Article View: mod.computers.vax
Article #4022Why is BACKUP/COMPARE so slow?
From: LLACROIX
Date: Tue, 31 Mar 1987 06:23
Date: Tue, 31 Mar 1987 06:23
35 lines
2382 bytes
2382 bytes
BACKUP/COMPARE (or the verification pass if /VERIFY is used) is really
slow on my system, and I'm wondering if anyone can either tell me what
I'm doing wrong, or at least offer sympathy :-). Because of its lack of
speed we don't do /COMPARE often, but we would like to afford the luxury
now and then.
The configuration: 11/750, VMS 4.5, RA81/UDA50, TU80.
The backup command to create the tape:
$ backup /fast /group=50 /record /ignore=(interlock) /list=xxxx.lis -
/image /rewind dua2: msa0:xxxx.sav /label=xxxx /save_set /buffer=5
The command to compare:
$ backup/fast/image/compare msa0:xxxx.sav/rewind/buffer=5 dua2:
When creating the backup, the tape drive frequently streams at 100 ips,
and streams at 25 ips most of the rest of the time. During the compare
run, the tape drive rarely streams at 100 ips and is often in 25 ips start/stop
mode. The compare pass takes at least twice as long as the original backup
pass.
The machine is not running out of CPU cycles--Montior consistently shows
more than 50% NULL time. If disk were the bottleneck then the tape wouldn't
have been streaming during the first pass--it wouldn't be getting data
fast enough. On the other hand, tape speeds aren't the bottleneck, otherwise
the tape drive would be delivering data at maximum speed (100 ips streaming).
So I conclude that the multi-buffering/read-ahead scheme for the comparison
pass is not as good as during the creation of the save set. That's neither
surprising or unreasonable, but perhaps disappointing. So...am I screwing
up in an obvious way? Any other suggestions? I can't do physical backups,
sorry.
=======
Les LaCroix
SPSS Inc., 402 Washington, Northfield MN USA 55057-2027
llacroix@carleton.edu (csnet)
Technical Details
Message-ID:
<8703311446.AA19801@ucbvax.Berkeley.EDU>Path:
arc1-ca.pugleaf.net!utzoo!utgpu!water!watnot!watmath!clyde!rutgers!mit-eddie!genrad!decvax!ucbvax!carleton.EDU!LLACROIXShow Full Headers
Complete Article Headers
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: arc1-ca.pugleaf.net!utzoo!utgpu!water!watnot!watmath!clyde!rutgers!mit-eddie!genrad!decvax!ucbvax!carleton.EDU!LLACROIX From: LLACROIX@carleton.EDU.UUCP Newsgroups: mod.computers.vax Subject: Why is BACKUP/COMPARE so slow? Message-ID: <8703311446.AA19801@ucbvax.Berkeley.EDU> Date: Tue, 31-Mar-87 01:23:00 EST Article-I.D.: ucbvax.8703311446.AA19801 Posted: Tue Mar 31 01:23:00 1987 Date-Received: Thu, 2-Apr-87 07:04:04 EST Sender: daemon@ucbvax.BERKELEY.EDU Distribution: world Organization: The ARPA Internet Lines: 35 Approved: info-vax@sri-kl.arpa