Hi Michael,
Thanks for your feedback.
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]
Sent: Monday, October 18, 1999 4:34 AM
To: Linux SCSI
Cc: Robin T.Miller; [EMAIL PROTECTED]
Subject: dt under Linux
Hi Robin,
I've downloaded and tried dt under Linux 2.2.12 (SuSE 6.1). Looks quite
promising. I'm going to spread this tool around my company, making
communication SW somebody definitely can make use of it.
I'm glad to hear it. It's gotten pretty popular at Compaq/DEC too.
I grabbed the latest source and found that I first need to do a 'make
depends' otherwise it will look for some RedHat include paths which I don't
have, of course. A simple 'make' will not do. Maybe the Makefile.linux need
some tweaking (I'll look into this asap) or the source-tar is not really
clean... No big problem, actually.
Ok, I'll see how to work around this. Perhaps remove the dependency list.
Here are some questions:
1. I'm intending to use dt for my SCSI dat tape (SONY SDT-7000). Can you
provide me with a script that determines:
a) Total capacity
b) Speed
When I test for tape capacity, I usually give 'dt' a very large data limit
like 'Inf',
and let 'dt' detect the end of media on writes/reads. In fact this is one
of our usual tests
to ensure EOM is reported properly by the tape drive and the O/S driver.
When testing tapes with which support compression, it's imperative to use a
non-replacing pattern, or 'dt' may run for days! For capacity testing, try
using:
# dt of=/dev/st0 bs=64k limit=Inf pattern=iot
To test variable block sizes, you can try:
# dt of=/dev/st0 min=1 max=64k incr=3 limit=250m pf=pattern_6
File 'pattern_6' is a random I/O pattern, or you can create your own.
For speed testing, we usually test with different block sizes, and to find
"best"
performance we usually enable POSIX Ascynhronous I/O (AIO) on Tru64 Unix.
This
latter feature is not fully supported on RedHat Linux, as 64-bit offsets
didn't compile
properly, and there was incorrect data corruption using their AIO thread
implemented library.
When doing raw performance testing, using "disable=compare" inhibits
filling the
data buffer on writes, and disables compare on reads, which gives better
performance.
The Linux tape driver will do buffering for fixed block tape devices, i.e.
write behind
and readahead, but this doesn't help you if your tape drive is using
variable length blocks.
2. Must I specify the device type under Linux when testing a tape device (I
guess: yes)?
You might have to, depending on the name of your tape device. Internally,
'dt'
will detect certain device names on non-Tru64 O/S's to set the device type.
For Linux,
since I'm using SCSI tapes, 'dt' detects 'st' or 'nst' to set tape device
type. Perhaps
this needs to be expanded?
3. Under Linux, I get 'fsnyc(): invalid parameter' or a similar message
when testing. Do I need to switch O_FSYNC off? fsync() seems to be
available (using SuSE 6.1 I still have to use glibc 2.0.x).
Yes, Linux doesn't like an fsync() to tapes, so I added a check in dtgen.c
to
avoid this operation based on tape device type. A "dtype=tape" should fix
this.
dt has lots of options! Can be a bit confusing... RTFM definitely helps!
Yes, that's why I finally updated the user's manual after 4 years :)
I'm overdue to put together a Frequently Asked Questions (FAQ) list.
Thanks!
Hope this helps get you started.
Feel free to ask more quesitons.
Cheers,
Robin
Mit freundlichen Gruessen / best regards
Michael Kwasigroch
FaxPlus/Open Development
________________________________________
eMail: [EMAIL PROTECTED]
INTERCOPE GmbH, Hamburg, Bebelallee 72
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
