Assessing disk performance with the iostat command
Begin the assessment by running the iostat command with an interval parameter during your system's peak workload period or while running a critical application for which you need to minimize I/O delays.
The following shell script runs the iostat command in the background
while a copy of a large file runs in the foreground so that there is some I/O to measure:
# iostat 5 3 >io.out &
# cp big1 /dev/nullThe AIX® operating system maintains a
history of disk activity. If the disk I/O history is disabled, the following message is displayed
when you run the iostat
command:
The following examples show that you must consider all available data on the system to
troubleshoot a problem. Disk history since boot not available.
The interval disk I/O statistics are unaffected by this.smit chgsys and then select
true from the Continuously maintain DISK I/O history
field.-
The following command stores three data samples on a system, which has three disks in the io.out file. These samples are collected at an interval of 5 seconds between samples.
# iostat -T 5 3 > /tmp/io.out & # cp bigfile /dev/null System configuration: lcpu=4 drives=4 ent=1.00 paths=3 vdisks=2 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 1.2 9.6 0.6 1.4 98.0 0.0 0.0 2.7 13:26:46 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:26:46 hdisk1 0.0 0.0 0.0 0 0 13:26:46 hdisk0 0.0 0.0 0.0 0 0 13:26:46 hdisk2 0.0 0.0 0.0 0 0 13:26:46 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.2 3.6 0.3 13.8 75.1 10.8 0.2 16.8 13:26:51 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:26:51 hdisk1 0.0 0.0 0.0 0 0 13:26:51 hdisk0 62.8 52428.8 205.4 262144 0 13:26:51 hdisk2 0.0 0.0 0.0 0 0 13:26:51 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.0 0.0 0.5 1.5 97.9 0.1 0.0 2.8 13:26:56 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:26:56 hdisk1 0.0 0.0 0.0 0 0 13:26:56 hdisk0 0.4 0.8 0.2 0 4 13:26:56 hdisk2 2.6 33.6 7.8 0 168 13:26:56Note: If the iostat command is run without specifying a time interval, the output indicates a summary of the system data since the last system reboot, and not the current values.- The first and third intervals show that the three disks were mostly idle, along with the CPU utilization, which is also shown as idle in the tty report.
- The second interval shows the activity that is generated by using the cp command, which was started for this test. This activity can be viewed on both the CPU activity (tty report) which shows 13.9% sys CPU and also on the disk report. The cp command took 3.14 seconds to run during this interval. In the report, the second interval shows 62.8 % for the hdisk0 disk under the tm_act metric. This example means that the hdisk0 disk was busy for 62.8 % of the time interval (5 seconds). If the cp command is the only process generating disk I/O to hdisk0, then the cp command took 62.8% of the 5 second interval, or 3.14 seconds, which is the total time the cp command took to run.
-
The following command stores five data samples with a 2 seconds interval between samples on a system, which has three disks in the io.out2 file:
# iostat -T 2 5 > /tmp/io2.out & # cp bigfile /dev/null System configuration: lcpu=4 drives=4 ent=1.00 paths=3 vdisks=2 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 3.0 24.0 0.4 0.8 98.8 0.0 0.0 1.8 13:29:51 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:29:51 hdisk1 0.0 0.0 0.0 0 0 13:29:51 hdisk0 0.0 0.0 0.0 0 0 13:29:51 hdisk2 0.0 0.0 0.0 0 0 13:29:51 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.5 1.0 0.2 8.2 85.5 6.1 0.1 10.1 13:29:53 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:29:53 hdisk1 0.0 0.0 0.0 0 0 13:29:53 hdisk0 39.5 30464.0 120.5 60928 0 13:29:53 hdisk2 0.0 0.0 0.0 0 0 13:29:53 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.0 0.0 0.2 21.5 62.9 15.4 0.3 25.7 13:29:55 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:29:55 hdisk1 0.0 0.0 0.0 0 0 13:29:55 hdisk0 100.0 83712.0 327.0 167424 0 13:29:55 hdisk2 0.0 0.0 0.0 0 0 13:29:55 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.0 8.0 1.3 7.2 87.5 4.0 0.1 10.4 13:29:57 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:29:57 hdisk1 0.0 0.0 0.0 0 0 13:29:57 hdisk0 20.9 16898.0 66.5 33792 4 13:29:57 hdisk2 10.4 82.0 19.0 0 164 13:29:57 tty: tin tout avg-cpu: % user % sys % idle % iowait physc % entc time 0.0 0.0 0.2 0.6 99.2 0.0 0.0 1.3 13:29:59 Disks: % tm_act Kbps tps Kb_read Kb_wrtn time cd0 0.0 0.0 0.0 0 0 13:29:59 hdisk1 0.0 0.0 0.0 0 0 13:29:59 hdisk0 0.0 0.0 0.0 0 0 13:29:59 hdisk2 0.0 0.0 0.0 0 0 13:29:59
- The first and fifth interval show that the three disks were mostly idle, along with the CPU utilization, which is also shown as idle in the tty report.
- The second interval shows the activity that is generated by using the cp command, which was started for this test. The cp command took 3.14 seconds to run during this interval. In the report, the second interval shows 39.5 % for the hdisk0 disk under the tm_act metric. In third and fourth interval shows 100 % and 20.9 % respectively for the hdisk0 disk under the tm_act metric. This means that the hdisk0 disk was busy for 100 % of the time interval (2 seconds) during the third interval and the hdisk0 disk was busy for 20.9 % of the time interval (2 seconds) during the fourth interval.