IObench is a generic I/O stress tool. It was born in willing just to understand what's going wrong sometime on system/ application/ or I/O side :-)
During execution, IObench reports its internal stats - what is supposed to be "seen" by application. Which very often helps to understand I/O subsystem limits and fix/understand scalability issues...
IObench kit is included into STAT-service of dim_STAT since v.8.2, as well web interface to IObench Scenario Generator.
- 1. Install dim_STAT
- 2. Install STAT-service on every server where you want to stress I/O subsystem
- 3. Check IObench is executing correctly:/etc/STATsrv/bin/IObench_v4 /etc/STATsrv/bin/IObench_v4 (dim) v.4.0 Usage: /etc/STATsrv/bin/IObench_v4 Options File [File2] Operation options: -h -- help -c command -- I/O command: read, write, rw -s size -- I/O size in KBytes -m mode -- I/O mode: seq, rnd -o order -- I/O rw-mode order chain: R=read, W=write, ex:'RWRRWR' (default: 'RW') -b position -- Begin offset (in
units) -e position -- End offset (in units) -t type -- I/O Type: buff, aio, osync -n Numb -- Finish after Numb I/O commands -l Limit -- Limit own throughput to Limit KB/sec -P -- Use pread/pwrite syscalls -L -- Loop I/O non-stop STAT options: -i ID -- Client ID for anySTAT (1:1024), -1 to disable anySTAT -k key -- SHMKEY for anySTAT -g group -- Group name for anySTAT Note: both -e and -b must be specified for random I/Os
|Generating I/O Load|
- 1. Generate a test scenario via dim_STAT web interface
- 2. Start generated script on your server(s)
- 3. Observe system and IObench stats via:
- collecting IObench stats directly via dim_STAT (pre-integrated)
- manually via:$ /etc/STATsrv/bin/IObench_STAT.sh 10