One of the new features with Exchange 2010 is the client access array. When configured correctly, this is quite a useful feature. In my view it is something that should be configured on all Exchange 2010 servers, even on a single server deployment.
Background
The full explanation of the CAS Array feature is available on Technet, but in short, the reason it was introduced was due to the changes in the way connections to the database are now handled. With Exchange 2007 and older, Outlook connected directly to the mailbox server (unless using Outlook Anywhere). With Exchange 2010 all clients now connect to the CAS servers. The CAS servers then manage the connection to the database.
With the Database Availability Group (DAG) meaning that an active mailbox could be moved between servers easily, connecting directly to the mailbox server wasn't really practical.
The simple way to think of a CAS array is like a virtual Exchange server. Clients see this virtual name instead of the actual name of either the CAS server or the mailbox server.
Why you should configure a CAS Array
If you are deploying multiple CAS servers, or a DAG, then a CAS array is pretty much mandatory. However if you are on a single server, or are separating the mailbox and the CAS role on to separate machines, then a CAS array is still of value.
If you have ever done a migration or disaster recovery, one of the key pain points has been getting Outlook to point to the new server in a timely manner. As long as the original server was alive, then Outlook will redirect to the correct server automatically. During a migration though, it may not be possible to get all clients to connect to the old server in a timely manner and the old server has to be removed.
However as the CAS array is simply a DNS entry and a small configuration in Exchange, it is completely under the control of the network administrator. A change to the DNS will make all Outlook clients point to another server.
If there is a possibility at any time in the future of additional Exchange servers being introduced, or the CAS role moved to its own server, the use of the CAS array from the start will become invaluable for easing that transition. All MAPI clients will use it, so as well as Outlook, this can also include things like Blackberry Enterprise Server.
CAS Array Configuration Notes
Ideally the CAS array should be configured before any mailboxes are moved to Exchange 2010. If you don't, then the clients that are moved will use the true name of the CAS server, and even after the CAS array has been configured, they will not change unless the mailbox is moved between servers or the Outlook profile is changed.
If CAS Array is therefore introduced retrospectively, it can produce mixed results if all clients haven't been updated with the new value some how.
You can use the CAS array with Network Load Balancing (NLB), but if the server has all of the roles and is also a DAG member, then you must use an external load balancer. Using NLB on the same server as the DAG is not supported.
A CAS array cannot go across Active Directory sites. Therefore if you are doing a two host DAG, with the second (passive) host in a data centre or similar, and have separated the AD sites, you will need two CAS arrays. In the event of a full failover, you will need to change both the CAS Array value on the database and the DNS. While this is a manual intervention, it does mean the process remains under your control.
The CAS array host does not have to be in the SSL certificate, simply because Outlook doesn't make any http connections to that host name.
You should not use the same host name for other services, particularly anything that is being accessed externally (like OWA), but you can use the same IP address and therefore NLB virtual IP.
For example, you could use outlook.example.local as the CAS Array host, then mail.example.com for OWA, SMTP, Outlook Anywhere etc.
If your internal and external domain are the same, then ensure the internal name doesn’t resolve, externally so no wildcard in the domain etc. Failure to do so will result in a confused Outlook, and will probably mean Outlook Anywhere has performance issues, if it connects at all.
Finally, on the DNS entry for the CAS array, turn the TTL time down. This will ensure that if you do have to change the host name IP address, it is picked up quickly.
Background and Configuration of the CAS Array: http://technet.microsoft.com/en-us/library/ee332317.aspx