Jerry writes:

> To correct the original problem, DBGeneral users should use the high-speed
> dataset/database erase option (4.4) to again, erase the database.  This
> feature works at the file extent level to fill all allocated file extents
> with zeros, which is why it is not subject to the problems encountered by
> DBUTIL/ERASE's cut-back/expand technique.  This is also why DBGeneral's
> ERASE feature is faster than DBUTIL (doesn't deallocate/re-allocate extents)
> and others techniques which may incur the performance cost of using file
> system writes.

DBUTIL doesn't deallocate/re-allocate extents.  For non-jumbo datasets,
it simply opens them with OUT access  ... that immediately causes a
call to fill_disc, which erases the extents (without deallocating/reallocating
them).

For some datasets, an extra pass is made, re-erasing
the data (this happens in a small percentage of the datasets)...using
an "FWRITE" loop.  HP is aware of this, and is working on optimizing it.

For jumbo datasets, the chunks are opened for INOUT access, and an
"FWRITE" loop is used to erase them.

However, in no case are the extents deallocated!

--
Stan Sieler                                          [log in to unmask]
                                     http://www.allegro.com/sieler.html