Generic Vs Specific Model (RosettaNet)

Hi All,

I need suggestions from you guys. Here are the the points for you guys to read and than suggest whats the best way to do it please .

We are doing RosettaNet Integration.
We have mutliple backends and multiple partners.
There are two ways for us to do the models.

We do the generic model where we Call the partner specific mapping and services by invoking the partner folder which is based on PArtner DUNS Number. Also, we get the DUNS number from Model Pipeline.

Second is to do the model with filters and specific to integeration.

In Generic Model We are not currently setting the Filters in the model and try to have one model per one PIP for every backend and frontend.

The reason we are doing it is that , there is a statement somewhere that having more models on the machine is unacceptable overhead and slow response times on Process Run Time / Monitor.

Here is the question I have.

What is better and how? Having 50 Transactions of 10 models , or having 50 Transactions of 20 models. Does it really matter if you have more models on machine than what we have in memory?

Following are the issues with Generic model.

Complexity of Development.
Complexity of Support.
Complexity of Deployment.

Hoping to have some expert openion.