Comments on: So what does Google use when they aren’t Googling?

By: Wes Felter

Wes Felter — Tue, 31 Jul 2007 18:18:57 +0000

Maybe I’m just cynical. I’d like to see systems that scale up and down, but fundamentally, scalable systems require extra engineering compared to non-scalable systems and that cost will usually be passed on to the customer. Just two recent examples: EC2 doesn’t scale down because it can’t run off-the-shelf software stacks (see http://news.ycombinator.com/item?id=37720) and Isilon doesn’t scale down because it appears to cost 2X as much per TB as Thumper. I imagine that Seval will also be charging a premium for their value-add, whatever it is. Also, non-scalable systems are getting bigger (e.g. 1Us now have 8 cores and 16GB RAM, Thumper is 24TB and probably soon 48TB) and thus pushing scalable systems up-market.

If there is hope, maybe it is in BigCo-subsidized open source like OCFS2, Lustre, and Solaris pNFS.

By: Robin Harris

Robin Harris — Mon, 30 Jul 2007 22:37:33 +0000

Wes,

How would that fit with “don’t be evil. Bwa-ha-ha-ha?”

I think the opportunity is for companies to offer products that enable the cluster computing and storage model at much smaller scale. Nearline storage clusters could be huge for backup and FIFO buffers for data on its way to archive. Is that what IBM Almaden is doing with the spin-out?

Robin

By: Wes Felter

Wes Felter — Mon, 30 Jul 2007 18:42:52 +0000

If you are conspiracy-minded, you might even wonder how much of the information coming out of Google is propaganda designed to lure competitors into an infinitely scalable build-it-yourself morass that looks great at warehouse-scale but actually costs more at small scale.

By: Open Systems guy

Open Systems guy — Sun, 29 Jul 2007 02:16:20 +0000

I read an interesting interview with Google’s CEO Eric Schmidt in Wired that answers some of this question- either way, it’s a good read. http://blog.wired.com/business/2007/04/my_other_interv.html, text follows:

“When you got here in 2000 it was a relatively new company run by people (Larry Page and Sergey Brin) who were at that point probably 26 or 28 years old. How did you convince them that this (inviting other executives to Google, soliciting their management advice, and installing a more systematic approach to running the company) was a good idea?

I don’t know. When I came here, I came because I liked Larry and Sergey. It was an interesting little company. I had no idea it would be successful. It was not broken. But we needed leadership.

For example, we had an accounting system which was an Intuit based system designed for five users, and they were using it for 20 people. It was too slow to use, so I suggested that they implement an Oracle system. It was a huge crisis. We ended up spending $100,000 for this. Larry and Sergey nearly had a cow over it (because they thought it was so expensive). A hundred thousand dollars is the cheapest Oracle system ever implemented in history I think.

How did you convince them that you needed to do this?

Well, it was actually very interesting. Larry and Sergey suggested that we should build our own, because most of the existing accounting systems weren’t any good. And I said, “I’m sure that’s true, but you’ll never get it audited,” And I thought that was a pretty clever argument. The auditors would never pass financials (generated out of software) that we built ourselves. And Larry and Sergey today will complain about the Oracle system, but they’ll also say “We had to get one that was auditable.”

I’ve always said that if the company were founded today on an empty lot, we would build the buildings brick by brick. We can’t imagine someone else building our buildings, we’d have to build it ourselves. This is a build-it-yourself culture. The good news is there’s no free land, and so we have to rent the buildings, rather than build them. But the culture is around building things. In that sense, by the way, it’s similar to some of the companies (Intel, Dell, Sony) that I mentioned earlier.”