Does moving SP code from DB to App Server help scale an application?

Question

We have an application developed using Oracle 10g Forms / 10g DB? All our processing is done using SPs. So they all run in the DB server. Even our Inserts/updates/deletes to a table are handled by SPs.
The site with the maximum simultaneous users (i.e. concurrent users) is one with 100 concurrent users.
We have prospective customer whose requirement is 300 concurrent users. Our application won't be able to handle it since the DB server is a single processor server with limited memory.
One suggestion was move the SPs to the App Server by moving them to the Form. Since OAS has a PL engine they will run in the App Server and hence remove the workload of the DB.
I don't buy this. My point is, even if SPs are moved to the app. server still the SQLs will run in the DB server, right?
So what is the advantage?

Answer 1

christian, I just modified the original post thinking nobody will reply since it's very long. Thanks a lot for the reply. For others and myself also here is my original question.
I have a problem like this: Take this scenario. We have a TELCO app. It is an E-Business Web Application (i.e. Dynamic Web Site) developed using ASP.Net/C#. App. Server is IAS and DB is Oracle 10g. IAS and the DB reside in 2 servers. Both are single processor servers.
The maximum simultaneous user load is 500. i.e. 500 users can be working in the system at one time.
Now suppose 500 users login at the same time and perform 500 different operations (i.e. querying, inserts, updates, deletes). Now all 500 operations will go to the App Server. From there the C# code will perform everything using Oracle stored procedures (SP). I.e. we first make a connection to the DB, SP is invoked by passing parameters, it will perform the operation in the DB, send the output to the App. Server C# code and we will close the Oracle connection (in App Server. C# code).
Now, the 500 operations will obviously have to wait in a queue and the SQLs will be processed in the DB server.
Now, question is how does CONNECTION POOLING help in this situation?
I have been told that the above method of using DB SP to perform processing will make the whole system very slow since all the load of the processing has to borne by the DB Server and since DB Operations involve disk I/O it will be very slow. They say you cannot SCALE the application with this DB Processing mode and you have to move to App. Server processing mode in order to scale your application. I.e. If the number of users increases to 1000 our application won’t be able to handle it and will get very slow.
What they suggest is to move all the processing to the App. Server (i.e. App. Svr. Memory). They also say that CONNECTION POOLING can help even further to improve the performance.
I have some issues with this. First of all to get all the data to the App server memory for each user process from the DB will not only require disk I/O, it will also involve a network trip. That will obviously take time. Again the DB requests to fetch the data will have to wait in the DB queue. Then we do the processing in the App. Server memory and then we have to again write to the DB server which again will require a network trip and disk I/O. According to my thinking won’t this take MORE TIME than doing it in the DB server??
Also how can CONNECTION POOLING help. In C# we OPEN a connection ONLY just before run the SP or get the data and we close the connection just after the operation is over.
I don’t’ see how CONNECTION POOLING can improve anything?
I also don’t see how moving data into the App. Server from the DB Server can improve the performance. I think that will only decrease performance.
Am I right or have I missed something?
Edited by: user12240205 on Nov 17, 2010 2:04 AM