NCSA and IBM Part Ways Over Blue Waters

An anonymous reader writes "IBM has terminated its contract with NCSA for the petascale Blue Waters system that was expected to go online in the next year. The reason stated was that NCSA found IBM's technology 'was more complex and required significantly increased financial and technical support by IBM beyond its original expectations.' The IT community is now wondering if NCSA will be renting out space in the new data center that is being built to house Blue Waters or if they will go with another vendor."

Comments

Translation

[andydread]

The reason stated was that NCSA found IBM's technology 'was more complex and required significantly increased financial and technical support by IBM beyond its original expectations.'

Translation: NCSA found that IBM was trying to lock them in with ultra proprietary technology that would have required IBM's expensive services for the life of the installation.

Re:Translation

[Anonymous Coward]

The reason stated was that NCSA found IBM's technology 'was more complex and required significantly increased financial and technical support by IBM beyond its original expectations.'

As usual the /. summary is misleading at best. The actual language used was:

The innovative technology that IBM ultimately developed was more complex and required significantly increased financial and technical support by IBM beyond its original expectations. NCSA and IBM worked closely on various proposals to retain IBM's participation in the project but could not come to a mutually agreed-on plan concerning the path forward.

Other tidbits from the real press release are that IBM terminated the contract, not NCSA, IBM is refunding the money paid to date, and NCSA is giving back the hardware delivered to date.

Translation:
NCSA found that IBM was trying to lock them in with ultra proprietary technology that would have required IBM's expensive services for the life of the installation.

That's a really dumb translation. Nobody expects a supercomputer to be commodity hardware. Just the opposite, as there is no such thing as a commodity supercomputer. Especially this kind of supercomputer, built in part to attain new performance records. When you buy something like that, you thoroughly expect vendor lock-in, expensive services, etc. There's only two or three vendors you can buy it from, and they're all going to be doing a lot of custom engineering for you, so proprietary is by definition what you're buying.

The real translation here is: IBM realized there was no way to deliver on the original contract without taking a huge loss, and tried to negotiate with NCSA for more budget, or maybe reduced system capability, but NCSA couldn't or wouldn't do that. (Probably couldn't, I doubt they can just scare up more money at the drop of a hat. As for backing off, when your project was funded to build a "petascale" computer, you're pretty committed to delivering a petaflop, so scaling back capabilities was probably not an option.)

Since the sides couldn't come to terms, IBM took a huge hit by terminating the contract. Yeah, they get their hardware back, but it's probably not very easy to sell to anybody other than NCSA. And they have to return all the money, which means they did a lot engineering work for $0, once again with few prospects of monetizing the work in a future deal.

As for NCSA, even though they get the money back they still lost a lot too. Years of development down the tubes, and they have to start over (if at all) with a new supercomputer capable vendor. From scratch. At 2011 prices instead of 2007 prices. Which might well be a disaster for them if they couldn't afford to give IBM enough money to finish the original system.

Why did IBM do this, and what next for NCSA?

[bridges]

Pretty surprising development, given the length of time that IBM and NCSA had been working on this. Dropping a contract like this essentially puts into question IBM's costing on future contract bids, so it's not something that they'd do lightly. It'll be interesting to see the scuttlebutt that comes out afterward to see how much of this was technical shortcomings and how much pure financial considerations from IBM. Maybe since IBM already got their big publicity for Power7 from Watson, they're being more profit-concious on future Power systems so they don't tie themselves to margins that are too low.

From the NCSA side, there will certainly be a fallback of some sort - NSF and NCSA are already working out those details according to recent reports. I'd guess that they go with a large Cray XE6 system, given that a pretty sizeable version of that system is already being stood up and ironed out (the Sandia/Los Alamos Cielo system), and Cray has a lot of history successfully standing up big systems (e.g. ORNL Jaguar, Sandia Red Storm, etc.). SGI Altix is the other alternative, I guess, and there's a pretty big one up at NASA now, though that'd probably be a riskier proposition than Cray IMO, and I expect that NCSA and NSF are going to be pretty risk averse on following up on this.

Cue the PERL / Beowulf cluster posts!

[FlyingGuy]

As in I could do what they do with a few lines of PERL and a Beowulf Cluster!!

Re:Cue the PERL / Beowulf cluster posts!

[bigtrike]

How many gigabytes long would each line of perl be?

Not really shocking news.

[Zero1za]

'was more complex and required significantly increased financial and technical support by IBM beyond its original expectations.'

Sounds about normal for an IBM gig then...

Job application

[wirelesslayers]

Now I know why they canceled the job exams (C, Perl and Linux Admin) I was about to do this week for a position at IBM-LTC. =(

Re:NNSA and IBM Blue Gene

[halfdan the black]

IBM does need to drop the price of Blue Gene, BUT Blue Gene is absolutely awesome to work on (I use Intrepid). Almost all the rest of the rest of the "supercomputers" out there like Cray are basically just PC clusters.

Typical

[lucm]

My experience with IBM is that every new software or equipment setup is painful, complicated and goes over-budget, but once things are up and running, it is rock-solid, so in the long run it is still the vendor I would trust the most for enterprise projects. Knowing them, I always take into account the extra oil and time that will be needed to make things go smoothly at first.

This is very different from a vendor like Dell, who takes good care of its new customers (especially the ones with deep pockets) and make sure that the delivery is on time and budget, but after a while problems start to appear (wrong firmware, obsolete drivers, etc) and pretty soon they tend to ignore you if they feel you won't bring new business in the next quarter.

In this case with the NCSA thing, it's a typical situation where budgets have no room for the fudge factor because the organization has a price-driven selection process, which is wrong.

Re:NNSA and IBM Blue Gene

[1729]

Blue Gene is absolutely awesome to work on (I use Intrepid).

Seriously? That's the first time I've heard that. What do you like about it? The buggy toolchain and CNK? The joys of (sort-of) cross-compiling? The I/O bottlenecks? The blazing fast (for 1999) CPUs?

The only way I can see BG/P being a useful machine is either:
1) All you need to do is run LINPACK
2) You're booting Linux on the compute nodes (in which case a commodity Linux cluster would probably be a lot cheaper)