A few years ago (okay, maybe more than a few), I recall reading a news story about Virginia Tech acquiring a lot (hundreds, maybe?) of a new model from Apple.
IIRC, the point was to experiment with comparing many small machines set up in parallel with much larger computers.
I do not recall ever having read anything about how those experiments turned out and whether they proved anything. Anybody know?
The idea of building supercomputers using massive arrays of CPUs in parallel has been around since the 1990s.
Such machines have been built with thousands of cores.
https://en.wikipedia.org/wiki/History_of_supercomputing#Massive_processing:_the_1990s
And yes, it's even been done with off the shelf PC CPUs.
I do not recall ever having read anything about how those experiments turned out and whether they proved anything. Anybody know?
The Virginia tech thing was done when Apple release that PowerMac Pro with Dual G5 processors.
They created a massively parallel supercomputer, and for a few months it was considered the fastest supercomputer on earth. It was dismantled two years later.
https://en.wikipedia.org/wiki/Beowulf_cluster
tells you more about the concept.
tells you more about the concept.
I've always enjoyed clustering with a copy of Beowulf clutched against my chest on a dark and stormy night. ;)
I've always enjoyed clustering with a copy of Beowulf clutched against my chest on a dark and stormy night. ;)
Or perhaps a couple of fire teams armed with AR15s chambered for .50 Beowulf.
https://en.wikipedia.org/wiki/.50_Beowulf
A Beowulf cluster is all on one local area network. That's a very different thing than SETI@Home.
It's not just processing bits and pieces of a data stream in parallel across different computers(SETI@Home). It's about getting a group of commodity micro-computers to act as one large computer.
A Beowulf cluster is all on one local area network. That's a very different thing than SETI@Home.
It is still the same concept though. Using large number of lesser powered computers to replace the projects that supercomputers used to once do.
It is still the same concept though. Using large number of lesser powered computers to replace the projects that supercomputers used to once do.
A Beowulf cluster isn't an attempt to use lesser powered computers to replace the projects that supercomputers used to do, it's an attempt at building a poor man's supercomputer.
That's the point, a Beowulf cluster is designed to act like a single large computer. A user accessing it remotely shouldn't be able to tell that it's a Beowulf cluster rather than a more traditional super-computer.