the CPYF to DDMF does not require exclusive allocation
Please explain how the above is accurate.
If I use CPYF to copy 10,000 files, and they are all being copied into one lib named Targetlib, then users MUST be locked out of TargetLib from the time of the first CPYF *REPLACE until the 10,000th file is replaced, correct?
So in effect the entire TargetLib requires no users to have access for the duration of the 10,000 CPYF executions.
Is this incorrect?
From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of CRPence
Sent: Thursday, September 05, 2013 3:07 PM
Subject: Re: Data Transfer
Applying a false implication\perception of how the CPYF via DDM
functions, then probably there would be no chance that the Copy File via
DDMFile would be able to compete. However accepting that the DDM is not
limited to single-row I/O and the CPYF to DDMF does not require
exclusive allocation [except for the clear to effect MBROPT(*REPLACE)],
then the answer is most definitely instead, that "It depends." And a
bulk copy is *always* best effected using FROMRCD(1) except when the
target data must be ordered. The biggest problem I see with CPYF, is
that use of CPYF will not be able to deal with isolation nor generally
/transactional/ update activity; at least not without great care, where
the care taken would likely be implemented with exclusive allocation of
the data. All that said... Save and Restore is almost always the
fastest and best for homogeneous movement of large amount of database
data; external media might be the best option to limit impacts mostly to
the source system, but very fast disk, available CPU, and large
pipelines can suffice for some scenarios.
On 05 Sep 2013 12:31, Stone, Joel wrote:
Wouldn't this approach take multiple times as long as FTP'ing a save
file? If he already has an issue with time, this would entail even
And, the test system would be unusable during the transfer (to
prevent any file locks since the CPYF would require exclusive use of
each file) correct?
The FTP approach he is estimating at 20 hours, wouldn't a
record-by-record CPYF thru a DDM pipe take many multiples of that
Hopefully I am wrong, but it seems that SAVOBJ is 10 times faster
than CPYF on a LOCAL machine (even with FROMRCD(1)). If a DDM pipe
is introduced, that is even more of a bottleneck I would think.
Monnier, Gary on Thursday, September 05, 2013 11:43 AM wrote:
If all you are interested in is copying the data from production
to development and you already have the physical and logical files
on the development system try using a combination of DDM and CPYF.