardware/connection speed assigned to this. Start render times are impressive as are Repeat View Page Load Time and Repeat View Start Render Time as a result of the quick launch out of the gate.

- Google is displaying LESS optimized code, their version has browser caching set to 4 minutes 59 seconds for example when much longer is appropriate on some sites. This is typical of a one setup fits all sites approach.

- Total page load time was slower in my tests despite the quick starts, the main reason looks like a case of too many files. My site makes 22 requests (and was a pain to get down to that) but after Google gets the page they run 50 requests to load it.

Conclusion, I don't suspect the quick start render time will remain quick as more sites are piled onto the servers which may negate the biggest advantage the service offers. If you're not into code and optimization on your own however you WILL see a faster site.

A question - el-cheapo sites on shared hosting will benefit most but will the owners of these sites want to pay their hosting fees AND Googles monthly fee for a small performance gain?

It's like Muscle Cars vs Rice Burners. Brute horsepower vs fine tuning. Both have pros and cons.

[TinkyWinky]

Will you walk into my parlour ? said the spider..

LOL - that made me laugh... as was roughly what I was thinking.

This idea will have to be binned soon - surely... is anyone dumb enough to trust their website to G... seriously?

[indyank]

AFAIK Google buys out and offers a platform for free, ONLY if they see a goldmine in it. The were running the back-end of the internet so far and now decided to run the front-end of the show too!

[DaStarBuG]

Tested a vBulletin powered forum (optimized already) and a Wordpress Page. Both very more optimized and faster on their own then with Google

Not interested :)

[londrum]

maybe i'm a cynic, but now we know why they started reporting page speed in WMT, and made webmasters think that page speed is part of the algo... it was to sell this.

i dont think that its a coincedence that all those free page speed tools came out months before they launched this.

[MarvinH]

Did Google buy webpagetest.org?

Pat, the helpful fellow who runs the site, is a G**gle employee, as far as I know.

[denisl]

Can we assume from this that G may put more emphasis on speed as a ranking factor?

[chrisv1963]

maybe i'm a cynic, but now we know why they started reporting page speed in WMT, and made webmasters think that page speed is part of the algo... it was to sell this

I don't know. According to WMT my website is slow. According to the test I ran it is fast and Google's Page Speed Service can not make it any faster and would actually make it slower. Why am I starting to have doubts about whether or not you should believe what Google claims ...?

[londrum]

im just saying its funny they've been giving away free tools to see how slow your site is, and then start selling something to make it faster.

when was it that they announced that site speed would form part of the algo? i cant remember exactly, but it's been less than a year. the free tools were probably released six months ago (i cant remember exactly).
...when you put all those timings together, the algo announcement and free tools smell like an ad campaign for this product.

the fact that you find WMT reporting different figures to their new product should ring alarm bells.
think about it... imagine if you took your car into a garage for a free MOT check, and they told you that the engine was running slow and the tyres were defalted and the brakes needed to be renewed... but they can make it run a lot better for £100. you might want to take them up.
But what if their free MOT test was wrong? what if their figures were all wrong. You would feel cheated.

That is basically what google are doing. Let's not beat around the bush about it, if their free tool in WMT is reporting wrong figures (...and if their other tool has different figures, then one of them is clearly wrong) whilst at the same time offering a product that can fix it, then what is that, if not false advertising?

last time i looked at WMT a major cause of my site slowing down was google adsense. so you've got a situation where google are offering to sell you a product to speed up your site, when it was another of their products which slowed it down! that is like them peeing on our coat, and then offering to clean it for $100.

the problem with the page speed figures is that they are totally uncheckable. how can you check them? google are taking times from all over the web, all over the world, and combining them. they are taking times with different browsers and different set-ups. you cant duplicate that.

[rlange]

denisl wrote:
Can we assume from this that G may put more emphasis on speed as a ranking factor?

I doubt it. Options for improving the load time of your pages have been around for quite some time; Google is offering nothing new here.

--
Ryan

[freejung]

4.7% improvement here, definitely not worth it even if I would consider pointing my domain at Google, which I agree is probably not a good idea.

If I really want further improvement, I'll put my images on a CDN.

[keyplyr]

When I did the test it said "invalid url, please try again..."

They must be using a method/UA than I block.

[tangor]

I have to wonder why Google wants to get in the "isp" business... then again, I know why... Hopefully, like so many G landgrabs on the net, the webmasters aren't going to buy into this one... One can only hope...

[Bewenched]

... All your Content Are Belong to Us ...

I about fell out of my chair on that one.

Would I want this service.... ummmm... no.

[davec]

I did a test and their optimized page would be 15.1% slower and 7% slower for the repeat - no thanks!

[MarvinH]

I noticed that the static resources from my site are, during the test, being returned from more than one different sub-domain, for example:

3-www-accel-pss.googleusercontent.com

and

4-www-accel-pss.googleusercontent.com

I wonder why they don't simply use one (instead of two or more) sub-domain. Each additional sub-domain necessitates a new DNS look up and a new initial connection. Each new sub-domain thus wastes perhaps 100 milliseconds or so. I wonder if only the comparison test is set up this way, or if also the final product works this way.

[tedster]

Each additional sub-domain necessitates a new DNS look up and a new initial connection. Each new sub-domain thus wastes perhaps 100 milliseconds or so

Using extra subdomains (hostnames) allows the browser to open and use more parallel threads at the same time. This can easily make up for the small loss that an extra DNS look-up requires.

[MarvinH]

Ted, the modern browsers can do 8, or even more, parallel connections, so I would not consider using multiple host names an advantage today. :-(

[tedster]

How does that square with the 40-60 http connections I see required for many web pages?

[MarvinH]

If a page has that many http requests, then I agree. But perhaps a better solution for those webmasters would be to check if there is a way to eliminate some of those requests.

I run a hobby forum (based on vBulletin 3.8.x). My home page currently has only 5 http requests, my forum pages have only 4 requests, and thread pages, if they contain no images posted by members, they also have only 4 http requests. So in my case, I would hate to lose even 50 milliseconds on an unnecessary DNS look up.

[freejung]

Marvin, if your forum is that light it'll be fast no matter what you do. This really gets to be a concern when you're serving lots of images. You can combine and minify your JS and CSS, you can use sprites for your backgrounds, but each img element needs a separate connection, no way around it (at least that I know of, if you know of any I'd love to hear it).

[pageoneresults]

Update 7/29/11: We were notified of a bug in the measurement tool that sometimes causes incorrect measurements. If your results indicated a slowdown on your pages, please run the tests again, and make sure you specify a fully qualified domain such as www.example.com. We apologize for any inconvenience and confusion this may have caused.

Page Speed Service: Web performance, delivered.
[GoogleCode.BlogSpot.com...]

[tangor]

Hmmm... Koolaide did not taste great first time around, must change recipe to keep drinkers happy. Personally, I would question any service as the source of any recommendations...

[brotherhood of LAN]

304 Not Modified is quite fast.

[MarvinH]

Not all sites may benefit from this service. Out of curiosity, I just did a test for www.google.com, and the new service would noticeably slow the site down:

Page Load Time : Original 0.823s, Optimized 1.054s, Difference +0.231s (28.1%)

Repeat View Page Load Time : Original 0.500s, Optimized 0.669s, Difference +0.169s (33.8%)

[Shatner]

Anyone else finding that their test doesn't work? I'm getting nothing from it.

Or are the results just confusing and incomprehensible?

[MarvinH]

Shatner, what exactly in the results do you find confusing and incomprehensible?

The results page shows you, in green numbers, the time the tested page could save by using the new service. If the test finds that your page is faster without the Google's service, then the results are presented in red font, and the numbers represent the time the tested page would actually lose if the Google's service was implemented.

There are two small "before and after" thumbnail images of the tested page. Under each, you can see a link to detailed test results. You can open both to compare in more detail. You will see that several tests (not just one) were performed. You will be able to see load times of each individual component of the tested page.

[londrum]

when i checked my speed against the figures given in WMT, it doesn't even match.

given that both sets of figures are provided by google, why should we trust the speed figures given by their new test?

[MarvinH]

I wouldn't expect the figures to match because the sources of the data are vastly different.

WMT uses data collected from your visitors who use Google toolbar or the Chrome browser. Such data depends on the geographical location of the visitors, and their connection speed.

The webpagetest.org comparison test depends on the geographical location of the test center you select.