Red Shift? Uh…
I just red this piece on Sun’s CIO Papadopoulos explaining his new usage of the word Red Shift to explain why Sun was doing so well. Whatever about the stupid appropriation of the phrase, I was interested to see what he was talking about - unfortunately as far as I can tell, there’s nothing really interesting about what he’s saying and probably some contradictions.
Let’s start with what was apparently his “aha! moment” (honestly, I really dislike that phrase, but they used it in the piece so I’m just rolling with it).
Papadopoulos realized that there are really two different application sets driving computing demand: one consisting mostly of newer Web-facing applications driving exponential growth in both user demand and computing requirements; the other comprising back-end systems that are growing at more historical rates.
Very insightful. Wow. Some websites are growing really, really fast! Ok… well, it’s true at least, so let’s see where he took this great insight. First he decides that the ones that are growing really fast are called “red shifting companies” and the ones that aren’t are called “blue shifting companies”. There’s some really clever analogies going on there, but I’ll spare you.
Papadopoulos argues that red-shift companies will enjoy exponential business growth in the coming years. Blue-shift companies–those whose processing needs aren’t exploding–will grow at about the same rate as GDP, he says.
Very bubbly talk here. If you have a lot of traffic you’re going to grow like crazy, if you are in some olden time business, you’re not part of the big growers. I’m not saying that it’s necessarily wrong, but I’m saying that at some point the twitters and youtubes and myspaces of the world actually have to start making a profit and a big one, to sustain themselves. The traffic’s important, but ultimately these are businesses and we have yet to see these high flyers prove that they actually are businesses.
The piece then lists a bunch of people who discuss their growing compute needs. Everyone’s worried about growing the datacenter, worrying about power and all that stuff.
The solution to all of this?
A shift to lower-cost, lower-risk utility computing, mostly on sophisticated “big iron” servers
This sounds like the opposite of what is going on. People are shifting to commodity computing, what Google’s been pioneering, building their datacenter out of low cost parts, just lots and lots of them. If any of them fail, there’s nothing interesting about that particular piece, just bring a new one online. Every piece is indistinguishable from the next and all of them are low cost components. Sun’s answer to this?
a jet-black shipping container. Prefabricated, equipped with a built-in cooling system, and crammed with 252 Sun servers, it’s one of the first of Sun’s so-called Blackboxes.
Go big or go home. Sun’s always had the heavy metal and I guess they decided they needed to turn that to 11. I don’t know if this is the way people are going to go. It seems to me like the trend I see is moving towards commodity computing - it gives you very granular control over your growth and turns hardware failure from something to be feared and dreaded to something that is expected and easily handled. The advantage of commodity is that the increments for computing are very small - upgrades and evolution are easy.
Moving to bigger and bigger iron reduces your granularity. What happens when you hit capacity on a Blackbox? You have to get another? That’s a lot of extra compute power. What happens a few years after you get the Blackbox as computing marches ever on, those machines are looking dated, the increment for improvement is much larger than with commodity computing.
On the flip side is grid computing as Amazon’s EC2 service. Here you offload all those worries onto someone else. This is a viable alternative, in my unenlightened opinion, to commodity computing. It abstracts the hardware away and you let someone else worry about that. This has the advantage of getting companies off the hook for maintaining their datacenter (moving all of that to a company who needs to focus completely on maintaining it) but removes the ability to really control the environment which some may be loathe to give up. Behind the scenes, though, I still see commodity computing as the way to build such a thing.
I just don’t see any advantage to going with really, really big hardware. But who knows, I always underestimate Sun, though. I mean, after they squandered their massive lead during the dotcom era I kind of wrote them off. I kept expecting to see them going out of business or getting acquired by another player but far from that, they’re slowly making their come back. So who knows how this whole thing will work out for them.
