IO Performance

[cfeitosa]

Hello,

Guys, I'm using HPUX 11 and Storage Hitachi.

Please I need a help.
How can I to monitor the performance of IO of my all disk directly on HBA?

I don't know if the iostat command is sufficient for do it or maybe I don't know to use and to interpret it correctly.

I don't have any tool for do that... so I need to do a script or to know some commands that I can to monitor the performance of my server and how is the flow of IO between my disks and the Storage (HBA, ports, LUN, and so far and so on).

I'm a little new with this environment, so please could someone give me a help?

Thanks a lot!
clefeitosa

[Steven E. Protter]

Shalom,

I'm sure Hitachi has some tools, you should check with them.

http://www.hpux.ws/system.perf.sh

Free, functional and might do the job for you.

SEP

[Alastair Donaldson]

If you are connecting through a SAN switch the switch itself should also have some tools to monitor SAN port statistics for your zones.

[Santosh Rao]

Hi,

The latest version of HP-UX (11i v3) has the ability to monitor IO performance on a per HBA Controller basis as follows :

sar -H

Other new features in 11iv3 for IO performance monitoring include :

1. sar -t (tape lun performance).
2. sar -L (per lunpath performance).
3. sar -dR (disk lun performance with seperate read and write performance data split).
4. Similar enhancements have been provided in iostat and Glance as well.

Thanks,
Santosh

[cfeitosa]

Hi,

Thanks for all answers!

In this moment I don't have any tools neither HP-UX (11i v3)...
So... I need to go to another way.

I already used the system.perf.sh that Steven said.

I'm studying...

[AndyMueller]

Hello,

have you considered creating a custom rept file in /var/opt/perf? You can customize it to record Global Disk, CPU, Network filesystem metrics. I use it to give me a monthly report on Global Statistics for my servers, which are extracted, uuencoded and emailed to me via root's cron every month. Works like a champ. Is that something you are looking for? If so, I'd be glad to explain in more detail.

Andy

[Bill Hassell]

iostat is archaic and simply useless for modern disk connections, especially with disk arrays and SAN switches. It assumes that only simple disks are attached to a simple SCSI interface.

If your storage goes through a SAN switch, use the switch tools to read the stats. They are much more accurate and solve all the issues with alternate links. Most fibre switches (Brocade, EMC, Cisco, etc) have both a command line interface using telnet as well as a web based interface. The switch must be connected to a LAN or for command line only, connected to a serial port. You can also look at fcmsutil depending on what HBAs you are using (man fcmsutil).

[cfeitosa]

Hi,

Andy, please could explain in more details or send me an example?

Thanks for all answer!

[skt_skt]

fcmsutil /dev/td0

fcmsutil /dev/td0 stat

fcmsutil /dev/fcd1 stat -s

[AndyMueller]

Clever,

send me an email axmuelle@sentara.com and I will explain in more detail.

[Patrick Ware_1]

I would love to have a more detailed explanation on the following series of options for fcmsutil also:

fcmsutil /dev/td0

fcmsutil /dev/td0 stat

fcmsutil /dev/fcd1 stat -s