Disk BottleNeck

[deepak_9]

Hi Experts,

Getting an error saying :-
<<
"Disk Bottleneck Probability=100% , End of Disk Bottleneck Alert".
>>
in my GlancePlus. I wanted to know what does this alert mean and how critical it is, what are the measures to take care of?

[Craig Falls]

Hi,

Basically GlancePlus has an "expert system" embeded in it which when the various aspects of performance drop check the things affecting performance and provide a somewhat informed decision about what is causing this performance drop. In the instance here GlancePlus is 100% sure that it's a disk bottleneck.

If you have a look at the /var/opt/perf/advisor.syntax file you can have a look at the way GlancePlus determines bottlenecks.

Now for criticality - well if you platform is like this ALL the time this is a critical problem. If it is only like this during specific jobs - can these jobs be rescheduled?

Measures to take care of it could range from newer faster disks/different storage technology (from internal SCSI disks to external fibre connected Array disks)

Before you can measure how critical it is and what needs to happen you really need to isolate what is running at the time the disk bottlenecks occur.

c

[Craig Falls]

Sorry my mistake, in the above I wrote that the information that GlancePlus displays comes from /var/opt/perf/advisor.syntax, well this is correct but the alarms are actually defined in /var/opt/perf/alarmdef .

c

[Tom Geudens]

Hi,
I don't know whether or not you already use it, but http://h21007.www2.hp.com/dspp/files/unprotected/devresource/Docs/TechPapers/UXPerfCookBook.pdf is a "must read" for bottleneck issues. I'm "fighting" an IO-bottleneck at this very moment too and this document is helping me along ... and it reads like a novel too :-) !

Hope this helps,
Tom Geudens

[Sanjay_6]

Hi Deepak,

Try this tuning guide,

http://docs.hp.com/hpux/onlinedocs/os/11.0/tuningwp.html

Hope this helps.

Regds

[T G Manikandan]

Hello,
This message is due to the I/O on the disk.
There is more read/write operations on the disk.
When these operations reach the maximum the performance monitor shoots this message.

check out the disk on which there is 100% disk I/O.
It is quite possible when you have lot of applications writing onto the same disk like oracle.
Distribute your application into various disks so that the I/O is also distributed for better performance.

Thanks