Thoughts on Google Fiber: Why is Google doing this (Part 1 of 3)

First, I assume I don’t need to explain the broader picture – Google wants to encourage the use of the Internet, especially to replace traditional media distribution systems, because Google stands a much better chance of making money if you watch video online than if you watch it on a TV. So, in that sense, a faster Internet means more use of the Internet, which means more money for Google.

There’s a deeper motivation for Google’s actions – simply put, they’re building a new Internet. They’re not starting from scratch, as some academic projects are considering, but instead they’re building incrementally. Some of what they’re building is parallel to the existing Internet, and some of it is through tweaks to the existing Internet. This is par for the course for Google – see Android, Chrome, ChromeOS, App Engine, and many, many more. But their work on the Internet is particularly interesting. Consider their moves in the Domain Name Service (DNS.) Google operates their own, public DNS Servers, and has proposed extensions to the protocol that would give a DNS server more information about what client is ultimately making the request. Both changes would allow Google more of an opportunity to where traffic goes on the Internet. Even better for Google, neither requires end users to make a change – for the most part, network administrators can make the change, who as a group are much smaller, more technical, and more Google-fawning than the population at large.

Where are they directing this traffic? To Google’s private Internet. In the 90s, in the dot-com boom, there was a huge explosion of installed fiber optic cable– way more than was needed. After the crash, Google (and others, including for example UW-Madison) bought up much of this cable, and likely laid a bit of their own to fill in the gaps. Some people refer to this network as GooglesNet, and they use it to connect their own data centers together, and to tie into larger ISPs directly. This has been a boon to Google; for example, it’s believed that Google doesn’t actually exchange much cash for YouTube video bandwidth, because they have agreements with ISPs to deliver it directly and they can bypass the large backbone providers.

That then brings us to why Google is considering getting into the ISP business: a serious roadblock to innovation is the exchange between Google’s private network and the ISP. By controlling both sides, Google can rapidly iterate and roll out new ideas quickly. Bob Cringley mostly gets this right, but I disagree with him a little bit: Bob thinks Google expects other companies to come up with the idea of what to do with all the new bandwidth, I suspect Google has the hubris to believe that it will be the primary driver of new applications. Cringley is certainly right about Google wanting to get deeper into the ISPs’ networks, though I don’t know that they’re going to go as far as actually placing data centers at the neighborhood/city level, even if they are only as big as one shipping container.

Placing data centers that close is important for two reasons: latency and bandwidth. Latency doesn’t seem to be that painful in the current Internet, especially with smart clients, and HTML5 will only make that better. Bandwidth matters if the connection out of the city is limited – that was the thesis that Brilliant Cities, Inc made at the last Broadband Telecommunications Regulatory Board meeting. In their model, deploying local fiber is great, but ultimately is a waste because everyone with their 1 gigabit connection chokes when they try to leave Madison and access the Internet. (IE, if the Internet were streets, we can have a lot of cars out on the road at once, so long as they’re all on different side streets. If everyone is trying to get onto the I-90 onramp, traffic is slow.) In Brilliant Cities’ model, and in Cringley’s model, there is a local data center, so most of the data is served without having to get on to the Interstate. (The Interstate is actually a good analogy – the UW’s Internet connection runs parallel to I-94, and you can see a couple of repeater stations fenced off next to the highway)

Nearly all ISPs do something like this already, forcing all outgoing web traffic to first go through their own web proxy so they can cache popular items, and we’ve been doing that for 15 years. The problem is many things aren’t easy to cache with today’s protocols.

For Google, having the ability to place data centers deep in the network, or as far in as they choose could be very interesting, or they may have enough bandwidth on their private network to handle the traffic going back to their major data centers. Regardless, when they have the control, they can experiment and discover what works and what doesn’t, and can rewrite their applications to support whatever new features they deploy in the network. They can also redefine what it means to peer with a network. There is some really cool work out of the other UW (Washington) looking at sharing data between data centers. If Google gets the interface right, they can replica it with other ISPs, and they don’t have to try and run ISPs everywhere. Sometimes Google gets into things just to set the direction, too. Its made noise about getting into the 700MHz radio auction, and promised that if it won the bid it would run the network in a very open fashion. Ultimately, they decided not to bid, but not before the FCC incorporated many of Google’s ideas into the bid requirements.

Ultimately, what’s the big application that Google wants to get in on? Personalized, in-show TV advertising. It’s insane that the same ads are shown to every person watching the show, but there’s nothing better infrastructure-wise today. Even just with the basic demographics it would get from knowing the name on the cable bill, Google could do a better job of not showing me ads that I will convert on anytime soon (like, for example, AARP-branded insurance) – but Google can do much better than that. By combining search history, data from applications, and the rest of the data exhaust that we leave behind every day Google will know nearly exactly who’s likely to be watching the TV show, and what ads make sense to show. The $70 Billion dollars that are spent on TV advertising could be far more effective, and advertisers could know and test exactly what works and what doesn’t.  It could be a tremendous economic boon to the country. If Google builds a network that can deliver, advertisers will demand it from the other network providers. And because no one comes even close to Google in this area, the other ISPs/cable companies will naturally partner with Google, giving Google a cut without Google having to run the network itself.
And by the way, if the above paragraph didn’t make you nervous about privacy, it should. I, like others in the field, are starting to struggle with the implications of what we’re able to do. To really have your mind blown, read this, and consider turning your cell phone off.


4 comments so far

  1. Humana Walmart-Preferred Rx Plan on

    Interesting article. I think google plans to monopolize the internet with this move. I can’t say now whether it is bad or good. I think good will be the more appropriate term to use for now.

    As of late, Kansas city is the first to receive the first google fiber. The lines will be installed between Kansas City, Kansas and Kansas City, Missouri, creating a solid backbone which later will branch out to all Kansas City consumers on both sides of the state line, providing download speeds more than 100 times faster than current broadband solutions.

  2. Race on

    What I find so inienesttrg is you could never find this anywhere else.

  3. click here on

    I’m shocked that I found this info so easily.

  4. Joul on

    That’s an expert answer to an inrnetstieg question

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: