The importance of innovative infrastructure at Google can't be over-emphasized. If you want to understand how Brin, Page and the top Googlers think, just say to yourself: Is there a way I can radically over-haul how infrastructure works?

Obviously, they've have done this for their computing power. And that worked so incredibly successfully in the following areas:

• Design your own distributed filesystem (GFS)
• Design your own racks for your datacenter
• Pick motherboards and CPUs for your racks only after testing your code on them
• Hire a chef for your start-up
• Get two VCs who don't split deals to split your deal
• Go public on your own terms - getting i-banks who don't split deals to split the deal

Do you think Google bought some package for SarBox compliance? Of course not! They've rolled their own by repurposing their source-code control system for that. Brin and Page are the kings of "rolling their own". That's what they do.

Now let's role-play. Put yourself in Brin & Page's shoes:

You want to minimize the housing / commuting problems for your talent. Do you:

• Pay employees to buy hybrids
• Design your own buses for mass transit
• Open remote work-centers
• Make all of Silicon Valley have free wireless
• All of the above

It's also why they'd buy a 767 instead of a Gulfstream for the corporate jet. You'll know when Google is in trouble when they starting buying off-the-shelf stuff.

¶ 12/30/2005

That post inspired me to do my own interview of Philipp. Below is part one.

Philipp, when did you decide to start a blog mainly about Google?

That was in around May 2003, while staying in Malaysia. I was reading Elwyn Jenkins' Google Village blog at that time and greatly enjoyed it, but it didn't have enough daily stuff to satisfy my information hunger, so to speak... so I created just the blog I was looking for to read at the time.

What did you do before you were a blogger? Was it related to 'outer-court.com' or was that just a domain you had for fun?

In my "job life", I've always been a programmer, even while maintaining my blog. Outer-court.com was just a domain name to host my first "homepage". I started out at (gasp) Geocities in 1997, then moved to Tripod, and a bunch of other free homepage providers. Finally I got so annoyed by them introducing ads and popups, so I got my own server. I guess only when you pay for something to you have the right to expect really good service.

As for the name itself, I was looking for court.com which was taken, so I chose the next best thing. I sort of imagine an "outer court" to be a meeting place kind of thing, so I liked it. It didn't have any specific meaning so it gave me the freedom to put online whatever I felt like at this domain, which later on included my blog at a sub-domain. Nowadays, I wish I would have chosen a domain instead of a sub-domain, but at that time I didn't know how important the blog would become to my life.

When I work on a project, I always like to have the domain name. And I think you have several cool domains, like 'feeeds.com'. In fact, I'm surprised that you don't own 'googleblogoscoped.com'. Why not?

OK, I guess I already answered that one now in the previous questions. The only bonus I can think of really is that with a sub-domain, people aren't linking so much to both www.blog.outer-court.com and blog.outer-court.com. Maybe that's the only bonus of this approach. Today, I believe moving is too much hassle for my visitors because they have to adjust bookmarks, RSS readers, etc. I could just do a forwarding from somedomain.com to my blog, but then again, I'd end up with several URLs people link to, and I like to keep it extra simple.

About how many domains do you own? Do you still look to buy domains, and do have a strategy when you look for new domains?

I don't buy domains unless I have a clear idea what I want to do with them. But I guess due to the experience with my blog, I'm more willing to quickly register something for a new project, just so it's easy for people to enter the URL. There's one thing that's a pain to change later on in a web project, and that's the address.

So let me check... I have 18 domains, plus a bunch of sub-domains.

Philipp, you mention keeping things simple, and many of your posts deal with simplifying the user experience on the web. You even have made several suggested redesigns, although I can't find the post where you simplified CNN and Yahoo and others.

I suppose you're referring to the ultimate reduction usability test, in which you remove everything which you don't use from a given site...

So I'd think that you were a UI designer at one point? Is that right, or are you mainly an expert hobbyist in that area?

I'm just a hobbyist in this area, though I can use this at my job as well. I did design work and usability work in my past. I don't think there's such a thing as "usability engineer" or "interaction designer" in Germany. Often, the job descriptions are split up into concept, graphics, management and programming. I actually prefer the synergy of overlapping teams and skills in a job. Usability really is common sense, but often people don't trust their instincts and this results in overtly complicated sites... or they're just not users of their own system.

Usability to me starts with the layout, the way forms and links are set up (As Spolsky says, every time you're presenting a button or link or checkbox, you're asking the user to make a decision), continues to the way a page loads (what loads first? what will be displayed while loading? will the loading progress look smooth?), and then down to the level of HTML and CSS.

I believe there's a paradoxical difference between something that looks good and cool and professional, and something that's usable. Take the Google logo, for instance... every designer will tell you it looks like crap (pardon my French), a very amateurish design. And I agree it's kind of childish, kind of amateurish. At the same time, people love it. Sometimes simplicity and some imperfection in design can lower the psychological barrier for starting interactions. I believe if you would have presented a flashy portal to someone in 1998, and you would have presented him the Google homepage, the portal would have appeared to be much more hip, cool, and technologically interesting to people. Of course recent internet history taught us a different lesson.

Do you think the web overall is improving, in terms of simplicity?

There are different trends. While some sites are putting effort into making things simpler -- apparently inspired by Google's success -- there's also a trend more towards portalitis. Ironically, Google now is starting to be part of that trend (just think of the Google Personalized Homepage, and its new API). But overall, I think the trend is towards simplicity. For example, there's search.yahoo.com, there's a new simpler Amazon product layout (I believe it's still in a test phase).

Still, not every company is "getting it". You might think it's obvious that simplicity is great for users, but then just go to look at the web reality and compare Yahoo products with Google products. If you analyze every link, every headline, every button, you will quickly see that only one of the two companies is completely user-focused at this point.

Another area of design you seem to be very interested is in digital character design - especially sprites and small logos. I'm thinking of the the characters in your chat room and the little logos you create for some of your frequent commenters. I also remember that you gave away logos just to test the services that create logos on the web... Can you tell me more about digital characters/logos, and your interest in them?

I like to design and draw. Logos are one way to be creative on a website. I also like to draw comic book characters, cartoons, stuff like that. When I was younger, I used to read a lot of comic books. I read Scott McCloud's "Understanding Comics" as a teen, and that made a big impact on my thinking... you can even apply his theories on usability, or programming. This comic book on comic books is really a great read, and it has a broad perspective, going back to social interactions of cave men. Really, you can apply everything to anything. Painting, programming, making music, and whatever.... these are all different ways of expressing ideas. Sometimes out of the synergy of these different areas, you get new ideas, which might turn into new sites or products.

¶ 12/29/2005

Despite what many people think, Google is still on the upswing, and will be MORE dominant in 2006 than they are now. Why? Simply because Yahoo is going to get weaker in the search-driven ad space, and Microsoft is not going to be a serious competitor.

The search-driven ad and lead marketplaces are "winner take all" with respect to inventory. Google is sewing up more inventory (e.g. more AOL) while their main competitors are splitting up the same inventory.

Google has the biggest liquid market for search ads, and for contextual ads (i.e. AdSense). By a LOT. And their lead will get bigger in 2006.

And if the ad market softens, guess who is still in the best position? The one with the most inventory. Google's black-box pricing flexibility doesn't hurt either.

The inventory advantage is huge because it creates more relevance. More places to show ads means more advertisers and more creative (ads) to show. It's easier to put good, relevant ads in the right place.

As Yahoo's contract with MSN expires and Yahoo loses the right to display ads on MSN in 2006, Yahoo becomes less attractive to advertisers since it has fewer impressions/clicks to sell. Yahoo will also revamp their ad system which will confuse people for a while.

Microsoft doesn't have the inventory clout of Yahoo or Google, and their current ad system is not very well understood by the SEM world - i.e. no one has figured out how to do better with MSFT ads than they already do with Google.

The fact is that most businesses and SEMs just don't have the time to muck around with Yahoo and Microsoft today. They spend 80% of their time and energy on Google.

¶ 12/29/2005

Here's Google:

Your search - "billy ray powers" googlebot - did not match any documents.

Here's Yahoo:

   1. Got Ads? - Google AdWords and Overture PPC Open this result in new window
      Got Ads? covers pay-per-click marketing and Google AdWords and Overture ad
 technology, including programming the Google and Overture APIs. ... might go to
 Google and search for: ["Billy Ray Powers" snowzilla] since "Billy Ray Powers" is
 the name ... faster to the USA Today than Googlebot today. It'll be interesting to ...
      RSS: View as XML - Add to My Yahoo!
      gotads.blogspot.com - More from this site - Save - Block

Go on with your bad self, Yahoo!

¶ 12/23/2005

Here are some of mine by month. Many of these I've already blogged about.

January: Google Calendar releases. Integrates with Gmail by sucking dates out of your mail and into your calendar.

February: Google Wallet or "Cash by Google" gets pushed out. Designed mainly to let you buy stuff from Froogle vendors.

March: Google buys a music startup like Pandora.

April: Yahoo releases revamped Y! SMS (aka Overture) so the auction model works like Google, based on relevance. MSN takes away a bunch of inventory from Yahoo (i.e. MSN itself).

May: Ask Jeeves is re-branded / renamed. Still remains irrelevant, however.

June: MSN and Yahoo fight for more inventory, buying up properties, but they struggle and Google is more dominant in PPC than ever.

I feel dirty, doing these cheap predictions posts.

¶ 12/23/2005

The problem with being number one, is that you attract the most spam. It's extremely lucrative to be able to spam the number one search engine. Google works extra hard to ensure that most spam doesn't get into the results.

But Google might not be as "responsive" to new stories as other search engines. Perhaps they are more careful and slower because they need to be more vigilant about limiting spam.

Here's an example: today the AP runs a story entitled: "Alaska Man Builds 16-Foot Snowzilla". But when you click on it, but there's no picture of the Snowzilla!

If you do the same search at Yahoo, you'll find the USA Today story, which does have a picture.

Perhaps Slurp is just faster to the USA Today than Googlebot today. It'll be interesting to see which bot picks up this post faster...

¶ 12/20/2005

Early talks were done in around 2029, as you may know. It was very important to us that Microsoft, even though a relatively small player at that time, would not bring their own corporate culture into the Googleplex.

And my favorite question that Philipp asks the holograpically projected Page:

You were once quoted as having said, “Asimov’s three rules suck.” What did you mean by that?

And I can't resist the comic brilliance of this one on how the 20% time project has evolved: "We’re actually considering a 20/80 rule for our engineers...". Read it all, as they say.

But for pure one-shot sublimity, I find this classic 'future trademark' is hard to beat:

One-Click TV™

In about 20 years from now, the internet and the TV have been completely merged. That means full interactivity for that innocent little box which years ago invaded humanity’s living rooms. A side-effect not even big media companies did foresee was that most people were actually quite happy with the ease-of-use of sitting on a couch and clicking through channels of mindless TV shows...

¶ 12/15/2005

Just wanted to update Carville for the modern web.

¶ 12/15/2005

Google's Base is one of the least-ready products they've ever launched. Did they not foresee that it would be massively appealing for spammers? Or maybe they did? It happens to be down right now, as I write this billet-doux. (It was out for about 3 mins total...)

Here's what the best use for base is: A spam honeypot. Google should basically take the logins of everyone who's uploaded more than 50 items, do a big WHOIS grep through their registry database and penalize every one of the websites of the uploaders in the main search index.

In other words, the only people using base so far are affiliate and porn spammers. Google could use that knowledge to increase the quality of its index spam filtering.

Just a thought...

¶ 12/14/2005

Paul Kedrosky asks why Alexa's Web Search Platform is a big deal. He wants to know why this is any bigger/better than what they had back in October.

Simply put, AWSP allows you to do computation on Alexa's computers. Those computers have the internet itself (> 100TB) in local storage. Back in October, Alexa's AWIS (which Paul referred to) only allowed you to access the data, and not do computation.

Before: Similar to Google or Yahoo search API - you can do a search programmatically.

After: Now you can write a C function that processes every page of a search result - and run it on Amazon's / Alexa's boxes.

See the difference? Amazon is creating an open San Diego super-computer center where anyone can schedule jobs.

Key concepts:

• Grid Computing Framework: Any programmer has access to a supercomputer environment similar to what an employee of Yahoo or Google would have.
• Internet in local storage: Alexa's crawls the whole internet incrementally. Your program can now have fast access to every document - and all the metadata (like links, file types, etc.)
• Web Services: If you build an application, you can publish access to it through an web service.

I think it's a big deal. We'll see how many people can meet the pretty high technical skill standards to start using it. I also think that even though access to the data, storage, compute time and bandwidth is all priced a $1 per unit, it could be a bit more expensive to do stuff than people might think.

¶ 12/13/2005

SiliconValley.com sponsored a discussion on the "Google Network" and there are some very informed comments in there about how Google's infrastructure is a competitive advantage.

After I posted my "Google skynet" comments on Robert Cringely's recent articles, I got a surprising dose of skepticism in several emails - "I'll believe it when I see it." was one theme.

A lot of people seem to doubt the veracity of Cringely's report. I was a bit stunned by this, because I think most experienced Google watchers know that Google's infrastructure is a huge part of what got them where they are today - these are the kids who designed their own racks so they could cram motherboards, disks and fans together pell-mell; then they built their own distributed filesystem.

I believe it's obvious that Brin and Page would build the next quantum leap computing architecture by extending how they leverage cheap hardware, dark fiber and worldwide reach. It's a logical progression. And they are really good at logic and logistics.

Don't you just know they'd love the idea of using helicopters to fly a datacenter from the ship right to the power station? Ship it over, plug it in, and turn it on. No different than installing a rack, really.

So I'm glad to see that the comments in the discussion group show that several others do see the outline of this type of strategy, and that Larry Page especially would be pre-disposed to an massive infrastructure advantage. This quote below is from David Vise, author of The Google Story

The dark fiber Google has been buying is for slashing peering costs per an edict from Larry Page, I'm told by Googlers. Larry has a penchant for saving money through dark fiber and looking for ways to reduce the cost of electrical power, another major consideration in the equation for Google as it deploys computers in data centers around the world.

And to those who doubt that Google would be building something like this, Page would probably shrug and say: "They'll 'get it' when they 'get it'".

¶ 12/08/2005

One more advantage Google has once it builds a world-wide super infrastructure, and creates a parallel internet is: Google will be the only place that has trained developers who can write applications on this kind of platform.

In my original post on their 'skynet' platform, I was a bit harsh when I wrote:

So while the public side of Google flounders around and creates barely passable beta me-too products like Base, Talk, Reader etc.,

I still think the apps are kinda weak for a company who promises "Great just isn't good enough", but now I realize that by the time Google deploys it's portable datacenters and builds out its bandwidth, the 24-yr olds who pushed mostly broken Google Base out the door might be experienced 27 yr-olds who can create something incredible.

And certainly Google is one of the two or three places right now that can train programmers to work on such an infrastructure. Yet another massive strategic advantage...

¶ 12/07/2005

Pandora has built a very cool app for discovering new music, based on their music-genre categorization technology. It's obvious that they are Flickr-type acquisition target.

Google should end up buying them. They'd be a great fit since Pandora has a unique technology in an area where Google is not already proficient. Unlike Kozoru, a search technology in an area that Google would rather develop in-house, Pandora adds something to the Google arsenal that Google isn't already trying to build.

Google could use Pandora as a base to take on iTunes, if they wanted to.

¶ 12/07/2005

Cringely breaks the silence on the rumors of a major Google innovation - worldwide datacenters in a box (actually a shipping container).

For the past year, Google has been secretly installing shipping containers full of Google CPU racks around the world. Cringely seems to think they are just starting, but I think they've been at it a while.

Google has several logistical problems which they solve ingeniously. They can't just go to the middle of Poland as Google and buy up the output of a small power station. No - they must secretly operate so the locals won't know its Google. They get the rights to the power they need to run and cool 4,000 CPUs in a box, without paying through the nose. They are doing this all over the world.

Memory is a similar problem for Google. They buy so much of it, that they need to launder their purchases so not to affect the market price. They need a "beard" to do the orders for them. Otherwise, they'd affect the prices, and pay too much. Just like the big institutions on Wall St. trading stocks. There is no liquidnet yet for Google to buy memory and power from, however.

Where does this lead? Simple way to put it: Skynet. (that's a shorthand analogy - I don't think the system is likely to decide to eliminate humanity...) It will be a system that delivers worldwide, incredibly powerful computing and bandwidth and is impossible to take down.

Cringely makes a nice analogy to Wal-Mart's real-estate strategy and an 'parallel internet' and I think that's true, but it's not emphatic enough. No one else will have the ability to deliver computing power like Google. No one besides Google is developing the OS to run such an infrastructure either.

In the follow-up article, he gets into the flow of what could be, envisioning a consumer box connecting to the Google skynet.

As a result, Google becomes overnight a major phone company, a major video entertainment provider, a major player in home automation and even medical telemetry.

Major is an understatement, since Google will probably plant several hundred of these boxes worldwide, and enable new markets for these services across the globe. Cringely sums up nicely:

Microsoft can't compete. Yahoo probably can't compete. Sun and IBM are like remora, along for the ride. And what does it all cost, maybe $1 billion? [Cringely later revises his estimate to $3B] That's less than Microsoft spends on legal settlements each year.
Game over.

What really intrigues me, beyond the incredible 5-year advantage this will give to Google, is how at a company with 5,000 smart people, a very small group creates and executes this genius-level strategy. There's probably 10 guys who are the brainpower behind this supercomputing platform.

So while the public side of Google flounders around and creates barely passable beta me-too products like Base, Talk, Reader etc., don't ever think that Google is running out of mojo. Bubble or not, Google has the chops to pull this off, and become the most important business in the world for at least 10 years. Is that worth $1000 / share?

¶ 12/06/2005

If you try to install an extension in Firefox 1.5 and get a warning that tells you "Software Installation is disabled", you need to use about:config to set the xpinstall.enabled variable to true.

Found by: this comment (ironically on a microsoft forum) solves the problem when you have Firefox 1.5 and you try to install an extenstion.

¶ 12/06/2005

A while back, I wrote that Yahoo needs to change their ad model to work like Google's. They need to change because Google's model works better, and makes a lot more profit. At least, Yahoo needs to have:

• a black-box auction that obscures how bidding affects ad placement
• incentives for more relevent ads, including tying click-thru-rates to ad placement

Of course, Yahoo realized this before I wrote about it and is working on it. But they are taking forever to get it working, I think. My prediction is now May '06 for Yahoo to have AdWords-like system rolled out.

Get Ready.

¶ 12/06/2005

Firefox 1.5 has issues. Wait for an update if you haven't upgraded already.

Among the things I've noticed:

• You can't install the developer toolbar extension - the necessary "software installation" option doesn't exist in the Options panel
• Changing the width of the browser window doesn't always cause a resize, so the little orange x in the tabs can disappear
• Table rendering order can cause odd effects

Of course, I'm assuming you are already using Firefox 1.0.7 or something. If you are still using IE, then just get Firefox 1.5 now and breathe the free air...

¶ 12/05/2005

Web Guerrilla has "a plan for click fraud" - i.e. he's gonna commit click fraud in the name of science.

The only practical thing to do would be to try and collect some objective data that would shed some light on how big of a potential problem click fraud may be. By that I mean auditing the performance of the various fraud detection systems being used by each of the PPC engines.

And that’s exactly what we are going to do.

Here are my predictions for his results based on the past year of doing click fraud investigations for several clients:

• Google will catch the unsophisticated attempts.
• Yahoo might.
• Google's overall content network click fraud rate is somewhere around 4-5%.

I believe click fraud is like any other statistical fraud, a fraudster can get away with it for a long time if they are careful and below the radar.

Having said that, if we don't hear from Greg in his promised timeframe, you can assume that he's found a lucrative way to cash in on the search engines lax enforcement against click fraud.

¶ 12/02/2005