IOR run

[hariharan-devarajan]

Key arguments

process-mapping: -F- file-per-process and -c- collective
interface: -a=<INTERFACE> values for INTERFACE:POSIX|MPIIO|HDF5
write-only: -w
block_size: -b=BLOCK_SIZE BLOCK_SIZE=256k 32m 1g here k is kilobyte, m is for megabyte, and g for gigabyte.
Select this based on how big a file u want. I would recommend at least 32mb block size
transfersize: -t=TRANSFER_SIZE TRANSFER_SIZE: Small requests (4k-32k) medium (64k-512k) good (1m-16m) and large (16m to 1GB).
file path: -o=. This has two parts a system's mount point (e.g., /p/lustre3) then path to file /p/lustre3/haridev/test_ior_64.

Metrics

If your looking at file output
Operation Max(MiB) Min(MiB) Mean(MiB) StdDev Max(OPs) Min(OPs) Mean(OPs) StdDev Mean(s) Stonewall(s) Stonewall(MiB) Test# #Tasks tPN reps fPP reord reordoff reordrand seed segcnt blksiz xsize aggs(MiB) API RefNum
190: write 565.67 565.67 565.67 0.00 565.67 565.67 565.67 0.00 2.71535 NA NA 0 48 48 1 0 0 1 0 0 1 33554432 1048576 1536.0 MPIIO 0

Just use the Mean(MiB) and StdDev as your measure.

if you use CSV: u need to add it to your IOR Run -O summaryFormat=CSV -O summaryFile=PATH TO FILE.csv
access,bw(MiB/s),IOPS,Latency,block(KiB),xfer(KiB),open(s),wr/rd(s),close(s),total(s),numTasks,iter
write,491.9527,125940.5447,0.0001,16777216.0000,4.0000,0.0050,266.4307,31.0752,266.4321,8,0
read,20352.3526,5210357.5328,0.0000,16777216.0000,4.0000,0.0003,6.4399,0.2269,6.4401,8,0
write,495.5930,126872.5599,0.0001,16777216.0000,4.0000,0.0067,264.4735,31.0806,264.4751,8,1
Use bw(MiB/s)

Sample IOR Run:

/usr/bin/flux run -N 1 --tasks-per-node=48 /usr/workspace/haridev/iflux/build/bin/ior -F -a=POSIX -w -b=32m -t=4k -o=/p/lustre3/haridev/test_ior_64