That's a strange response.  From MSKB 815180 we know that the max active
users per EVS in an active/active design is 1900.  We also know that this
represents 40% of the max capacity of a node.  We can infer that a single
node is good to 3800 active users or so, still leaving 20% headroom.
Given this, A/A/A/P should suffice.  Of course this assumes that all users
are active at the same time.  If your concurrency rate is lower, you could
go higher.  I've done 7500 users per EVS/active node in low concurrency
environments.

On the front end, it depends on quite a few things as well, including usage
pattern, hardware ssl offload, etc.  In most cases two FE or three FE
servers with NLB will suffice.

Note:  Each node in an Exchange MSCS cluster requires a Windows 2003
Enterprise and Exchange 2003 Enterprise license.  Front end servers require
only Windows 2003 and Exchange Server 2003, not the enterprise versions.

Understood.
Hope it works out well for you.  I noticed that John posted a response
questioning the suggestion of a 6-way cluster (a/a/a/a/p/p).  From what
you've said to date, you're not likely to pull that off.  My past experience
tells me that folks unable/not willing to read the docs are not good
candidates for clusters in the first place.  To successfully use Microsoft
(or any vendor for that matter) clusters, you really need a high level of
discipline to keep from melting down your environment.

One thing you'd find in that documentation I pointed out is why the
recommendation to use 6 nodes (4 active and 2 passive). That's important to
understanding what makes the cluster work. Understanding why you would even
want to cluster in the first place is incredibly important.  Cluster is very
expensive and even if money is no object, is highly complex and takes a
great deal more time and discipline to make work. IMHO, you'd be better off
using stand-alone BE servers.  Likely just max them out (HP makes some
really nice DL 58x series machines; something comparable might be useful to
you). Disk layout is incredibly important as you scale this up.  If you
instead went with multiple smaller BE servers, say a farm of them, you could
split the users down and therefore your risk of an outage due to hardware
failure, significantly, however your maintenance would be more difficult as
well.  The tradeoffs continue as you continue down the line figuring out
what's important to you and what's not.

In the end your question was very simple.  The information needed to
accurately answer, guide you is not present.  Your results and the answers
you receive based on your simple question will vary.

Like I said, I wish you all the best and hope it works out well.

Al