Re: Topas - wrong disktransfer numbers?

From: Jean-Marc Monnez (monnez.jean-marc@AGORA.MSA.FR)
Date: Wed Nov 13 2002 - 05:44:12 EST


Holger,

I think topas says 458 kilobyte per disk per second (not Megabyte) in
your results.
This would lead to a global 1.77 Mbytes/s.
If I am right, it's not 10 times too much, but 100 times less than from
other figures.

I can't help with 5.1, hereafter my results on 4.3.3 :

from /usr/bin/topas (perfagent.tools 2.2.33.0)
Disk Busy% KBPS TPS KB-Read KB-Writ
hdisk0 89.9 8781.4 2195.3 8781.4 0.0
hdisk1 0.0 0.0 0.0 0.0 0.0

from /usr/bin/iostat (bos.acct 4.3.3.0)
Disks: % tm_act Kbps tps Kb_read Kb_wrtn
hdisk0 87.2 8439.2 2109.8 84392 0
hdisk1 0.0 0.0 0.0 0 0

Problem with 5.1 ?
Regards

-- JMM

-----------------------------
Jean-Marc MONNEZ
MSA / AGORA / ATD
monnez.jean-marc@agora.msa.fr
-----------------------------

Holger.VanKoll@SWISSCOM.COM wrote :

> Hello,
>
> I am doing some benchmarking of my internal hdisks (dd </dev/hdisk0-3
> >/dev/null bs=256k); Aix 5.1.
>
> Topas tells
>
> Readch 183.3M
> Writech 183.3M
>
> So 183 Megabyte per second are read (from disk) and written (to
> /dev/null).
> Thats also what I roughly measure using date ; dd ... ; date.
>
> The disc-statistic is what I dont understand:
>
> Disk Busy% KBPS TPS KB-Read KB-Writ
> hdisk0 100.3 458.5K 114.6K 458.5K 0
> hdisk1 100.3 453.9K 113.5K 453.9K 0
> hdisk3 100.3 455.6K 113.9K 455.6K 0
> hdisk2 99.3 455.8K 114.0K 455.8K 0
>
> Thats 458.5 Megabyte per disc per second. Way too high transferrate
> (10 x times too much).
> Or am I misinterpreting?
> Furthermore I wonder why I see Readch at 180M (as before) and Writech
> at 0 (or other way round) sometimes at higher monitoring-intervals.
>
> Iostat 10 gives
> Disks: % tm_act Kbps tps Kb_read Kb_wrtn
> hdisk0 99.8 46903.1 11725.7 469148 0
> hdisk3 99.9 46632.3 11658.1 466440 0
> hdisk2 100.0 46906.3 11726.5 469180 0
> hdisk1 100.0 46617.5 11654.5 466292 0
>
> Looks correct (184M/sec).
>
> I only have one aix5.1 node here. Can anybody confirm that topas gives
> 10 x too much disk-transfer-numbers?
>
> Regards,
>
> Holger



This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:19 EDT