Date: Wed, 22 May 1996 21:09:32 +0200 (MET DST) From: Wilko Bulte <wilko@yedi.iaf.nl> To: jmb@freebsd.org, freebsd-scsi@freebsd.org (FreeBSD SCSI hackers) Subject: Tape performance measurements Message-ID: <199605221909.VAA01659@yedi.iaf.nl>
next in thread | raw e-mail | index | archive | help
Hi there
( I copied -scsi because I think there might be some interest there)
Below is the essence of Jonathan's message that triggered my
measurements.
> > > if you have a tape drive that your use successfully
> > > plese send me the dmesg output for the drive
> > > any quirks that the drive exhibits
> > > any tricks you needed to get the drive working
Environment: Asus TP4XE, P100, 16 Mbyte, 2.1R, used with the
-current dump (has the kbytes/sec stats).
2Gb 5400 rpm Quantum on NCR810, NCR810 shared
the same SCSI bus with the tapes.
DLT measurements:
-----------------
The dmesg entry:
(ncr0:6:0): "DEC TZ87 (C) DEC 9B3C" type 1 removable SCSI 2
Uncompressed, 10Gb mode.
bash# ./dump 0ubBf 32 10000000 /dev/rst2 /local
DUMP: finished in 717 seconds, throughput 1247 KBytes/sec
Compressed, 20Gb mode
bash#sh# ./dump 0ubBf 32 10000000 /dev/rst2 /local
DUMP: finished in 630 seconds, throughput 1418 KBytes/sec
Slightly hacked dump, allows blocking 64
bash# ./dump 0ubBf 64 10000000 /dev/rst2 /local
DUMP: finished in 657 seconds, throughput 1360 KBytes/sec
Comments: I had expected the 64 blocking to be better than the 32.
My experience is that the compression works better and the
drive is therefore faster.
I cannot keep the tape streaming, start/stop operation hurts
a DLT's performance really big. On a Sun Sparc 20 with Solaris 2.5
with the (identical) disk and the DLT on _seperate_ SCSI buses
I got almost 1.9 Mbyte/sec for an ufsdump of /usr. When tape
and disk were on the same bus on the Sun I also had start/stop
operation, and went down to something like 1.4 Mbytes/sec.
(Somebody wants to donate me a second NCR810 ? 8-)
Exabyte 8200 measurements:
--------------------------
The dmesg entry:
(ncr0:5:0): "EXABYTE EXB-8200 2687" type 1 removable SCSI 1
DUMP: finished in 700 seconds, throughput 227 KBytes/sec
Comments: If I remember correctly the Exabyte is speced for 240kbytes/sec.
So, running at spec I think.
Archive Viper 525 measurements:
-------------------------------
The dmesg entry:
(ncr0:3:0): "ARCHIVE VIPER 2525 25462 -007" type 1 removable SCSI 1
bash# ./dump 0ubBf 32 10000000 /dev/rst0 /usr
DUMP: finished in 845 seconds, throughput 188 KBytes/sec
Comments: this is on a 525Mbyte tape. Speced @ 240 kbytes/sec for 120 ips
tapespeed. 180 kbytes/sec for 90 ips speed.
The tape is kept streaming without problems.
-007 is no-good firmware. I'll get -011 and retry the test
Wilko
_ __________________________________________________________________________
| / o / / _ Wilko Bulte email: wilko@yedi.iaf.nl
|/|/ / / /( (_) Private FreeBSD site - Arnhem - The Netherlands
--------------------------------------------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605221909.VAA01659>