README 1.31 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Test Cases to evaluate darshan-parser --perf performance calculations.

Logs
----
350665
  * IOR single shared file with multiple iterations
  * 350665.output
  * harms_IOR_id350665_2-8-78651_12.darshan.gz

350690
  * IOR file-per-process with multiple iterations
  * 350690
  * harms_IOR_id350690_2-8-83458_1.darshan.gz

350892
  * IOR with both single file and file-per-process modes
  * 350892.output
  * harms_IOR_id1350892_2-9-74258_1.darshan.gz

Check Performance
------------------
Add all the performance test runs (both read and write) and the compute the average. 

awk 'BEGIN {sum=0;count=0;} /^(write|read)/ { sum+=$2; count+=1; } END { print "avg: " sum/count; }' xxxx.output

darshan-parser --perf log.gz

Compare the four methods to the calculated average.

IOR Commands
------------
# File Per Proc with Multiple Files
qsub -q prod -A Operations -t 60 -n 512 --mode vn --env BG_COREDUMPDISABLED=1 IOR/src/C/IOR -a POSIX -F -b 256M -g -m -i 4 -K -o <path>/ior.out -r -w -t 4m -v

# Single Shared File with Multiple Files
qsub -q prod -A Operations -t 60 -n 512 --mode vn --env BG_COREDUMPDISABLED=1 IOR/src/C/IOR -a POSIX -b 256M -g -m -i 4 -K -o <path>/ior.out -r -w -t 4m -v

# Both I/O methods in one run
qsub -q prod -A Operations -t 60 -n 512 --mode vn --env BG_COREDUMPDISABLED=1 IOR/src/C/IOR -f single_and_multi_file.ior