jfreeman wrote: > > Danny, > > I agree with Eric's fast cluster building with boot disks, (I did it > myself the first few times) and like the practical examples, but you > will _really_ want to do remote booting of the nodes as your normal > process, it will make node level install of software, OS, node > configuration, versioning, etc. much much easier. > > A very good example page for this is the: > > Linux Terminal Server Project > > http://www.ltsp.org/ > > It will get you well on the way... > > Good luck, > > Jim Remote booting is slick, but is the advantage really that clear for this kind of cluster (embarrassingly parallel, diskfull, and none-too-big)? The NFS share on the master node holds the install image(s) and BLAST executables. Any node specific features can be handled with environmental variable calls. I suppose if you are updating the kernel often, then remote booting is a far better choice, but who does that with production machines (not me, call me chicken). You still have to go through the DHCP/MAC routine, regardless. That being said, I am going to try ltsp too. Thanks for the pointer, Jim! -- Eric Engelhard - www.cvbig.org - www.sagresdiscovery.com