High performance computing is a tough business. Customers always want more performance. Few want to pay more for this performance. Many vendors want to serve this market but precious few are pushing more than boxes. Its easy to push boxes, and the really low end vendors, likely having been burned as suppliers of low end windows machines, decided to work on clusters.
Occasionally there are some good ideas. Some really good ones. And they languish for a while until some money can help make them a reality.
Many of us have been talking about the era of personal supercomputing since 1998-ish onwards. The SGI GenomeCluster product was a great example of this. A simple interface to NCBI BLAST that ran it on a cluster through a queuing system. But it was interactive, basically a command line tool. And that was the idea. End users want to think their clusters are interactive, are things they can use when they need. That power is immediately available. Or at least seemingly so.
This idea was one that I and a few dedicated others pushed. It was a good one, and it still is a good one. Give an end user a great deal of computing power on demand. No, not the IBM marketing bit, but make it really easy to get to. Make power available without thinking about it.
Along came Orion MultiSystems. Their idea was put a cluster of lower power consuming, and unfortunately, lower power chips together and give it a simple interface, and make it available somewhat inexpensively. The basic idea, a cluster available on demand (again, not the IBM marketing bit), is a great one. It was the implementation that left some of us wanting.
Well, not just that. The pricing. A 12-way cluster would run about 10k$US. Neat, huh? Until you look at the performance of these 12-processors. Compare it to a dual dual-core Opteron. The Opteron would be a shared memory machine. You could put a sizeable IO system on it. Not the Orion. The dual core machine would run under 7k$ for a comparible memory config, a better IO system, shared memory, …
Lets not talk about the 96 processor unit for 100k$US. This was obviously not designed to be a volume play.
So Orion had some money from VC’s. They were on the board. Like other startups, they were burning through cash while trying to build a market. They didn’t have a natural market niche, and a ready made customer base. They had to build that. Which meant they had to evangelize and teach their customers. Which is even more of a risk.
But it was a bold unit, maybe not the best designed, but v2 would be much better. Their choice of launch partners wasn’t the best either. None of the partners could drive significant volumes, which is what they needed to grow revenue.
Again, there is the coolness factor in having a cluster on or next to your desk. Others are doing a copy-cat right now. Well, sort of. Some of those copy cats are going to be really interesting BTW.
I had heard from various sources that after the first RIF at Orion, that the VC’s had said something to the effect of “not another dime”.
The best way to kill a good idea is to starve it to death. Then again, if the idea is bad, it is a waste of money to throw good money after lost money. Trasmeta’s decision to exit chip making foretold the problems that Orion was going to face. This forced an expensive complete redesign at a point where capital was badly needed to grow the market. I don’t fault the VC’s. They probably made the right decision in the end. Orion’s idea was great. Just not their implementation.
Personal supercomputing will take off. Supercomputing is a rapidly growing and very interesting market. The demand for computing cycles is far outpacing the supply. Creative products are coming onto the market to significantly increase the performance and price performance, not to mention the power consumption, and related. Orion pushed an idea that has been around for a little while, and this idea will survive Orion’s demise.
Hopefully capital won’t be harder to come by for supercomputing ventures. This market needs fresh ideas to turn lots of old markets around to newer concepts and excite customers. Fresh ideas die without the capital to birth them, and a steady, sure, experienced hand to guide them along the path to market. Both have been lacking.