[Info-vax] TK50 - this is annoying...
Paul Sture
nospam at sture.ch
Wed Oct 10 06:40:26 EDT 2012
In article <5074b4c7$0$47944$c3e8da3$dd9697d2 at news.astraweb.com>,
JF Mezei <jfmezei.spamnot at vaxination.ca> wrote:
> On 12-10-09 18:30, Stephen Hoffman wrote:
>
> > http://downloads.quantum.com/sdlt320/handbook.pdf page 2.3
> >
>
> very pleasant to see Quantum acknowledge Digital's leadership/expertise
> and good design for what became the DLT system.
>
> People poo-poo the TK50, but the general design has lasted until today,
> so it couldn't have been all bad. Sure, they changed the positioN/number
> of rollers, but they kept many of the principles (such as a read head
> after the write head to check that the data was written properly).
>
> When discussing how fast the TK50 wasn't, was this due to the tape
> advance speed relative to data density ? Or was it just how the
> controller on the Q-BUS was implememted ?
>
> It seemed to be that once the TK50 had found your data, it wasn't so
> slow loading it. (aka: data transfer speeds didn't seem that bad, but
> the "seek" to get to it was terrible (understatement)).
IIRC backing up a pretty full RD54 (150MB) took somewhere between an
hour and an hour and a half for the first tape. I would set off my
weekly full backup on an evening and go home, putting the next tape in
the following morning.
I got some tips on BACKUP block sizes etc out of the documentation which
came with a TK70 drive, and that probably helped. IIRC the VAXstation
2000 was restricted and could only write blocks of 16K maximum to TK50s.
> A "DIR MUA0:" to find the name of the savesets on it wasn't a "go out
> for coffee", it was a "go back home for dinner, watch TV, get 8 hours of
> sleep, wake up, get breakfast and goto office hoping the command had
> ended" type :-)
I remember it more as a lunch break rather than going home thing, but I
would put it into a batch job, since a typo followed by CTRL/Y might
have you waiting ages for the tape to finish the current operation.
BUT. In the V4/V5 era VMSINSTAL would use DCL READs on the source drive
to display a list of available products on that drive. That of course
involved traipsing all the way to the end of tape, before you even got
started on a software installation. There was a similar waste of time
if you specified the wrong product name as a parameter to VMSINSTAL.
I used to avoid this by COPYing the lot to disk first, and installing
from there. Software installation from disk has always been
substantially faster than tape. Faster than from CD too.
The same philosophy applied when CD distributions came along. If you
have a spare drive, VMS installations can be greatly speeded up by doing
a BACKUP/IMAGE from the CD to that drive and booting from that.
One note about doing VMSINSTAL from disk. I came across many a
suggestion that you needed to put savesets into [000000]. This was only
true for the VMSINSTAL kits that insisted they went there. Properly
written (and tested) kits would allow you to specify the source as being
in a standard ddcu:[directory]
--
Paul Sture
More information about the Info-vax
mailing list