Imagine a Beowulf Cluster of Penguin Computers

Pii writes "News.com is running a story about Penguin Computing acquiring Scyld Computing, a company founded by Donald Becker, of linux ethernet driver and Beowulf cluster fame. Becker will stay on as Penguin's Chief Technology Officer, and the companies claim they don't expect any layoffs as a result of the merger."

Comments

Good.

[Usquebaugh]

Prehaps now they'll have enough time to get the DEC Tulip driver working 100%

lay-offs

[Rosco P. Coltrane]

the two companies don't expect to lay off any staff, a representative said

Right, so they'll have 2 PR divisions, 2 marketing divisions, ... The key word here is "expect". My former company's management didn't expect layoffs either, but funnily enough they did 6 rounds before sinking completely.

This said, Donald Becker is cool, Penguin Computing is cool (I toyed with an alpha box from them for a while and I was very inpressed), so I reckon the result should be uber-cool.

Re:lay-offs

[BlueShades]

Right, so they'll have 2 PR divisions, 2 marketing divisions, ... The key word here is "expect". My former company's management didn't expect layoffs either, but funnily enough they did 6 rounds before sinking completely.

If you read the article, you would have seen that both companies are keeping their perspective names and business model the way it is. The only thing that is merging is the cluster technology on their systems to compete against bigger companies. Doing this, both companies will most likely create a department for testing the clusters on their hardware and providing hardware, software, and support in one bundle to the customers. By trying this move, the company can brag about selling everything you need for extreme computing.

It reminds me of the Oracle radio commercial:
We have the violinist, pianist, and the guitarist.

In this case: The Penguin, The Wulf, and The Beast

Re:Anyone actually use a beowolf cluster?

[paitre]

Protein folding, for starters.
The point you're missing is that a beowulf class supercomputer is -MUCH- cheaper to put together than some "super-powerful" single machine (forgetting that most of the -really- big machines are themselves clusters of a sort).

Money played (and plays) a rather large role in the whole COTS supercomputing arena :)

Yes.

[mhore]

We have 2 that we use for simulations in the Physics department. Well, i say 2...the 2nd one hasn't actually been assembled yet.

The reason we use the beowulfs is because the problems are such that they're easy to break up into chunks. Consider the "system" (in our case) to be a cube with stuff inside that we want to process. We can break the cube up into smaller cubes and process those chunks, and then reassemble (must like is done with folding@home, seti@home, and the like).

The difference? Well, instead of a program taking a day to run, it will take a few hours. Instead of taking a week, it may take a day.

Sometimes a problem doesn't require a Beowulf. If you require several simulations, and the total CPU time amounts to 1 month...then you'd do just as well running the simulations on nodes and just waiting a month. If you REALLY want to know what's happening with a certain set of conditions, it's often times very useful to use the cluster to find out in a matter of hours.

Mike.

Amen to that.

[pr0ntab]

I've got about 9 plushie penguins sitting on my desk (hint, that's how many Altus's we have running our nasty compute jobs)

I am QUITE CERTAIN this merger will mean even more gnarly Penguin configurations!

Sun better watch the hell out. (go ahead fanboys, flame away)

It's always the cheapskates who complain.

[hndrcks]

Ahh... cheap shit Tulip cards and RealTek 8139s... now those are the kinds of quality hardware you can depend on! And you wonder why you have networking problems.

Becker has spent a LOT of his time adding Linux support for poorly documented, mis-configured, total crap hardware - and you cheap assholes complain. You should read the kernel newsgroup archives - some developers suggested a few years back that they dump RTL81XX entirely because the firmware sucked so bad. You whiners should be thanking Scyld for their work.

Or maybe buy a decent NIC instead.

Re:It's always the cheapskates who complain.

[Anonymous Coward]

the funny thing is, that winnt drivers dont exhibit the same god damn problems that becker's rtl8139 does. christ, the amount of collisions you get at 100/full duplex is ridiculous. the 8139too is a much better driver overall.

i will say this though - the sis900 driver (and firware), written by the boys at sis, sucks far more ass than any of becker's drivers.

at least becker's can autoneg reasonably well.

the problem is that my company likes the $0.25US/per onboard price on an rtl8139c...

Re:It's always the cheapskates who complain.

[molarmass192]

Having spent some time working on a Linux hardware driver for an undocumented PCI card, I think most people have no clue at the amount of work involved. Deciphering bus calls when there's little to no documentation to hint as to what the calls are for is like building a puzzle without the final image to look at. Getting a piece is easy, but figuring out where it fits relative to other pieces is one mean task. Anybody who has the audacity to just sit back and complain without either filing bugs or submitting patches should just migrate back to Windows, their whining certainly will not be missed. Also, good point about cheap-shit hardware, you're not going to end up with a Ferrari if all you have is parts for a tricycle.