Building a cluster to run Molpro

[Kirk Peterson]

Martyn,

I have a AMD-based cluster (dual node MP2000+ processors) that I've been very happy
with running molpro on. In our configuration we have 10 slave nodes and 1 master node.
All nodes, however, have local disk (2x35 Gb Ultra160 SCSI). This is absolutely
necessary. We have 3 GB per processor and I wouldn't recommend anything less than
2 Gb (1Gb/node). For interconnect we have myrinet, but I've recently seen from
the Stuttgart group that perhaps gigabit ethernet might be sufficient (and alot
cheaper!). We use the PGF compiler and since we use MPICH for parallel molpro, the
compiler has little to do with the parallelization (i.e., buy the workstation version
of PGF).

best wishes,

Kirk

PS - if you have a little extra money, it might be worthwhile to buy the newer AMD
Opteron processors, which are 64 bit.

------
Kirk Peterson
Washington State University
Department of Chemistry
Pullman, WA 99164
(509) 335-7867
kipeters at wsu.edu
http://tyr0.chem.wsu.edu/~kipeters

On Wed, Jul 16, 2003 at 10:04:50AM +0100, Wheeler, Dr M.D. wrote:
> Dear All,
> We are putting together a new multiprocessor (8 or 10) machine to run molpro and would like some advice on what hardware would be recommended by experienced users.
> Which is better, Intel Xeon or AMD MP??  Is a sensible configuration to have one mainboard acting as a master node (with striped pair of HDD for efficient disk I/O) and the other mainboards acting as slaves??  Or is it better to have a more general set up that has disks connected to all of the nodes??  How much memory per processor is recommended >1GB??
> 
> Also, is the PGF compiler best, the standard version appears to parallelize for up to 4 processors, I guess the CDK is required for bigger applications.
> 
> Any suggestions are welcomed, thanks in advance,
> Martyn
> 
> ----------------------------------------------
> Dr. Martyn D. Wheeler
> Department of Chemistry
> University of Leicester
> University Road
> Leicester, LE1 7RH, UK.
> Tel (office): +44 (0)116 252 3985
> Tel (lab):    +44 (0)116 252 2115
> Fax:          +44 (0)116 252 3789
> Email:        martyn.wheeler at le.ac.uk
> http://www.le.ac.uk/chemistry/staff/mdw10.html