Performance of ODBC vs. other access methods

["Dave Odom" <Dave.Odom@xxxxxxxxxxxx>]

Nathan, Vern, et al,

I, too, have found certain implementations of ODBC fairly efficient in
tools like QMF for Windows when talking to non-DRDA machines, but when
using tools like Crystal Reports and its implementation of ODBC, my CPU
goes out of sight.  So, in the case of vendor's tools it does depend on
how they implement ODBC and the rest of their processes.   

> From what I've been able to experience and glean from others so far
when building remote access apps or tools the following are the best
performing access methods:

- Have your app call System i5 apps written in some native compiled
language with imbedded SQL therefore creating static SQL   -  Best
access
- App calls application APIs stored on the i5 written in a high-level
language
- App calls REXX application that calls SQL or application stored
procedures 
- App calls SQL or application stored procedures using the IBM .NET or
the Java "provider"
- App calls SQL or application stored procedures using the older OLE
provider 
- App calls SQL or application stored procedures using ODBC provider
- App sends SQL strings through the above "providers" using the same
order of preference as above  - Worst of these possibilities

Something like that.   I'm sure you all will put in your opinions as to
my list.    I'd also like to hear about the basic premise that ODBC is a
slow performer/"provider" compared to OLE,  the IBM .NET provider or the
Java provider and certainly slower than calling stored procedures or
native language APIs.

Thanks,

Dave