[jibi]

By the way, if these are only in your test environment, I would HIGHLY recommend changing the names prior to moving into production. Name them something like BES01-A and BES01-B. It avoids a LOT of confusion down the road.

Also, if you're going to be hosting more than 2,000 users, you'll want to move the BlackBerry Administration Servers to their own system, if that is possible. The Core BES components are not resource intensive, by any stretch of the imagination (not much of a notch more than BES 4.1.x), but the BAS services are memory hogs and could impact your BES servers with that number of users on them. As for the total number of users, I would take what hdawg says as applicable to BES 5.x as well... it doesn't seem scaling has improved much (assuming you're using their recommended hardware configurations and scaling guidelines) - adding BAS onto the same servers doesn't help with those figures.

My current QA environment is as follows:
BlackBerry Administration Service:
Load Balancer VIP Name: bbadm-us.domain.com
Server 1 Name: basqa01.domain.com
Server 2 Name: basqa02.domain.com

BlackBerry Enterprise Server (Core BES):
Server 1 Name: besqa01-a.domain.com (Primary)
Server 2 Name: besqa01-b.domain.com (Standby)

BlackBerry Configuration Database (SQL 2005 w/ Mirroring):
Server 1 Name: bbdbqa01-a.domain.com (Primary)
Server 2 Name: bbdbqa01-b.domain.com (Standby)

Failover components will have the same server name with a partner designation (-a and -b). Pooled components will have incremental server names (01, 02, 03, etc). Not only is this best practice for these sort of configurations, but it is also the most practical method for implementation in this scenario.