On Thu, 2012-11-29 at 03:57 -0500, Massimo Biancucci wrote:
> From my observations, compression saves disk space but it's not a
> "good idea" from a CPU (and it's obvious) and elapsed time (less
> obvious) point of view.
Elapsed time? Better see for yourself if you haven't already. I saw
elapsed time improvements on an ESCON-attached V2X2 with selected
compressed datasets ("write once read many"). As you say, it depends.
> For instance, in a simple read-compress/write-compress program (few
> logic), CPU consumption is really high (60% more) and Elapsed too (25%
> more) if compared to read-compress/write-on-tape.
Compression is more expensive than decompression.
> Do you use compression ? What's the rule of thumb do you use ?
I cherry-pick datasets that I know will benefit and set them to an
appropriate DATACLAS in the ACS routine. You don't want to broadly
apply this because users who want to update records in place will be
ringing you up in short order.
--
David Andrews
A. Duda & Sons, Inc.
[email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
