IOT or Cluster or neither

I am evaluating a good table design to store data with the following characteristics.

• Large number of records >200million rows
• Tables have a unique "id" for each row (generated using sequence).
• The rows can be logically grouped per "user_id"
• The rows are primarily accessed per "user_id"
• The distribution of the rows is not even among "user_id" i.e. Most users have <50 rows but there are a few who have >10000 rows.
• The tables need to have some secondary indexes.
• The tables have "minimal" updates and "moderate" inserts.

In order to increase the performance I have the following options.

A) IOT with primary key similar to "userid,id"
B) Cluster the table on user_id (index)
C) Cluster the table on user_id(hash)

Since the distribution of the data over "user_id" is not equal I'd like to get feedback from folks who have found themselves in similar situations and what solutions they have chosen and why.

Appreciate any input/comments

-- 
Adi Alurkar
aalurkar_at_linkedin.com
--
http://www.freelists.org/webpage/oracle-l