Lab Facilities

Currently, Our labs have a cluster with totally over 1000 cores and 8GB memory for each core.

  • 4 x Asus 1642 Xeon 5345 with total 8 core + 24 GB memory
  • 2 x IBM X3350 Xeon 5355 with total 8 core + 16 GB memory
  • 2 x Supermicro Xeon 5430 with total 8 core + 64 GB memory
  • 4 x Supermicro Xeon 5410 with total 8 core + 16 GB memory
  • 1 x IBM X3650 Xeon 5530 with total 8 core+16 thread + 48 GB memory
  • 1 x supermicro AMD 2431 with total 12 core x 48 GB memory
  • 1 x supermicro AMD 6174 with total 48 core x 192 GB memory
  • 4 x supermicro Xeon 5630 with total 8 core x 48 GB memory
Tutorial to use Lab facilities

Linux official website:  http://www.linux.org

Latex official website:  http://www.latex-project.org/

OpenOffice official website:  http://zh.openoffice.org/

Subversion official website: http://subversion.tigris.org/

DateAuthorContentSupplement
2007/08/07Tian-Li YuLinux short tutorial (zip)
2007/08/14Yuh-Renn WuLatex manuscripts and examples (zip)
2007/08/21Chen-Mou ChengSubversion tutorial (ppt, pdf)Online Free O’Reilly Book
2007/08/28Tian-Li YuC/C++ programming & make (zip)
2007/09/04Yuh-Renn WuMatlab tutorial (zip)
2007/09/11Jason ChangParallel & Distributed Computing (ppt)
PBS tutorial
Examples of PBS script for submitting jobs (job.sh)
#!/bin/bash
#PBS -S /bin/bash
#PBS -N grid-pi
#PBS -l nodes=1:ppn=1,mem=1gb,walltime=100:00:00,nice=15
#PBS -M xxx@xxx.xxx.xx
#PBS -q long
#PBS -m abe
#PBS -o /home/username/pbs_out
#PBS -e /home/username/pbs_out
cd /home/username/workingdirectory
./gridpi.exe > gridpi.info &
./a.out > a.out.txt &
wait


Note:
/home/username/pbs_out is a directory, not a file, you need to create this directory

“wait” command is very important, the program will wait until the the previous command is finished.

xxx@xxx.xxx.xx is your emal address

ppn=1 ppn: How many CPUs you need per node. If you are running a serious job, ppn=1. If you are running a parallel job, 1-8 are suggested values. You may not be able to find a workstation with ppn larger than 8. Only a few nodes has cpus larger than 8.

mem=1gb : How many memory you estimate your program will use. If you know your program needs memory larger than 1GB. You have better to request the memory you need. Or your program might crash due to insufficient memory.

-l nodes=1 : node number. The default is 1 except that you are running a MPI program.

Submitting jobs
qsub jobs.sh
Query job status
qstat
qstat -n
qstat -f
qstat -q
show all jobs running
showq
delete the job
qdel (jobid)
BASH Script Example
Examples of BASH SCRIPT (submit.sh)
#!/bin/bash
file[1]=”t1.sh”
file[2]=”t2.sh”
file[3]=”t3.sh”
file[4]=”t4.sh”
file[5]=”t5.sh”
file[6]=”t6.sh”

for (( i=1 ; $i<=6 ; i=$(( $i+1 )) )) ; do

mkdir ${file[$i]}
cd ${file[$i]}
for (( j=1 ; $j<=6 ; j=$(( $j+1 )) )) ; do
mkdir $j
cp ../a.out $j
cp ../gridpi.exe $j
eval “sed ‘s\\/home/yrwu/2\/home/yrwu/2/${file[$i]}/$j\\g’ ../excu.sh | sed ‘s\\grid-pi\grid-pi${file[$i]}$j\\g’ > $j/excu.sh ”
eval “qsub $j/excu.sh”
done
cd ..
done
Post Views: 5,809