how to calculate disk throughput

[NDO]

Hi All I dont know if this is the right group to post this message, but here I go: I have a rx7640 connected to a NetApp, and I need to know if I can run a specific application as long as this NetApp storage system gives me 850IOPS. So how can I calculate how many IOPS the NetApp will give me. F.R.

[Turgay Cavdar]

You can use iorate :

ftp://ftp.emc.com/pub/elab/iorate/

[NDO]

Hi According to the pdf file, this tool is destructive, so I cannot run this tool on a prod system. Do you have a better suggestion?

F.R.

[Turgay Cavdar]

Read tests are not destructive, but write test of course destructive. You cannot do any write test that  is not destructive on raw devices, this is the nature of write operation.

I think you are using netapp over NFS file systems, so if you use different file names then write tests also not destructive.

Regards.

[Steven E. Protter]

Shalom,

I calculate disk throughput with sar.

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

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

SEP

[NDO]

Hi

What I really want is how many IOPS I will get from netapp disks?

[Pete Randall]

I can only suggest that you run some test against these disks and measure your throughput using sar or glance.

[Dennis Handly]

>as long as this NetApp storage system gives me 850IOPS.

Are there any tools on the NetApp side to measure the NFS IOPS?

[NDO]

Hi

wouldn´t know if from the storage side there is any tools to calculate IOPS, but, someone mentioned that I could use sar tool. Any hints in how to go about using that tool?

F.R.

[Pete Randall]

The someone who suggested sar, Steve Protter, also include links to scripts that he has set up, presumably to do exactly what you're asking.  Try clicking on the links and take a look at the scripts, then click on the kudo button to indicate your appreciation.

[NDO]

Hi

Sorry If I did upset you, I will, like always do, click on kudos.

FR

[Pete Randall]

Not upset, just trying to make a suggestion.  Glad to see you always do!

[NDO]

Hi

so what would be the right measure:
1. the sar -H 2 2, that gives me :
HP-UX dbnode1 B.11.31 U ia64 07/12/11

13:37:41 ctlr util t-put IO/s r/s w/s read write avque avwait avserv
%age MB/s num num num MB/s MB/s num msec msec
13:37:43 fcd0 90 4.73 140 130 9 4.70 0.03 1 0 7
13:37:45 mpt0 1 0.01 2 0 2 0.00 0.01 1 0 6
fcd0 89 4.84 184 139 45 4.62 0.22 1 0 6
mpt4 1 0.01 2 0 2 0.00 0.01 1 0 5

Average fcd0 90 4.79 162 135 27 4.66 0.12 1 0 6
Average mpt0 1 0.00 1 0 1 0.00 0.00 1 0 6
Average mpt4 1 0.00 1 0 1 0.00 0.00 1 0 5
That gives me the IO column

Or the sar -d -R 2 2, that gives me the number of reads and writes:

HP-UX dbnode1 B.11.31 U ia64 07/12/11

13:38:35 device %busy avque r/s w/s blks/s avwait avserv
13:38:37 disk6 1.00 0.50 0 2 32 0.00 5.80
disk11 1.50 0.50 0 2 32 0.00 6.42
disk65 1.50 0.50 5 4 168 0.00 1.72
disk66 0.50 0.50 4 4 136 0.00 0.56
disk67 58.00 0.50 88 2 30118 0.00 6.41
disk71 0.50 0.50 4 2 160 0.00 0.56
13:38:39 disk6 1.50 0.50 0 3 35 0.00 14.40
disk11 1.00 0.50 0 2 33 0.00 8.85
disk65 18.00 0.50 32 3 5867 0.00 5.11
disk66 1.50 0.50 3 3 43 0.00 2.19
disk67 58.00 0.59 112 1 30518 0.02 5.23
disk71 0.50 0.50 3 0 96 0.00 1.15

Average disk6 1.25 0.50 0 2 34 0.00 10.96
Average disk11 1.25 0.50 0 2 32 0.00 7.77
Average disk65 9.75 0.50 19 4 3018 0.00 4.43
Average disk66 1.00 0.50 4 3 90 0.00 1.26
Average disk67 58.00 0.55 100 2 30318 0.01 5.75
Average disk71 0.50 0.50 3 1 128 0.00 0.78
dbnode1[24]/root #


FR

[Brett Williams]

Glance can also be used to show the throughput.  Under the IO By Disk section ("u" as the hot-key) you can see the Physical IO for the given device asssociated with the NetApp.  Glance works best when invoked within an hpterm.  Glance shows you the information in real time and can summarize the information since the utilty was invoked.  If you want to export the historical data from the server then use the "extract" utility to report on the information you're looking for.