No, Google doesn’t intend to become a national Internet Service Provider, despite its new plan to build a number of optical networks to serve homes and businesses at up to one gigabit-per-second. The real plan is half Xerox PARC and half Tom Sawyer.
When the Computer Science Lab at Xerox Palo Alto Research Center was organized by Bob Taylor in the early 1970s to revolutionize computer, network, and printing technology, there was a conscious decision to live 10 years in the future. The CSL would build devices that could be expected to make economic sense in 1980, not 1970. This was a huge leap, because it meant the amount of memory in each device would be 64 times as much as made economic sense in 1970 when 1K was a lot. Yet think of it, a 64K PC was the norm when IBM introduced that product in 1981 (base was 16K!) so the numbers were about right. Only by embracing future limits, no matter the cost, was PARC able to achieve so much (Ethernet, Graphical User Interfaces, laser printing) in its first three years of operation.
Part of Google’s inspiration, then, for building-out a few residential and business optical networks is to do the same thing. Because not all smart people work at Google and even more so because the smart people who do work at Google don’t generally think or operate like the rest of us, it will be very useful to see what normal folks actually do with that much bandwidth.
There will be a few surprises, I’m sure, but not many. For the most part Google is hoping to inspire current ISPs — mainly cable companies — to follow its lead, like Tom Sawyer did when getting his friends to whitewash that fence. Google wants to set an example for how to do local networks right and get the Obama Administration to codify that methodology through the Federal Communication Commission. Then they want someone else to do the actual heavy lifting.
And it will probably work, not so much because Google is brilliant but because the cable TV companies are ambitious. We’re entering an era where cable operators will have a real cost advantage over telcos in expanding residential bandwidth, thanks to DOCSIS 3.0 modems.
I’m the third DOCSIS 3.0 customer in Charleston, South Carolina and the first residential customer following two law firms. I did it I suppose to write this column but even more so because I have some heavy video activity coming-up and thought I might need the extra bandwidth, which is substantial. The important thing to understand about DOCSIS 3.0 technology is that it’s not a big deal, really. It’s just channel bonding.
Where earlier cable modems had users on each subnet sharing a single analog video channel (generally channel 80), DOCSIS 3.0 devices can grab several channels and aggregate bandwidth. Think about it, under this scenario if a cable system operator were to abandon its analog signal entirely in favor of a total IP solution that would mean a 100X increase in shareable bandwidth on each subnet — subnets that are already for the most part interconnected by fiber. That’s 30 gigabits-per-second or more to be share in your neighborhood alone for a cost that amounts to about $300 compared to the average $1350 per customer Verizon is spending to install FiOS fiber.
Some cable companies will use DOCSIS 3.0 to take down the local phone company, which will be hard-put to compete. And they’ll have support from the TV manufacturers as well as cable box makers. My new 58-inch Panasonic Plasma TV has an Ethernet port on the back and all Panasonic’s competitors would like us to buy new TV’s too. And don’t forget who is America’s largest maker of cable boxes — Cisco. You think they don’t want IP TV? Heck, they trademarked the term.
While what Google intends to install is fiber (or so they are saying right now) the ultimate beneficiaries of this project may be more traditional cable plants running mainly thick old coax.
Google wants to nudge this along because their ultimate goal isn’t to be an ISP but to live in the data center of the ISP providing us data with ads to go with it. They want to drop one of those shipping container server farms into the parking lot of every cable head-end in America, ultimately providing gigabits of data without having to pay anything for bandwidth.