Is Google really a search engine? Is Facebook a social networking site? An interesting article about Facebook’s new open-source datacenter architecture. Google isn’t really a search engine, it’s just a means to track eyeballs for marketers. Facebook isn’t really a social networking site, it’s a means to track everything about those eyeballs for marketers.
The idea that Google and Facebook are somehow competing with one another in the datacenter space may sound odd at first, given that most people are used to thinking of Google somewhat vaguely as an ad-supported software company. But as we're fond of pointing out, Google is essentially a maker of very capital-intensive, full-custom, warehouse-scale computers—a "hardware company," if you will. It monetizes those datacenters by keeping as many users as possible connected to them, and by serving ads to those users. To make this strategy work, it has to hire lots of software people, who can write the Internet-scale apps (search, mainly) that keep users connected and viewing ads. Since the price of Google ads is set largely independently of Google's cost of delivery, every dollar of efficiency that Google can wring out of one of these large computers is a dollar that goes to the bottom line.
Why Facebook open-sourced its datacenters
Why don’t we have the five-computer mentality of the past, with redundant parts and power systems, instead of thousands of commodity machines spinning away? Does the world really need more than 5 computers? Unlike the aforementioned article, a cluster of data centers is not a computer in my mind.
Perhaps we need a biological computer? By the definition above of a data center cluster being a computer, an assembly line could be considered a biological-mechanical computer. What if the assembly line was controlled by a small pond containing neurons? No more data center, no more power consumption, no more cooling issues.
Here’s to hoping they haven’t figured out how to get those neurons to multiply yet… sounds like The Matrix to me.
