Large Scratch Disk Available to Programs during execution on the HKUSP2 System

We have recently installed two 9.1GB disks on two batch processing nodes (spf3n10e and spf3n11e ) of the HKUSP2 system. While the adding of these two disks are intended for use by the Gaussian 98 software package which requires a huge scratch disk space, users requiring a large working disk space during their program execution (serial or parallel) can also use this large disk facility by using the following loadleveler command setup.

The procedure below is for reference only.  G94/G98 users are advised to check the web-page for the Gaussian software on HKUSP2 for the updated procedure which may be required in case there have been system modifications since the writing of this document. 

#!/bin/csh
#@ job_name = test123
#@ initialdir = /d12/home/cc/cmwoo/test
#@ output = $(job_name).$(cluster).out
#@ error = $(job_name).$(cluster).err
#@ job_type = parallel
#@ class = g94
# Use 1 or 2 processors depending on whether 
# your program is serial or parallel.
#@ min_processors = 1
#@ max_processors = 2
#@ checkpoint = no
#@ queue

#Clear the temporary disk space on those nodes
foreach node ( $LOADL_PROCESSOR_LIST )
  rsh $node /usr/local/bin/cleartmp1
end
#Run the paralle program and use /tmp1 as local scratch disk space.
poe my_parallel_program.exe
# If it is serial program, use this command instead : 
# ./my_serial_program.exe

Should you have any queries relating to the use of this large scratch disk space, please feel free to contact the undersigned.


 

Kwan Wing Keung
Tel: 2857 8631
E-mail: kwk@cc.hku.hk

Woo Chat Ming
Tel : 2857 8632
E-mail: cmwoo@cc.hku.hk


[Contents] [Next] [Previous]