diskless clients? beowulf-newbie seeks advice
cblack at eragen.com
Fri Jun 22 12:15:04 PDT 2001
On Fri, Jun 22, 2001 at 10:55:08AM -0700, Brian LaMere wrote:
> why does every guide around talk about diskless clients? I mean...disks are
> stinkin cheap nowadays...
[rant about boss-man making silly hardware choices for cluster
[other stuff about pvm, networks, etc deleted]
> Just a little genetics research firm, needing some serious horsepower to
> start running big hammer and blast jobs. The data we have now is just the
> bare minimum we need to get by, but if we had things like a working beowulf
> the scientists upstairs would start making, since they'd be able to use it,
> much more data. They hired me on as the unix guy here knowing I don't know
> squat about beowulfs, but that I really want to learn :) Got "how to build
> a beowulf" <grin> and I've read the manuals for pvm, mpich, lam-mpi, etc,
> and several other beowulf how-to guides. All are about diskless. Is
> diskless better? Is it just better because its cheaper? Are there other
> reasons its better? Would having gig-ethernet in the boxes instead of hard
> drives be far better performance-wise?
It seems like you are doing bioinformatics type stuff (BLAST, etc).
We have found that many of these types of applications are close
to I/O bound, and especially when you consider having 50+ nodes
stream lots of data over the net at the same time, using a local
disk sounds better and better.
We use local disks here (we do bioinf stuff as well) and have a copy
of some core data on each node. When we need more data, we push it
over via rsync before computation begins. The performance of this
is very nice and only requires a 10GB fast EIDE disk per node.
Management isn't bad since we have scripts to push data out and
use systemimager (www.systemimager.org) to install/manage node
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 232 bytes
Desc: not available
More information about the Beowulf