Results 1 to 6 of 6
Hi all,
I am trying capture the disk IOPS from OS ... I tried to capture with iostat.. but basically i have a doubt whether the r/s (read requests ) ...
- 08-19-2009 #1Just Joined!
- Join Date
- Aug 2009
- Posts
- 3
DISK IO stats in LINUX
Hi all,
I am trying capture the disk IOPS from OS ... I tried to capture with iostat.. but basically i have a doubt whether the r/s (read requests ) and w/s write requests are equal to actual disk IOs at storage level ..
I was just wondering how to correlate using IOSTAT , SAR and VMSTAT commands from other tools like AWR in oracle databases ??
Thanks in advance for the thoughts.
Phani
- 08-19-2009 #2Linux Guru
- Join Date
- Apr 2009
- Location
- I can be found either 40 miles west of Chicago, or in a galaxy far, far away.
- Posts
- 8,974
You can specify a device to iostat to see what the stats are for a specific device, which can be a specific drive partition such as /dev/sda1. I think that's about as close to "storage level" that you will get.
Sometimes, real fast is almost as good as real time.
Just remember, Semper Gumbi - always be flexible!
- 08-19-2009 #3Just Joined!
- Join Date
- Aug 2009
- Posts
- 3
Ya for the specific device am able to get with options .
like sar -d and iostat -b but then the stats within sar, vmstat and iostat are not matching and moreover in the output it has tps (total requests per second), but what I read from searching is that requests are at the device driver level ...now does that mean there can be any difference with hard disk IOS at storage level .. here what does it mean from device driver level ???
Thanks
Phani
- 08-19-2009 #4Linux Guru
- Join Date
- Apr 2009
- Location
- I can be found either 40 miles west of Chicago, or in a galaxy far, far away.
- Posts
- 8,974
A device driver can handle i/o requests for more than one device. Also, these stats are very much point-in-time data sets, so running one and then another will likely give differing values. They are only tools for scoping out the general health and load on the system and are intended for longer term analysis, such as I/O stats over some period of time.
Sometimes, real fast is almost as good as real time.
Just remember, Semper Gumbi - always be flexible!
- 08-19-2009 #5Just Joined!
- Join Date
- Aug 2009
- Posts
- 3
You mean same driver can handle requests for deices like sda sdb etc then .. anyways we are getting stats for sda and sdb etc... individually . and moreover riostat and sar will give results for over a period (cumulative)..
actually when i compare these over period results with some third party performance tools the results far beyond differing.. so am not able to figure out
Thanks
Phani
- 08-19-2009 #6Linux Guru
- Join Date
- Apr 2009
- Location
- I can be found either 40 miles west of Chicago, or in a galaxy far, far away.
- Posts
- 8,974
Well, you know that old saw about statistics? There are lies, there are damned lies, and then there are statistics...
Sometimes, real fast is almost as good as real time.
Just remember, Semper Gumbi - always be flexible!


Reply With Quote