Google Web Accelerator

This thing is pretty cool. It's sped up forum browsing quite a bit. Only works for IE and Firefox 1.0+ though:

2. How does Google Web Accelerator work?

Google Web Accelerator uses various strategies to make your web pages load faster, including:

* Sending your page requests through Google machines dedicated to handling Google Web Accelerator traffic.
* Storing copies of frequently looked at pages to make them quickly accessible.
* Downloading only the updates if a web page has changed slightly since you last viewed it.
* Prefetching certain pages onto your computer in advance.
* Managing your Internet connection to reduce delays.
* Compressing data before sending it to your computer.

http://webaccelerator.google.com/

I'd really recommend you dorks try it out.

polster wrote:This thing is pretty cool. It's sped up forum browsing quite a bit. Only works for IE and Firefox 1.0+ though:

More specifically, it only works with those browsers under Windows. Mac and Linux (and free/open/netbsd, and who knows what else) Firefox users are out in the cold.

gleam wrote:More specifically, it only works with those browsers under Windows. Mac and Linux (and free/open/netbsd, and who knows what else) Firefox users are out in the cold.

Do you mean Safari?

Read this about that.

Aaron Deacon wrote:

gleam wrote:More specifically, it only works with those browsers under Windows. Mac and Linux (and free/open/netbsd, and who knows what else) Firefox users are out in the cold.

Do you mean Safari?

Them too, but I meant firefox on mac/linux/other OSes. I specifically mentioned firefox since firefox on windows is supported. It's awfully easy to make a cross-platform XPI, so I don't know why they didn't. Presumably they're hooking into Windows in some super sekret ways. A shame.

The link I linked to says the two things I thought-- one, it's AOL circa 1994 all over again, two, if we're mostly on broadband do we need this?

I guess since I'm Mac it's an academic question, but...

More recent news items say that Google isn't anonymizing, they're passing the IP address to the site via an X-HTTP header.

I'm not sure how this accelerates, then, if they're not actually caching, or if the IP-pass happens asynchronously as a courtesy to the site.

Most proxy servers pass a header with the client's IP, usually called X-Forwarded-For. This helps the destination web server count its unique visitors, etc etc. If google is issuing a HEAD request to determine if the page has changed, then they're probably sending the X-Forwarded-For header then..

Most big companies use caching proxies like this, but internally. At my company, which is fairly small, we cut down on bandwidth for web by 10%. Larger companies see much bigger gains. Large files that are cached on the local server are also served to our users much more quickly, so they don't mind that, either.

A lot of big websites also use "reverse proxies", servers that sit in front of the web server cluster and do caching at that level. It keeps the back-end web servers from having to serve up little images and such, and instead focus on the resource-intensive dynamic pages.

The "they're stealing my copyrighted material" fears are silly. Caching proxies have been used for a long, long time -- and the user is, in some cases, completely unaware that they're going through these "transparent" proxies. The dialup services that claim to offer faster page loads than normal 56k send all of your web requests transparently to a caching proxy, which does the exact same thing google is doing.

I doubt using robots.txt will stop google's accelerator from working -- and cmon, let's be serious, do you want to lock out people who are using a google service? Plus, their caching proxy will lessen the load on your web servers, which means lower bandwidth bills and faster response from the machines. Many webmasters do all they can to make their site MORE cache friendly, since these caches are good for pretty much everyone.

You can read much more about caching proxies here:

Caching Tutorial

If everyone went through google's cache, I'd imagine chowhound could finally stop begging for money for their bandwidth bills!

-ed

One more warning about Google Web Accelerator:
It works by prefetching every link on the page while you're sitting and reading the original page. The original intent of prefetching was that only items in the "link" tags in the header would be prefetched, but GWA grabs every button and link.

This can have severely negative effects: Delete buttons get clicked (and javascript confirmations ignored), every item gets added to the shopping cart, etc. etc.

This sucker's still in beta, give it some time to settle down. As it is, I'll probably have to rewrite sections of my shopping cart software to tell prefetch servers to go pound sand.

JoelF wrote:One more warning about Google Web Accelerator:
It works by prefetching every link on the page while you're sitting and reading the original page. The original intent of prefetching was that only items in the "link" tags in the header would be prefetched, but GWA grabs every button and link.

This can have severely negative effects: Delete buttons get clicked (and javascript confirmations ignored), every item gets added to the shopping cart, etc. etc.

This sucker's still in beta, give it some time to settle down. As it is, I'll probably have to rewrite sections of my shopping cart software to tell prefetch servers to go pound sand.

Indeed, that's a bug. It's only supposed to prefetch pages specifically tagged as prefetchable. Firefox/mozilla, but not IE, also pay attention to those prefetch tags. If you do a google search with firefox, your browser will start prefetching all the top results in the background. You can read more about this here.

nb: Caching proxies are more useful when they sit somewhere on your outbound path, since then they actually can save you bandwidth. Google's web accelerator won't save bandwidth at all, but it should help with response time, since I'm sure the GWA is sending you to the closest (in either geographic or network terms) datacenter on their network. Instead of loading that 300k worth of html and images from japan, you're loading them from a datacenter in, I think, Elk Grove. (yes, I know the compression may save some bandwidth, but most web servers and browsers support compression anyway)

And, obviously, google has a lot more bandwidth than pretty much every site on the web, so you may see downloads happen faster, even if you still transfer the same number of bytes.

What I'm most curious about is whether google is using any kind of inter-cache communication. That is, if someone using the Chicago proxy cluster requests a file not in that cluster's cache, does it check to see if any of the other google proxies have the file, or does it go directly to the source?

Another thing to consider is that if GWA gets popular enough it could, essentially, put Akamai and its competitors out of business. Since they work by serving end users with geographically closer copies of static files, like images, Google would be stealing their role. Obviously, GWA isn't likely to ever get the 100% market penetration required to wholly obsolete akamai, but I'd still be a little concerned if I were in their path.

-ed