A colleague recently alerted me to a Network World article about how Exchange 2007 will shake up messaging. Whilst Exchange Server 2007 (formerly codenamed E12) will bring significant improvements that will require careful consideration and planning, I found the article to be highly misleading and thought I’d probably better set the record straight.

Firstly the article states that the new role-based architecture has the potential to require up to 5 types of Exchange server to be rolled out (up from just 2 with current versions); however it’s not that simple. The five roles are:

• Edge transport (message hygiene).
• Hub transport.
• Client access.
• Mailbox server.
• Unified messaging.

Exchange Server 2003 and earlier do not have message hygiene or unified messaging capabilities (so that counts for two of the new roles); however many organisations will have a separate product already performing message hygiene functionality so even that is not really an additional server to deploy (simply a case of replacing a third party product with a Microsoft one). Also (and crucially), four of the roles (all except edge transport) can be co-hosted on a single server if required. What the new role-based model really provides is flexibility in designing an Exchange server infrastructure.

The move to 64-bit architecture has come in for much criticism from some people but quite simply that is the way things are going. All servers sold in the last 18 months or so by a tier 1 OEM (basically HP, Dell and IBM) have had 64-bit capabilities even if they have had a 32-bit operating system installed so the Longhorn Server wave of products that will hit us in 2007 are time to “get with the program”. A 64-bit architecture removes many constraints (e.g. memory limitations) and allows for applications such as Exchange to scale more effectively, allowing larger mailboxes and greater consolidation.

The new clustering features are where there is the most uncertainly at the moment (features may well come and go before release); however the article refers to one user who doesn’t want his server to fail over from LA to Chicago. That may well be the case for some, but for many wouldn’t it be good if we could easily fail service over between two data centres? In any case, Exchange Server 2007 is likely to support three forms of clustering (on mailbox servers) – the current Microsoft cluster service, local continuous replication and continuous clustered replication (for geoclusters) so there are many options. As for only mailbox servers supporting clustering – so what! All other Exchange server roles either hold transient data or perform a client access role – load balancing is probably more appropriate.

My final issue is that the article points out that upgrading from Exchange Server 5.5 is not supported. Whether or not there are many organisations using it, Exchange Server 5.5 will be 4 releases old when Exchange Server 2007 hits the streets and is already unsupported. It is time to drop legacy platforms in order to make better use (dare I say “leverage”) Active Directory more effectively. This is another case of needing to “get with the program” – e-mail is being viewed as more and more critical by organisations and should not be left languishing on an outdated and unsupported platform.

As one would expect after 4 years (and for a major release), there are many additional features and enhancements planned for Exchange Server 2007, each with their own implications that need to be considered during the infrastructure design and implementation planning. Much of the information I have on Exchange Server 2007 was supplied under NDA but everything I’ve written here is available publicly and other information sources include the excellent Microsoft Exchange team blog, as well as Microsoft UK’s Eileen Brown (and of course my own ramblings here).