diskless nodes? (was Re: Xbox clusters?)
troy at osc.edu
Fri Dec 7 06:47:10 PST 2001
On Fri, 7 Dec 2001, Velocet wrote:
> On Thu, Dec 06, 2001 at 02:02:04PM -0500, Greg Lindahl's all...
> > On Thu, Dec 06, 2001 at 10:11:44AM -0500, Velocet wrote:
> > > I am sure a number of others' use
> > > of clusters also fit into this range, thus my wondering.
> > Given the poor tone of both this discussion and the one about paying
> > more for nodes, most people are being quiet.
> Discussing diametrically opposed technological solutions does
> not necessarily make it a 'poor tone'. I apologize if my posts
> came off that way.
> It is a black and white choice "disk or not to disk" and people decide one way
> or the other, and I am wondering what makes them decide. I have information
> which indicates that for the parameters on the systems I setup it works quite
> well. I need someone to point out to me where my reasons are just dead wrong,
> or where I dont have enough experience - perhaps they can anticipate where I
> will go wrong before I find out the hard way. However, so far, others have
> raised points that I've indicated dont necessarily indicate that a diskless
> system is a bad choice all round (admittedly I havent gotten through
> all of today's posts yet).
> Hopefully, even though the issue is more or less black and white "do
> I get diskless nodes or not", we can all still learn something about it by
> discussing it.
Well, there's nothing keeping you from keeping the root filesystem on NFS
and using local disk for swap, /tmp, and /var. We do that on our Pentium
III and Itanium clusters, and it seems to work pretty well. The biggest
problem is user education ("No, 'cp file /tmp' does not copy it out to all
the nodes in your job, use pbsdcp instead").
Troy Baer email: troy at osc.edu
Science & Technology Support phone: 614-292-9701
Ohio Supercomputer Center web: http://oscinfo.osc.edu
More information about the Beowulf