Ask Donald Becker

This is a "needs no introduction" introduction, because Donald Becker is one of the people who has been most influential in making GNU/Linux a usable operating system, and is also one of the "fathers" of Beowulf and commodity supercomputing clusters in general. Usual Slashdot interview rules apply, plus a special one for this interview only: "What if we made a Beowulf cluster of these?" is not an appropriate question.

Re:One question...

[kiolbasa]

If I recall, the definition of a Beowulf cluster does not specify Linux specifically, only a free operating system.

Look it up [canonical.org]

Re:One question...

[The Turd Report]

But, there have been beowulf clusters made out of Windows boxes.

Re:Considering you're on the board at scyld..

[gowen]

crynwr...And why don't you people like vowels

Crynwr is a Welsh word, and in Welsh both 'y' and 'w' are vowels. So thats a pretty good ratio...

Re:One question...

[jahjeremy]

Note: Only logged because AC is giving me formkey errors.

This isn't a very well-informed question. Beowulf does not specify a particular platform.

From the Beowulf FAQ [canonical.org]:
[Beowulf is] a kind of high-performance massively parallel computer built primarily out of commodity hardware components, running a free-software operating system like Linux or FreeBSD, interconnected by a private high-speed network.

Please mod accordingly. Let's not waste Becker's time or one of the ten questions on ill-informed pablum refuted in the first question of an FAQ.

Re:Message Passing vs. Single System Image

[fgodfrey]

These two ideas aren't mutually exclusive. The Cray T3E is a single system image machine, but applications running on it are almost exclusively message passing in nature. My opinion on why there aren't proliferations of SSI clusters is because they are a lot harder to build. If you go with a set of seperate machines, which means you don't have a single *memory* image, getting the various kernels involved to all talk to each other is non-trivial. If you go with a single memory image, then you're not really doing a cluster, you are building a real supercomputer. Examples of single memory image machines of large size include the Sun Enterprise 1x000 line, the SGI Origin 2000/3000 series, the Cray T3E, and the not-quite-in-full-production-yet Cray X1.

As for the 32 bit address limit, it's already a problem. For large scientific code, 4GB per processor is already not enough. Now, people live with it, but that doesn't mean they like it. Intel's 36-bit addressing hack doesn't help, either, since you still have a single-virtual-address space limitation of 32 bits. This is probably the biggest motivation to go to a 64 bit architecture. Note that this problem also applies to large databases.

Re:One question...

[bhsx]

No, there have been high-performance, super-computing clusters built on Windows OSs (w2k, iirc). You can be quite sure that M$ doesn't call them Beowulf.

Re:Which Network gear manufacturer?

[0x0d0a]

I like Becker's drivers, but I ran into a problem with his Tulip ones -- on a *massively* overloaded Ethernet, if you get 16 retransmits failing and so the transmit fails, the driver does a full reset of the card. This makes the card not send data for about two seconds, which means on an extremely overloaded Ethernet, the card isn't that useful.

Right now, I'm using a 3c905b card (though it isn't a Becker project) with great success.

I think Linus likes eepro cards, IIRC from lkml.

Why no Linux drivers for IPSec Ethernet cards?

[Quietti]

I wanted to start using crypto-enabled Ethernet, only to find that Donald Becker has not made drivers for these and that he asks people to directly contact 3Com or Intel for their non-GPL drivers [scyld.com] instead. What's preventing Don from writing his own GPL drivers for those cards? Is there some US crypto export restriction law that directly forbids it? The same condition appears to affect several Gigabit cards too: please contact the manufacturer for their non-GPL driver. What's the deal?