Re: Fw: Fw: forcing native access with websphere and toolbox

Alex Garrison wrote:

It definitely does have a huge impact on our dasd utilization for just the
reason you stated.  However, 8.5gb drives for our as/400 are only $1600
installed, so dasd is cheap.
-----------
Incredible.  I just bought a 20GB, 7200RPM drive for $200.  While prices
are coming down, $200/GB as opposed to $10/GB is still a big margin.


Legacy impact: again definitely a concern.  We had to recompile/modify all
rpg programs that use these files.  All the rpg code is ile sql rpg.  We
found the evals still were ok, but moves had to account for the fields
being
twice as many bytes.  It took about a day to fix all that.  However, we
were
very fortunate in that our design isolated the web system from the main
body
of our legacy system (millions of lines of rpg and thousands of files)
through a series of data queues.  Any data in/out of the web systems move
through these data queues, so we really only had to fix the legacy code
that
interfaced the web databases to the dataqueues.
-----------
What I am going to do in my classes is make sure that I do as much local
caching of information as possible.  I'll let you know how that goes.  But
since my primary technique involves intercepting the 5250 buffer, I'm not
going to have a whole lot of luck.  Of course, I suppose I could do it by
making sure all my external data structures for the data transfer are set
up with CCSID 13488.  Will RPG automatically convert a character field from
a non-unicode database file when it is moved into a subfield from a unicode
data structure?  If so, that could make things very interesting... I wonder
if the RPG conversion routines are faster than those in the JVM?  I bet
they are...


When you are desperate for performance, you will try almost anything.  In
balance conversion to unicode was rather painless, and a 25% performance
gain is nothing to sneeze at.
-----------
No, not at all.  One way I found to greatly enhance speed in my technique
was to not translate the unused portion of the 5250 emulation buffer (I set
an arbitrary 2000 character buffer, and most record formats used 100 or
less characters).  A simple thing, but it boosted performance from several
seconds to sub-second response time.  I was quite impressed.

Joe


+---
| This is the JAVA/400 Mailing List!
| To submit a new message, send your mail to JAVA400-L@midrange.com.
| To subscribe to this list send email to JAVA400-L-SUB@midrange.com.
| To unsubscribe from this list send email to JAVA400-L-UNSUB@midrange.com.
| Questions should be directed to the list owner: joe@zappie.net
+---