Oracle slow with Multinet - anyone else same problem

Alpha OpenVMS 7.2-1, Multinet 4.2 with all ECO`s, Oracle 8.1.6

When I connect to Oracle using TCP/IP Multinet, it is more than 10 times slower than when using mailboxes. It`s possible that there is more overhead but it`s more than that.

Does anyone have the same problem?
We have Gembase and also Cobol programs using Gembase and both are slower when connecting via TCP/IP (10 minutes => 100 minutes) The only thing I can see is that when using IP, the processes (application and Oracle server) are always waiting on something. No CPU, no IO. Just wait time. The total number of CPU seconds and IO count is the same upon job completion, except that with IP it takes 10 times the elapsed time.

I am trying to identify the cause (with the help of Multinet support), but with no luck yet.

Please respond if you:

a) Have the same problem
b) Have UCX and can compare speed with the 2 type of connections
c) Know of a way of correcting the problem

The workload must be on the application as well as on the Oracle side. An UPDATE XX SET XYZ=0 in SQL*PLUS for a 1,000,000 rows table will not do the job 'cause Oracle does it all. No IP traffic. A Cobol program doing a loop to retrieve each row sequentially would be good.

Thanks for any assistance you can provide!

--

Syltrem
http://pages.infinit.net/syltrem (OpenVMS related web site)
To reply to myself directly, remove .spammenot from my address