As of now My test application is simply creating a number of "dummy
records" (random data), but the final version is going to do all
kind of data manipulations (adding, changing and deleteing data).

I have uploaded both the IBX and IBO versions of my testprogram, and
I am running it on a dual Xeon 2.4 GHz (somebody else suggested that
IBO has this problem in a Dual CPU environment). Both version
contains a backup of an empty database or the DLL for building my
test database. These 2 version can be found at:

http://www.liljendal.dk/files/DBThreadTestIBO2.zip and
http://www.liljendal.dk/files/DBThreadTestIBX.zip

Normally I am using IBO (for my unthreaded/GUI-enabled) applications
and I would naturally prefer to use IBO for my (comming) threaded
service. However if I using IBX can have 500 records a second and I
using IBO can only have 200 then IBX is going to be my choice (I am
hopeing I've just made a stuid error in the IBO version).

Pelle

> There's probably a more ideal way to do what you are doing.

Perhaps you

> could let us know more what you are up to.
>
> IBO tends to be more stable than IBX in multi-threading

environments. There

> are numerous people who are here in the IBO fold because they

couldn't get

> IBX stable enough.
>
> Perhaps there is some trade-off in performance for the stability,

but my

> guess is you are probably doing something in a less than ideal way

that

> doesn't have anything to do with multi-threading.
>
> Please provide more information. Also, if you could provide a

sample app

> demonstrating this it would be awesome.
>
> Thanks,
> Jason Wharton