SQLNet 2 & Win95/NT Client

Ok.....here's the deal.

There is a 16 bit application in use at this site which requires 16 Bit versions of SQLNet 2 and Oracle ODBC driver (Ver. 7.1 to be precise) to connect to the database, which operates on a DEC Alpha 8400 running OpenVMS 7.1, with ORACLE 7.3 at the back end. The back end is a rocket, and so it should be, with two lightning fast CPU's and 2.0Gb of RAM.

The problem is this....we are currently benchmarking performance and hence are running a fairly simple SQL script at the various network layers (eg. directly at the backend, on a client via. SQLNet, and through the ODBC layer), and are receiving very inconsistent results at the client end (ie. SQLNet and ODBC layers).

The SQL script simply creates a table, issues five hundred separate insert statements on the table, selects all records from the table, deletes all records from the table and finally drops the table.

At the backend it is consistently completing this task within 5 - 8 seconds (and this includes the connection to the database through SQLPlus). However, at the SQLNet and ODBC layers (for Windows 95 / NT client machines only) the results vary any where from 35 seconds and sometimes up to 6 minutes - at the SQLNet layer (slightly more for the ODBC layer)!!!

Initially we thought we had a problem with the network, being a large site with several subnets and thousands of computers scattered over the campus LAN/WAN. So we picked up the client PC and moved it within feet of the DEC Alpha...on the same subnet (in other words Mohammed went to the Mountain!!!), and ran the same tests over again. The result......no difference!!! The response times were exactly the same, and the client Win95 PC was still performing inconsistently, from 35 or 40 seconds to upwards of 6 minutes (again at the SQLNet layer).

Next step...move from the Win95 client PC to a Unix client and run the script again...and again...and again..and so on. The results were consistently coming back at around 40 seconds (at the SQLNet layer) - time and time again. Just to prove it wasn't a fluke, we then tried a MacIntosh client machine and got the same consistent results we did for the Unix client.

Hmmmmmm.........some opinions have been formed here, but I thought I'd throw this one out onto the Internet and see what I get back....any ideas??? And yeah, I know, you don't have to be rocket scientist to read between the lines...just give me the facts!!! Received on Thu Jun 04 1998 - 05:40:40 CDT