In <H000132102922173@MHS>, on 07/06/99
   at 11:38 AM, Thomas Denier <Thomas.Denier AT MAIL.TJU DOT EDU> said:

>My site is using an IBM 3494 tape library with a VTS (Virtual Tape Server) to
>store ADSM data.

>Most specific volume mounts of virtual tapes for ADSM take about five minutes.
>I understand that the process is more complicated than mounting a real tape,
>but even so, five minutes does not seem an impressive performance for what is
>advertized as a highly advanced tape technology.

I don't have a VTS, but I have attended session at SHARE about the VTS.  I
came to the conclusion that a VTS was not good for ADSM.  You seem to
be confirming it.  I thought it would be particularly bad to be doing 
collocation to a
VTS hosted sequential pool.  You would be constantly thrashing the disk storage
that pretends to be many tapes as you recall a whole tape just to extend it by
a little.  And I would not want to give up collocation to make the VTS work 
better.
(I have 1200+ nodes.)

During a restore, do you have to get the whole tape image back before the CCW to
do a high speed locate will complete?  That would be very bad.

>I have occasionally seen virtual tape mounts for ADSM take forty minutes or
>more with no obvious cause. The people who manage the VTS insist that these
>episodes are normal behavior (they usually say something about the length of
>time required to forward space a ten gigabyte stacked volume). I find this
>claim implausible in the extreme, but I don't know how to refute it.

Ask the VTS managers if any of the tapes have lost VCR.  The real tapes and
drives should be able to position to the data you request within 15 seconds of 
the
actual mount.  If VCR has been lost, the high speed search on the real tape is 
disabled.

I have similar problems with STK 9840's.  They some lose their MIR (same idea 
as vcr)
and the a restore can take 40 minutes.

Quickly doing the numbers in my head, I would think that any mvs mount for a 
400Mb
3490 should be completed in under 90 seconds.  10 sec for robotics, 15 sec for 
hi-speed
search and 65 sec to read the tape and write the disk.


>Are there any IBM documents with concrete data on the expected mount times for
>virtual tapes?

--
--------------------------
--------------------------
Bill Colwell
Bill Colwell
C. S. Draper Lab
Cambridge, Ma.
bcolwell AT draper DOT com
--------------------------
=========================================================================