> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
> Behalf Of Chad Cooper
[snip]
> I am working with a commercial product that does context
> sensitive searching
> of tech docs. The core engine was written in C++ 32-bit DLL's, wrapped by
> Java. It talks to a MS SQL 2000 server through a JDBC>ODBC bridge. The web
> server is IIS 5, and the web pages are ASP pages that talk to a COM +
> object. The COM+ Object is actually a third party COM+>JAVA CLASS bridge,
> which ties into the Java wrapped 32-bit dlls. We have to run
> SUN's JVM for
> Win32 because Microsoft does not support its own version of the JVM.
>
> So this is the path the data takes for each web request
> BROWSER>ASP>IIS>COM>Sun JVM>DLL>compare data to binary database>sun
> JVM>JDBC>ODBC>MS SQL>ODBC>JDBC>Sun JVM>DLL>Sun
> JVM>CLASS>COM>IIS>ASP>BROWSER.
Good heavens. ODBC by itself will eat a ton of cycles. This is for text
search? And it uses the RDBMS to do the text searching? Why not just use
an off-the-shelf search engine that has ODBC (or even better, native) and
Web server hooks? One that starts with 'V' comes to mind.
> The vendor is asking us how many people they expect will be using the
> product concurrently. I think that they feel there might be a performance
> issue with their product, but they are not disclosing it to us.
> They are planning on making this a off-the-shelf product.
I'm guessing that ONE concurrent user might see performance problems...
when the system isn't being maintained, which I'd guess will be often, given
the number of components.
Nick
