If you are considering creating an Exchange Server cluster, one of the biggest decisions you'll have to make is whether to create an active/active cluster or an active/passive cluster. In this article, I explain the differences between the two Exchange clustering models and the pros and cons of each.
The concept of active/passive and active/active clusters has to do with how many cluster nodes are being used. In an active/passive cluster configuration, at least one node acts as a hot spare. This hot spare is unused unless a failover occurs. In an active/active configuration, each of the cluster nodes is actively running one or more Exchange virtual servers.
Although this concept sounds simple (and it is), there are actually a lot of limitations that you need to be aware of. The biggest limitations have to do with Exchange virtual servers, which you can think of as the number of logical Exchange servers that your cluster is running.
If you were to create an active/passive cluster, the cluster could contain up to eight nodes (assuming that you were running Windows Server 2003 Enterprise Edition and Exchange Server 2003 Enterprise Edition). Of those eight cluster nodes, up to seven of them could be running an Exchange virtual server. However, each node can only run one Exchange virtual server.
Therefore, the rule of thumb is that you take the total number of nodes in the cluster and subtract one -- that's the number of Exchange
As I mentioned before, in an active/passive configuration, one cluster node is being used as a hot spare. Should any of the other nodes fail, the Exchange virtual server that the node is running will fail over to the spare server.
The active/active cluster configuration does not reserve a node as a spare. All of the cluster nodes can run one or more Exchange virtual servers. Microsoft advises staying away from active/active clusters though, and imposes some serious limitations on them in order to avoid potential disaster -- and with good reason.
For starters, an active/active cluster is limited to two nodes. If you want more than two nodes, you will have to use the active/passive configuration. If you are running an active/active cluster, you must never have more than 1,900 simultaneous connections to a node, and the individual CPUs must never exceed 40% capacity for more than 10 minutes at a time. Otherwise, bad things can happen.
So why the CPU and connection limits? If one of the nodes fails, there is no hot spare waiting to pick up the slack. The server's entire workload is shifted to a node that is already servicing at least one Exchange virtual server. If either node has above a 40% average CPU utilization, a failover will likely overwhelm the remaining node.
Another active/active cluster configuration limitation is the number of storage groups that each Exchange virtual server can support. An Exchange 2003 cluster running in active/active mode has the same storage group limitations as a standalone Exchange server (four storage groups with a maximum of five databases each).
The reason for this limitation is that, if a failover does occur, all the cluster's Exchange virtual servers would be running on a single node. The node has a physical limitation of four storage groups with five databases each. Since Exchange does not allow this limit to be exceeded, you must guarantee that the sum total of all Exchange virtual servers does not exceed what could be run on a single Exchange server.
The active/passive configuration does not impose this limit, because a node never runs more than one Exchange virtual server.
At this point, I'm sure you've realized that an active/passive configuration is greatly preferred to an active/active configuration in most cases. You may be wondering, however, if there is ever any benefit to using an active/active configuration.
The only time that I could see running an active/active configuration would be if the cluster needed to host Exchange virtual servers for two or more tiny companies, and hardware and software costs were a huge concern. If you've only got a few users in each Exchange organization and never expect those organizations to place much demand on the servers, an active/active configuration is an option.
One thing that you need to be aware of, though, is that an active/active configuration is much more prone to virtual memory fragmentation than an active/passive configuration. This means that you will have to occasionally reboot the cluster in order to avoid fragmentation-related errors.
About the author: Brien M. Posey, MCSE, is a Microsoft Most Valuable Professional for his work with Windows 2000 Server and IIS. Brien has served as CIO for a nationwide chain of hospitals and was once in charge of IT security for Fort Knox. As a freelance technical writer he has written for Microsoft, TechTarget, CNET, ZDNet, MSD2D, Relevant Technologies and other technology companies. You can visit Brien's personal Web site at http://www.brienposey.com.
Do you have comments on this tip? Let us know. Related information from SearchExchange.com:
This was first published in June 2005