> >My comment would be to move the actual tape handling part out of DDR
> >entirely and let the data storage be handled by something else, eg a
> >pipe connecting to the input or output of the DDR engine. Then it
> >wouldn't matter how we were storing the data, and all the positioning
> >stuff could be out in user space, so DDR wouldn't need to know or
care.
> >Then we wouldn't need CMSDDR any more.


> I don't see how that removes the need for CMS DDR.  If anything it
seems
> to strengthen it because standalone DDR wouldn't have access to
whatever
> is doing the tape handling.

Sorry -- CMSDDR is a specific thing. It is a specifically modified
version of DDR that can store it's output in CMS files rather than on
tape. The Germans modified it and released it for download on the VM
Download page. It's very handy, but it's not part of the standard DDR
package. 

>  I don't think you meant we wouldn't need
> standalone DDR because that will always be needed as a last resort for

> when CMS is not available, and therefore it has to be able to read
what
> CMS DDR wrote.  This requires the media handling function to be part
of
> standalone DDR, which seems contrary to your comment.  Please clarify
in
> case I misunderstood you.

No, I agree that standalone DDR is necessary, but it's a matter of how
you use it, and if we want to ask IBM to do some work on DDR, I want to
get the most bang for the buck out of that work. 

I approach a standalone restore from the point of getting a minimal VM
system up ASAP, then use that to restore the rest of the system. Now
that you can get a full 3390-3 + IPLable DDR on one tape, I approach the
problem as having one "minimal" boot tape that gets me to the one-pack
system, then I've got CMS. I'd then use the pipe suggestion to handle
the tape positioning, because at that point, pipes can handle the
labels, and DDR just eats what it's fed and restores it to the disk it's
told to use. That one-pack system changes so rarely (thank heavens for
dynamic configuration in CP) that one unlabeled tape is considered
acceptable. (It could even be a labeled tape if I were willing to IPL
the tape twice in standalone mode to get past the label). 

An example of what I'm suggesting might look as follows: 

A) Writing to a labeled tape: 

PIPE DDR STREAM INPUT A | DDRCONN DDRDATA | TAP1 SL 

Where STREAM INPUT A might contain something like: 

SYSPRINT CONS
INPUT 123 3390 FOOBAR
OUTPUT STREAM DDRDATA
DUMP ALL 
END

DDR gets modified to create a connectable IUCV resource with the
specified name on the OUTPUT card. We write a pipe stage DDRCONN to
connect to the specified service and copy data to the output of the
stage. We then use whatever pipe function we want to store the data (in
the example, write to a standard label tape). 

B) Restoring from a std label tape

PIPE TAP1 SL 4 | DDRCONN DDRDATA | DDR RESTORE INPUT A

Where RESTORE INPUT A might look like: 

SYSPRINT CONS
INPUT STREAM DDRDATA
OUTPUT 123 3390 FOOBAR
RESTORE ALL 
END

In this case DDRCONN eats standard input, and creates a IUCV resource
with the specified name. DDR then reads that as input and writes to
output as specified. 

C) Standalone case

Preparation: 

Create a one pack VM system (you don't need spool or page space, as this
thing is really only for the purpose of running parallel restores).

MOVEFILE standalone DDR to front of tape and leave the tape there.
DDR one pack to same tape.

In Standalone case:

IPL one pack system tape.

DDR restore one pack system to spare disk using the usual methods.

IPL from one pack system. 

Log in as MAINT.

IPL 190

And you're to the point where you can use the pipe method to control
tape positioning again. 

I just want IBM to spend their development money to get us the biggest
bang for the buck. I think that having the I/O from DDR as something we
can get our hands on with other tools would give us a LOT more options. 

Does that help?

-- db

Reply via email to