improve I/O

[Jorge Prado_1]

Hello,
Which the parameters of the kernel that should be adjusted are and to that value to improve the performance of I/O toward an array of disks SCSI?.
O.P.: HP-UX 11.11
Server: rp2405
SCSI: Ultra160
Thank you

Jorge A. Prado T.

[LucianoCarvalho]

hi Jorge,

Take a look at this doc.
http://www.dutchworks.nl/htbin/hpsysadmin?h=3&dn=71082&q=performance%20tuning&fh

regards,

[LucianoCarvalho]

search for doc ID KBRC00000947.

regards

[Michael Steele_2]

In general, increasing I/O performance is a factor of block size or spindles. While kernel parameters are important, they are not the only consideration.

Since you have a disk array more spindles can be replace with more I/O pipes into the disk array.

With file systems mostly dedicated to one process larger block sizes can increase I/O performance. Conversely, if a file system has many processes associated to it a smaller block size can increase performance. Crossing large block size with a file system with many processes can cause a I/O bottleneck, so its important not to cross the two.

Disk arrays have other factors to consider some of these include ACP pairs, raid type and number of luns.

[Tim Sanko]

To improve performance, I would like to know what kind of Array, whether you are using Raid 1,5,S etc.


Many questions I need answered before I ask you to
send me SAR/measureware output...

TJS

[Steven E. Protter]

Attaching a sar performance data collection script.

SEP

[Jorge Prado_1]

Hello,
Explanation:
we have an RAID Fortra configured in RAID 5 and through a Ultra2 Wide scsi card.
The Ultra160 SCSI card has a tapes library qualstar 6430.

Steven, I attach files created by their shell.

Thank you.
Jorge Prado T.

[Michael Steele_2]

/respaldo has some issues at c0t1d1. Note the following. Here you have disk bottlenecks at c1t15d0 and c3t15d0:

16:43:37 c1t15d0 83.00 0.50 159 1506 5.30 5.79
c3t15d0 87.00 0.50 152 1368 5.37 5.51

You have several issues with c0t1d1:

16:47:50 c0t1d1 100.00 2.72 198 11584 17.08 47.33
16:47:51 c0t1d1 100.00 1.43 191 11200 11.17 45.21
16:47:52 c0t1d1 100.00 1.55 199 11712 10.04 44.20
16:47:53 c0t1d1 100.00 3.32 197 11458 22.02 50.24

Including an I/O bottleneck:

16:45:20 c0t1d1 100.00 18.25 298 17456 67.01 46.83

Your disk bottleneck also shows up with 100 %wio :

16:38:04 0 0 100 0
16:38:05 0 0 100 0
16:38:06 0 0 100 0
16:38:07 0 0 100 0
16:38:08 0 0 100 0

Could you check logtool?

STM > TOOLS > UTILITY > RUN > LOGTOOL > FILE > VIEW > RAW SUMMARY.

Note the first and last dates of transactions and calculate the difference. If the difference is short, like 4 hours, then this is important to note. Now read down the report of hardware addresses and observe the integer numbers in parenthesis. Anything over 150 in this 4 hour period should be called into HP for replacement.

[Michael Steele_2]

Let me clarify that /respaldo resides upon c0t1d1, the disk array while your O/S resides upon c1t15d0 and c3t15d0.