Everyone know EJB could do distributed computing.So I have some questions here.If someone could clarify my concept.I will be grateful and thanks in advance.

1.In my idea,there are 2 kinds of distributed computing:
a)many request at a same time,and the web server need load balance
b)one request but the computation load is heavy(since the logic is complex,it cost many CPU time),so it's better distribute its load to serveral computer
is that right?
How to implement the 2 situation by EJB?
How to implement on Weblogic or other application server

2.about JNDI:
if there is a Bean A with JNDI name "smaple\A" on computer 1,there is a Bean B with the same JNDI name on computer 2,
should they conflict?
if EJB client and EJB server deployed on different computers,when the EJB client want to find the EJB server by JNDI,it must specify the URL of the EJB server located.But if there are 2 computers deploy the same EJB server and how the EJB client could find.(we use 2 comupters so that if one is down another could work)

3.is BMP is slower than CMP?
as we can see in the code,every time BMP store or load a data from database,it will generate a statement from connection, and close it after using.So I don't think the performance will be good.Compare to use DataSource directly,is Entity Bean will be better?Why?