sar -> wio

[j773303]

I see wio always 45% in my sever, while using
sar -r to view. Does anyone tell me, what's mean of wio, and why it is reach 45%?

[John Poff]

Hi,

The 'wio' value in sar is the percentage of time that processes have spent waiting for I/O. Take a look at 'sar -d' and see what the avque and % busy values are to see the average queue length and percentage busy. You could have some disk bottleneck, or you could have an issue with mount options for your filesystems. As with all tuning problems, the answer is, it depends.

Better still, if you have Glance installed you can see much more than what sar will show you.

What model of system do you have? What version of HP-UX? What kind of disk storage? What kind of applications does it run?

JP

[Norman_21]

Hi,
My $.02:
I would suggest that you do a search, there are alot of threads out there. I found some good ones for you.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x585eef70e827d711abdc0090277a778c,00.html

Another one:
http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x43b44a988422d711abdc0090277a778c,00.html

Hope this help.
Thanks and good luck.

[Michael Steele_2]

Anything above 15% is not good and indicates a disk or tape bottleneck.

What is the value of %idle for sar -u 5 5?

For %busy in sar -d 5 5 anything above 50%?

For anything in sar -d 5 5 where %busy is above 50% is avwait > avserv ?

For avserv in sar -d 5 5 anything above 20 ms?

What is the value of po for vmstat 5 5?

...sar -v 5 5?

Percent total utilized in swapinfo -tam?

[Steven E. Protter]

In my humble opinion, you are overutilizing disk resources. To many applications with not enough memory, you are probably paging in and out of swap like crazy.

The question at hand is whether this is a short term problem or a long term issue. This can be done with further research. To that end, I'm attaching my production data collection script. It collects the same data background over any time period you set in the script.

SEP