Karl,

I'm the AMISYS developer that worked on the memory-mapped performance
enhancements that relate to the two files you mentioned (MMS0800S and
MMS0800C).  As you probably know, these files are flat-file copies of
the SUPER-TABLE and CODE-DETAIL data sets that are maintained in sorted
order by their corresponding TurboIMAGE keys (SUPER-KEY and
DETAIL-CODE#).

The performance enhancements achieved by the use of these files is not
so much the result of them being memory-resident (or not), but rather
because they can be searched by a binary search algorithm using
memory-mapped I/O techniques.  In benchmark testing I did, doing lookups
on the memory-mapped files was approximately 15 times faster than doing
the corresponding TurboIMAGE DBFIND/DBGET sequence to retrieve the data
from the data base itself.  Because of the large number of accesses that
AMISYS makes to the SUPER-TABLE and CODE-DETAIL data sets to retrieve
configuration information, we've found in out testing that the use of
these memory-mapped files can have a significant impact on the claims
adjudication process (typically somewhere around a 15% improvement in
run times).

The use of these memory-mapped files is optional.  If you purge these
files, AMISYS will "fall back" to doing conventional DBGET/DBFIND
sequences to retrieve SUPER-TABLE and CODE-DETAIL records.  You may wish
to try things out with and without these files to see if you can see a
difference.  As with almost any performance issue, YMMV.

John Schick
Sr. Technical Development Advisor
AMISYS, LLC

-----Original Message-----
From: Karl Hancock [ mailto:[log in to unmask]
<mailto:[log in to unmask]> ]
Sent: Thursday, October 18, 2001 12:59 PM
To: [log in to unmask]
Subject: Amisys performance - files in memory?


Once again trying to address our Amisys 10.0 performance problems.
Seems that one of the differences between previous versions and 10.0 is
that two files were created, MMS0800S and MMS0800C in the data group.
The ideas was that with everyone accessing these files then these files
would always reside in memory and that would increase performance over
dataset reads.

I'm wondering if there is anyway to determine if a file or parts of it
are residing in memory at a particular moment?

Thanks.

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *