From: Lukas Beeler
Your way of benchmarking is ... interesting. Compare apples to apples.
Sorry, I'm not going to model my applications or architecture precisely after SAP's. I already admitted that the comparisons were rough. If you find little value in that, that's okay.
use full commitment control and journalling for everything
It would really surprise me if SAP's benchmarks included journaling. I don't think any major benchmarks have that requirement. If so, none that I'm aware of mention it. Not that I'm against journaling.
Regarding commitment control, the only time I've needed it was in a banking application where multiple transactions were accumulated in work files temporarily, then posted as a batch, and required a number of tables to be in balance.
But my order entry applications follow a traditional header-line model, where the order is entered, then a series of line items are entered. Each entry adds or changes a record on the server, immediately.
A number of thick-client applications accumulate order headers and line-items in PC work files, then post them to the server as a batch. They may need commitment control. But I don't subscribe to that model.
Under my header-line model, there's no need for commitment control. But RPG-based commitment control doesn't add much overhead, if or when needed.
What you're currently doing is benchmarking a light-weight web
application against a fully-featured framework.
Well, I subscribe to the old adage - make everything as simple as possible. In this case, the IBM i HTTP server forwards browser requests to an RPG-based server which natively interfaces with the database, then returns an HTML/JavaScript response.
-Nathan.
midrange.com
