Disk I/O

[Gopinath rao_1]

i would like to Gatther disk i/o information .

Reads per sec and writes per sec .
sar -d gives reads+writes per sec.

thank you in advance
Gopi

[A. Clay Stephenson]

By far, your best statistics are going to be gathered by Glance. If you don't have Glance, you can install a trial version using your Application CD's.

[Roger Baptiste]

hi!

Use glance -> option u and then select any disk you want by option S . That gives the read and writes on a disk.

HTh
raj

[Jeff Schussele]

Hi Gopi,

Then you should use Glance - specifically the graphical version gpm.

From it you can get an I/O by disk report that will give reads & writes separately.

Glance is a purchase item but you can download it & get a free trial period. But it's definitely worth the money to keep it.

Rgds,
Jeff

[Gopinath rao_1]

i need to script . as i have to collect the data for 2 weeks for our Disaster Recovery planning to build the wan pipe.

i would like to know if we have any command or script which could help me.

Thank you
Gopi

[Michael Tully]

You can also use the measureware interface of the glance product to also collect data. From this you can extract information into valuable reports. There are literally hundreds of reporting metrics that can be used. CPU, Disk IO, memory usage. Glance is a terrific product and well worth the expenditure.

[Michael Tully]

You can use the 'extract' part of the measureware program to collect information. The program 'extract' has a user interface to guide you. Once you have installed both components from your application CD, you can try it, but wait a couple of hours so there is at least some data to experiment with.

[harry d brown jr]

The accuracy of glance/measureware versus sar is like comparing a mercedes to a yugo (Glance/measureware being the mercedes and sar being the yugo). If you use sar, you might as well just guess, because your guess is as good as sar's.

If you are doing disaster recovery planning, then the cost of glance/measureware and perfview should never be an issue. You really should HIRE either HP or another firm to assist you in the data collection and analysis. If you don't, then you are wasting your time and our time.

live free or die
harry

[Sanjeev Kohli_1]

Hi ! Gopi

You can run the following sar command to collect data for the next 2 days.

# sar -o sarfile 5 (secs) 10 (no. of times) --- you can increase the secs and no. of times parameter option forthe command to be run for the next 2 days.

After the the command has stopped running or you have interrupted it , run

# sar -AF sarfile > sar.file
# pg sar.file

This file would have collected the data as per your requirements .


Have fun,

Sanjeev

[Mott Given]

Harry, I am very interested in your comment on sar comparing it to a Yugo, as our company expends quite a bit of effort collecting sar stats that are graphed and posted to an internal web site. What specific things have you noticed about sar accuracy compared to Glance?

Mott Given

[Bill McNAMARA_1]

try
iostat -t

It gives pretty nice info.
not as comprehensive as Glance as mentioned, but maybe what you're looking for.

Bill

[harry d brown jr]

By Doug Grumann,
http://h21007.www2.hp.com/dspp/files/unprotected/UsingGlance.pdf

an extract:

I am often asked why the data shown in Glance sometimes differs from the data shown by tools such as sar, vmstat, iostat, and top. Most often, the root cause of discrepancies is due to the underlying instrumentation. Glance uses Special tracing features in the HP-UX kernel that are translated into performance metrics via the midaemon process. Tools like top get their data from counters in the kernel that are maintained by sampling instrumentation. Because a tracing methodology can capture all system state information, it is more accurate than data that is obtained via periodic sampling.

live free or die
harry

[Michael Kahlke]

Is there any way to get separate read/sec and write/sec data other than through glance/midaemon? I am currently using sadc and parsing the binary output it produces. I've had to reverse-engineer the binary output as there seems to be no documentation on the data format. I've found fields that correspond to the queue lengths and total IO/sec but not separate read IO/sec and write IO/sec.