Exchange 2007 Server Roles and You

Before the Beta 2 release, I wanted to write a column about server roles in Exchange 2007. A lot of misinformation is circulating, largely because Microsoft hasn’t done a good job of answering the questions Exchange administrators are asking about the new roles. Microsoft's new FAQs at http://www.microsoft.com/exchange/preview/faq.mspx will help somewhat, but as a public service, I want to address some of the questions I hear to help clarify what you need to know about the new server roles.

One question I often get is whether these roles can, must, or should coexist on a server. The Exchange 2007 server roles are modular, and you can install (or remove) them independently of one another. You can put four of the five roles (Mailbox, Hub Transport, Client Access, and Unified Messaging--UM) on a single server and dynamically add or remove any of these roles at any time. The Edge Transport role has to be installed on a separate server, but that’s by design; remember that it’s intended to live in the perimeter network, while the other server roles should be restricted to the internal network.

That explanation leads to a related question: Do you have to have an Edge Transport server role? The answer, of course, is no. You get some additional security and message-hygiene features if you do, but the Hub Transport server role can be easily configured to interchange messages with the Internet. As for the other roles, the Client Access, Mailbox, and Hub Transport roles are mandatory; you must have at least one of each in your Exchange organization. (Well, I guess you could get by with no Mailbox roles, but that wouldn’t be a very useful setup!)

Exchange admins who use clustering tend to ask lots of questions about Exchange 2007’s cluster support. The Mailbox server role can be clustered, but the other roles cannot. This makes perfect sense when you think about what those other roles do: They process transient data, and they all provide dynamic discovery tools so that clients and other roles can locate the currently available instances of a role. For example, if the Client Access server that your mailbox points to is down when you launch Microsoft Office Outlook 2007, the Outlook Autodiscover process (which runs when Outlook starts, when the Client Access server or Mailbox server fails to respond after a given interval, and at an administrator-specified interval) will locate another available Client Access server and Mailbox server.

Speaking of clustering, the answer to another common question is, Yes, with cluster continuous replication you can build a hot-site failover system based on a two-node cluster that has no single point of failure. Mailbox data can be replicated between cluster nodes so that the surviving node can seamlessly pick up operations if the primary node fails. This is a big improvement over Exchange 2003’s clustering implementation.

What about scalability? At this point, that’s a hard question to answer. The only people who really know what kind of loads the new roles can handle are at Microsoft, and they’re not saying much. (The UM team, however, has said that a UM server should be able to handle about 100 concurrent phone calls.) It’s premature to start making architectural decisions based on preliminary scalability numbers or claims; as we get closer to product release, I expect to see better-defined parameters for how to size servers to handle the various roles.

What should you be doing to get ready for Exchange 2007? That’s a fertile topic of discussion, and one that I’ll start talking about next week.

End of Article