Saturday, March 27, 2004

A Thermodynamic Argument For Google's Continuing Success

Okay, this is going to be a big pile of hooey because I don't really know what I am talking about, but what else is new? I've been fascinated by how Google has done its search engine system. Their data center involves thousands if not tens of thousand of dedicated servers in order to almost instantaneously come back with results for queries from a world of users.

I've also thought about the most likely competitor, Microsoft, and how they could catch up. (I was reminded of this by this article). And looking at it from a "thermodynamic" viewpoint it seems like they cannot succeed. My assumption is that Google uses open source servers, probably running lean versions of unix or linux. These operating systems are well known for stability AND what it is more important, the very low maintenance needed for operating them in farms through scripting. Also, their kernel can be pretty small.

Microsoft would also have to use farms, but they would be hampered by running a version of a Microsoft server operating system. From my experience, the very high maintenance needed for these servers is from difficulty in globally scripting these machines. The MS kernel is a huge beast and a sizable fraction of the energy in these farms is going to go into keeping alive that kernel in memory - at the expense of computing power.

So from an energy viewpoint alone, it would seem that the energy involved in maintaining a MS server farm (from pure joules of electricity to the logistics of maintaining thousands of windows machines - and the linux/unix machines do not need GUI code resident in their memory) can be a magnitude greater than a linux/unix farm. And when we are talking about a business such as search engines - these considerations of energy could be the difference. In a server farm "war" where Google and MS escalate by building bigger and bigger farms - Google might seem to have an insurmountable energy advantage - it will cost less to build and house bigger farms. In any escalation, Google would always win. Of course, MS will have the humongously bigger pockets and they could Reaganize Google out of competition by outspending them - but it might be close! And also, MS could use their monopoly to force better economics from their hardware vendors at the expense of Google - kinda like the way US and the USSR used third-world countries in the cold war.

The other side of the coin is of course is that evolution teaches us that big, energy wasting dinosaurs lose out in the long run to smaller, energy conserving, and bigger brained mammals. So, I think it is more likely that a more sizably efficient technology or revolutionary algorithm will be the winner in the search engine wars - and MS history tells us that scenario is more likely to come from Google or a totally different third party in their garage. MS is not historically known for coming up with more efficient or revolutionary technologies/algorithms - look at Word...


Post a Comment

<< Home