Clustering Solutions for Exchange Server

Have you given up clustering as a viable solution for Exchange Server 5.5 reliability? Hang in there; help is on the way in Platinum.

If you've already explored the possibility of deploying Exchange Server on Microsoft Cluster Server (MSCS) clusters, you already know the scoop. If you haven’t explored this possibility, let me give you the short story. Exchange Server 5.5 Enterprise Edition was the first version of Exchange Server to support clustering. This support included only the bare necessities, which translated to support for Active and Passive configurations only. As a result, Exchange Server can run only on one cluster node at a time. In addition, Exchange Server has no native support for the clustering APIs. Microsoft chose to simply support the generic resource DLLs for MSCS. By supporting MSCS via the generic resource DLLs, Exchange Server is severely limited in its failover capabilities in a cluster. Essentially, Exchange crashes on one node and does a soft recovery on the other during failover. Clustering in any environment adds a degree of management, which is the case for Exchange Server. Operators must not only understand Exchange Server operational procedures, but must also add clustering to the mix. This bundle of caveats for deploying Exchange Server 5.5 on MSCS clusters has made justifying it and finding adequate return on investment difficult for most organizations.

Does Platinum provide any relief? In my opinion, Platinum will provide true enterprise support for clustering. Microsoft was able to make the necessary architectural changes in Platinum to provide a decent clustering story. One of Platinum's most touted features will be the multiple information store (IS) support. Along with this, Platinum introduces the concept of an information storage group (ISG). An ISG is a JET/Extensible Storage Engine (ESE) instance that manages one or more ISs sharing a common set of transaction log files. As applied to a clustering scenario, the ISG is the unit of failover in a cluster. For example, in a four-node Windows 2000 (Win2K) cluster, with each node running Platinum, you could configure three ISGs per node. You can configure failover preferences for the ISGs so that if one node fails, the cluster will distribute the three ISGs from the failed node among the three remaining nodes (for a total of four ISGs each). Although this scenario is just an example, it might be the most common scenario organizations deploy.

I know what you’re thinking. If the Exchange Server 5.5 scenario makes management complex, the Platinum scenario will be much more complex. It's true that Platinum will bring a new administration paradigm. Companies will have to rewrite their backup and restore procedures, and staff will have a steep learning curve with all the changes that come in Win2K, Active Directory (AD), MSCS, and Platinum. But if you start thinking along these lines now, you’ll be ready when Platinum hits the streets.