I keep reading everywhere: "just load the tables in parallel." What exactly does this mean? Do I just run two or more shell prompts in the OS and run sqlldr or imp on both of them? Is there a fixed limit on the # of sqlldrs that can be run at one time? Or is it just HW and OS dependent?

And how does Oracle manage the tasks it's supposed to handle? I.e. there are 3 users.
User #
1 System is dropping a schema
2 scott is loading an employee table of 1M records
3 Jerry is doing an ad hoc query that is doing a 6 way join where two of the tables are 20 Million rows and only 1 table is indexed.

How will Oracle start to schedule the tasks? FIFO? Least resources used runs most? Most resources used is run most? Equal time slice multi-tasking. Can this be tuned?