Providing High Availability for combined CAS, Hub & DAG servers…

One of the great changes in Exchange 2010 over Exchange 2007 is the ability to combine both CAS & Hub roles on mailbox servers configured as DAG, this gives you the flexibility to have a highly available solution with just 2 servers.

Note that you can still combine the UM role on the same servers however it’s not recommended because of the performance impact…

The challenge is that if you are having a requirement for load balancing the CAS or Hub traffic among the two servers you will not be able to use the Windows load balancing service “WLBS” and this is because simply it’s not supported to enable both WLBS and Clustering services on the same server. So the question now is how can you provide the required high availability to the CAS and Hub roles while maintaining the DAG and without using WLBS? to answer this question let’s split the roles.

Hub role:

Hub servers are automatically load balanced by Exchange 2010 for any servers to server submissions, you don’t need to configure any type of load balancing mechanism to load balance the mail submission traffic among the Exchange servers, however some of the environments are having SMTP clients or applications that use Hub servers to send/relay email to either internal users or to external users. To load balance these clients traffic there are two methods that you can use:

  1. Use a hardware load balancer and configure it to load balance the SMTP traffic across the Hub servers (preferred)
  2. Use DNS round robin (not preferred, and this is due to the fact that if one of the servers is experiencing a failure there will be a possibility that one of the clients will get this server IP as a response to the DNS query and will not be able to send the email however the mail will be released with the next retry attempt)

Note that by any method you will choose it’s a must that you don’t use the selected load balancing method to load balance the server to server communication…

CAS role:

CAS servers traffic can be load balanced using any of the below methods:

  1. Use a hardware load balancer and configure it to load balance the CAS traffic “Web & MAPI-RPC”
  2. Use ISA/TMG web farm publishing to publish and load balance the traffic for internal users however this will work only with the CAS web services such as OWA, Outlook Anywhere, etc… but not with the Outlook MAPI-RPC traffic if you are configuring your CAS servers in a CAS Array
  3. Use DNS round robin however as mentioned before, clients can get a DNS response for a failed server and then they have to re-attempt the connection to receive the next answer from DNS

If we will have a quick analysis to the above options, using a hardware load balancer is definitely the recommended option if you are going to combine the CAS/Hub with a DAG.