Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ihnp4!ptsfa!lll-lcc!ames!oliveb!jerry
From: jerry@oliveb.UUCP
Newsgroups: comp.unix.questions
Subject: Re: TK-50 tape backups -- could they go faster?
Message-ID: <1409@oliveb.UUCP>
Date: Mon, 15-Jun-87 16:06:40 EDT
Article-I.D.: oliveb.1409
Posted: Mon Jun 15 16:06:40 1987
Date-Received: Wed, 17-Jun-87 01:45:40 EDT
References: <241@kosman.UUCP>
Reply-To: jerry@oliveb.UUCP (Jerry F Aguirre)
Organization: Olivetti ATC; Cupertino, Ca
Lines: 24

In article <241@kosman.UUCP> kevin@kosman.UUCP (Kevin O'Gorman) writes:
>Running dump(8) on a MicroVAX, Ultrix 1.2A -- and I'm getting REAL tired
>of waiting for the tape.  The tape unit is a streamer, and the software
>is not fast enough to keep it in streaming mode, so I wait and wait and

Modifications to dump were distributed on the net.  They are also in the
4.3BSD dump.  The basic change is for dump to fork several copies of
itself.  This creates a "ring" of dumps that write to the tape as soon
as the previous process finishes.  The effect is to provide double
(actually multiple) buffering.

I tried the changes here and the improvement on a regular tape drive was
about twice as fast.  A friend at sci tried it out on a MicroVAX-TK-50
combination and said the improvement was about 5-8 times.  This is
because the streaming comes into effect.  If you have the source to your
dump routine then I could send you the patches.  Otherwise DEC will
presumably include this, along with other 4.3BSD changes, in a future
realease.  Someone should really bug them to make the dump inhancements
immediately available, expecially to TK-50 users.

A side effect of the dump speedup is a greater load on the system.
Depending on the speed of the CPU vs the tape, dump can create a load of
3 or more by itself.  This is, of course, not a factor if the system is
single user.