Exchange Server

PDPHell · Guest

Looking for anyone's input on a proposed Exchange 2003 design.

Currently have 7 old NT4 based Exchange 5.5 servers - 5 mailbox/PF, 2
Bridgeheads split across 2 locations.
6000 Mailboxes.
Outlook XP clients

Its proposed that this is all migrated to another site 400+ miles away
running on a win2k3 - Exchange 2003SP1 - Four node cluster with San
storage. No local Exchange servers, no bridgehead/front end servers.
Outlook 2003 Clients using cached mode. (200mb limit)

The WAN is currently limited to 24Mb by its slowest link to an MPLS
cloud with potential to be upgraded to 100mb. Eventually 1Gb links are
planned. Like any WAN, the full bandwidth is not guaranteed/shaped
exclusively for exchange.

Some test lab analysis has been done using Loadsim, using a profile
based on our typical usage which was determined as "heavy+" using a
combination of stats and guidance from the excellent document - "Client
Network Traffic with Microsoft Exchange 2003"
Although this provided some figures, it does not consider latency and
the lack of QOS on the network link at present.

Does anyone have any experience running anywhere in the region of 6000
mailboxes over a WAN connection.
After a lot of searching, i cannot find any decent exchange "network"
sizing documents or case studies. It seems that everyone with this
amount of mailboxes would always recommend local servers.

Advice, comments, ridicule - All Appreciated !

Jonathan Norris · Guest

I have done this several times. It works well assuming you have Outlook 2003
with the latest service packs running cached mode or http(s)/rpc.

What you will want to do is ensure they/you prepare the user community for
the change.

Additionally I would always follow best practices and ensure they/you know
your user profiles to ensure this is the proper solution. I have done this
with rather suspect trans-continental links so its rather forgiving.

Also there is the option of having users use the OWA interface if they are
not very heavy users.

If in doubt I recommend using a pilot to ensure the solution will work as
advertized.

The issue with publishing such documents is that (we) can't make too many
assumptions. Every environment has its limitations and its usually too
complex to to provide the proper guidance.

This is where most folks hire consultants.

Al Mulnick · Guest

I've seen a lot more over a lot less do just fine. I've also seen a lot
less over much more do horribly in terms of performance.

I agree with Jonathan that you should run this as a pilot before declaring
it a success. The variables in your environment will change over time as
well, so be aware of the risk and communicate it well. There should be no
surprises when the PTB decide that all related infrastructure will be run on
the WAN and no local server resources will exist and suddenly you get
complaints of slow mail because somebody decided they should download a SQL
db for "testing" purposes. The list of risks goes on, but you get the idea.

I disagree with Jonathan about the consultant however. I doubt a consultant
will be able to tell you anything pertinent in an environment not yet built.
You'll have to do the work.

One last thing: you'll want to ensure that your servers in the core site are
not a performance bottleneck. That means your GC's and Exchange servers are
the fastest part of the email chain at any given point under any given load.
You'll also want to be careful of Anti-Virus settings to be sure that
client-side-mail-aware AV doesn't impact your performance in ways you don't
expect. ;)

Al

"Jonathan Norris" <JonathanNorris@discussions.microsoft.com> wrote in
message news:20B33FAE-85D9-477F-B6BD-2FE80F30E1D6@microsoft.com...
[quote]I have done this several times. It works well assuming you have Outlook
2003
with the latest service packs running cached mode or http(s)/rpc.

What you will want to do is ensure they/you prepare the user community for
the change.

Additionally I would always follow best practices and ensure they/you know
your user profiles to ensure this is the proper solution. I have done
this
with rather suspect trans-continental links so its rather forgiving.

Also there is the option of having users use the OWA interface if they are
not very heavy users.

If in doubt I recommend using a pilot to ensure the solution will work as
advertized.

The issue with publishing such documents is that (we) can't make too many
assumptions. Every environment has its limitations and its usually too
complex to to provide the proper guidance.

This is where most folks hire consultants.

	Exchange Server Forum Index -> Design	All times are GMT
Page 1 of 1