Are you implying that you can use more than 1 channel per tape drive? 
(effectively, you can't)

A task, in this case, a backup or restore, can only use 1 path to the
DASD at the same time, and 1 path to the tape at the same time.  So your
thruput speed is limited to the slowest (disk/tape) path you have.

In the case of 8 ESCON channels to the 3494ATL, you can have 8
backup/restores (or 4 tape to tape copies), going at the same time,
without loosing performance on the tape side.  But if you are actively
using only 1 tape, you are only using 1 channel.  So you are limited to
17 MB/s ESCON, being written to the tape at 9 MB/s (3590B) or 14 MBs
(3590 E/H) without tape controller compression.  If you are using tape
controller compression, then the 17 MB/s will be your bottleneck.  An
that won't feed the beast.  And the drive will shoe shine, which is very
bad for performance.

Tom Duerbusch
THD Consulting

>>> Alain Benveniste <[EMAIL PROTECTED]> 11/27/2008 12:54 PM >>>
Tom,

At restore time we have 8 ESCON per tape drives in a 3494ATL.

alain


Le 26/11/08 20:30, « Tom Duerbusch » <[EMAIL PROTECTED]> a
écrit :

> Channel/device speeds
> 
> If you are on ESCON, with 17 MB/s and going to an IBM 3590 tape
drive, for
> example...
> 
> The B drives write to tape at 9 MB/s.
> The E/H drives write to tape at 14 MB/s.
> 
> If you use tape controller hardware compression, say 3:1, you need to
send at
> 3 times the data rate as the drive can write to the tape.  i.e. the
controller
> on a B drive, will have to receive data at a 27 MB/s, in order to be
able to
> write to the tape at 9 MB/s.  Not going to happen with ESCON.  The
E/H drives
> are even worse.  You need FICON channels to handle these drives.
> 
> And, you need FICON channels on the DASD to be able to read the data
at these
> high rates.
> 
> Otherwise, you shoe-shine the drives. Very time expensive.
> 
> For my 3590B drives with 10 GB tapes, it takes about 75 minutes to
fill the
> tape with tape controller compression turned on over ESCON channels.
> 
> So, what kind of drives do you have and what kind of channels are
they
> attached to (both disk and tape)?
> 
> Tom Duerbusch
> THD Consulting
> 
>>>> Alain Benveniste <[EMAIL PROTECTED]> 11/26/2008 11:35 AM >>>
> We are several guys to misunderstand something about compression.
> Here is the situation:
> Our interrogations come from MVS but I think it is true fo VM too.
> 
> We did 2 backups with DFDSS, 1 with the COMPRESS option and 1
without.
> For the 1st one, the restore too 3h20, the 2nd 6hrs ! 25 output tapes
for
> both tests. This was made on 2084 with 10 procs,and really enough
mips.
> 
> Ok the software compression is very cpu consumming, hardware no. The
> question is what a software compression does, IDRC doesn't. Several
of us
> think the algorithm used is very similar from one to the other, so !.
Why
> that difference ! Others think, that IDRC is off, even if it is coded
in the
> system (I verified this twice). Others says, we didn't pay to use
IDRC.
> Others remember we had to install IDRC in the old past. So we don't
have it.
> 
> I'm looking for a firefly in this dark mess.
> 
> Alain Benveniste 
> 

Reply via email to