SUMMARY: measuring tape IO speed

• Previous message: selcuk karaca: "patch problems..."
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Date: Mon, 15 Dec 2003 18:24:28 +1000
To: 'Tru64-Unix-Managers <tru64-unix-managers@ornl.gov>

Many thanks to the multiple respondants, most people suggested the use of
'collect':

collect -st -i 10
collect -i5 -st

And to see aggregate performance for all disk and tape add -T qualifier:

collect -i5 -st -T

or tapex to measure statistics. Historically, disk devices are more
interesting so you see things like iostat [Dr. Thomas P. Blinn].

Other people offered pointers that /dev/zero and /dev/random aren't going to
get you nice accurate numbers. /dev/zero is highly compressable so the
numbers will look huge, /dev/random should be almost uncompressable [Charles
Ballowe].

More suggestions to ensure that the tape drive is on a seperate SCSI bus.
Personally, in these days of SCSI Ultra320, this doesn't seems to be such an
issue.

Regards,
Kevin

-----Original Message-----
From: Kevin Fleming [mailto:kevin@w617.com]
Sent: Thursday, 11 December 2003 8:12 PM
To: 'Tru64-Unix-Managers
Subject: measuring tape IO speed

since iostat doesn't allow the option of measuring IO speeds to a tape
device - how can it be measured.

I'm looking to indepedantly test IO performance - so the possibility of
using dd is limited. One option is to

dd if=/dev/zero of=/dev/tape/tape0 count=5000 bs=128k

or if=/dev/random, then workout the transfer rate per second, however would
like the same kind of output as iostat.

thanks,
Kevin

• Previous message: selcuk karaca: "patch problems..."
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]