One of our production server,
with below details runs fine with 100 or less users.
When more than 200 users come,
we feel a lot of buffer busy waits, when any one of the table
goes in db_file_sequential_read for more than 5 seconds.
And also reading in undo is very poor.
Attaced a statspack report for your reference.
Os level, the datafiles are with RAID 5 (Though I told the sysadmin
that raid 5 will not suit, he told this will not be a big problem).
Redo log file s are in RAID 1.
SOlaris 8.
Oracle 9i.