We are currently developing a custom ERP system on V5R4M5 (soon to be 7.1). Our business model dictates packaged products require too much customization.
Our first test of the ERP system was a few weeks ago. With the write-once, share-many type architecture our programmers have written many SQLRPGLE GET programs that are called by other programs. These GET programs are also using SQL mostly in the form of SELECT * for getting the data. I am trying to get them to be column selective so indexing will help but we still have a memory issue.
As users drill through menu options from order entry, to parts lookup, etc. their jobs can exceed 120MB. The further they drill down the more activation groups they acquire since they are all *NEW. They are using work files instead of dynamic memory.
What is the best approach for better memory utilization? Because the activation groups are using so many modules and we cannot share the activation groups I fear the more users we put on the box the worse it will get. I believe more direct SQL coding in the programs without trying to share service programs is the answer but I'm hitting a wall with the CIO and programmers.