JPAContainer with large data

Flow
1

I make some tests using JPAContainer.
At this moment i want to know something about how efficient is JPAContainer with large datasets.
I have an table with 100.000 dummy rows
With trivial initialization see that all data (number of rows) from my table are loaded into container.
My questions:

  1. data are loaded entirely, or are fetched only id’s, then when table request new items to be rendered, are fetched next rows.
  2. where are stored JPAContainer in memory? To server or on client?
  3. How much memory need on server to handle supposing about several hundred clients. Application have more tables, not all so big obviously.

Thanks!!!