How to integrate C++ server in weblogic

Question

I have developed a C++ class implementing a kind of server. It
receives requests from clients, processes the requests (by
accessing ODBC databases), and returns the results back to the
client. Then it remains waiting for the next request, never
finishes.
Now I am trying to integrate this server into a Weblogic
environment to take benefit of scalability, availability,
performance, ...
I would like to know what are my choices:
a) I think that servlets are not suitable as I do not want my
server to finish execution after serving a request. But what
about servlets + EJBs. Could my server be implemented with a
combination of both?
In that case I suppose that I cannot take advantage of the JDBC
conection pool from my C++ class, can I? ...
b) I could make a java wrapper for my C++ functions in order to construct an RMI object. I wonder if this is possible.
c) I could implement my server as a CORBA object. Then the
weblogic server would have an active IIOP connection pool for
say 5 or 10 instances of my CORBA object. Do I really need to
purchase Weblogic Enterprise? Can I do something else with the
Weblogic Server?
In that case the ODBC connections would be established at the
CORBA startup time.
c) Other?
Thanks in advance.

Answer 1

Comments inline...
David Ruana wrote:
Eduardo, does your solution require the Weblogic Enterprise product or just the Weblogic Server?No, WLS will not work with WLE until the next point release of WLE.
Is a C++ ORB the same as a CORBA object?Yes.
>
>
On the other hand, I have wrapped the C++ class by means of JNI so that I have the corresponding java class available. Then the implementation of a RMI server is quite straightforward. I am just beginning with RMI but it seems far more easy that CORBA, isn't it?Yes, the only trouble is that iff your C++ object is colocated with the WLS server, you could take the whole server down iff the C++ code faults. What would also work, without this side effect, is if you wrapped the C++ code in a WLS client and bound the client side RMI server into the WLS server.
As I see it, the main advantage of this choice would be that the Weblogic Enterprise product would not be required (since Weblogic Server itself supports calling RMI from servlets, for example).
Please some further advise on this issue would be appreciated.
Thanks. David
Eduardo Ceballos <[email protected]> wrote:
Host your C++ server on a C++ ORB. BInd a reference to it into the COS Naming/WLS JNDI tree. Delegate calls from (into) the C++ object using WLS support for IIOP. This makes it so that the native code does not take down your server when (if) it goes south.
David Ruana wrote:
I have developed a C++ class implementing a kind of server. It
receives requests from clients, processes the requests (by
accessing ODBC databases), and returns the results back to the
client. Then it remains waiting for the next request, never
finishes.
Now I am trying to integrate this server into a Weblogic
environment to take benefit of scalability, availability,
performance, ...
I would like to know what are my choices:
a) I think that servlets are not suitable as I do not want my
server to finish execution after serving a request. But what
about servlets + EJBs. Could my server be implemented with a
combination of both?
In that case I suppose that I cannot take advantage of the JDBC
conection pool from my C++ class, can I? ...
b) I could make a java wrapper for my C++ functions in order to construct an RMI object. I wonder if this is possible.
c) I could implement my server as a CORBA object. Then the
weblogic server would have an active IIOP connection pool for
say 5 or 10 instances of my CORBA object. Do I really need to
purchase Weblogic Enterprise? Can I do something else with the
Weblogic Server?
In that case the ODBC connections would be established at the
CORBA startup time.
c) Other?
Thanks in advance.