Chapter 8: Leveraging Exchange Clusters

In his book In Search of Clusters , Gregory Pfister defines a cluster as “a parallel or distributed system that consists of a collection of interconnected whole computers which are utilized as a single, unified computing resource.” While the concept of clustering has been around in the computer industry for many years, the technology is relatively new to the Windows environment. In Microsoft’s endeavors to make the Windows technology ever more reliable, clustering technology is one area to which they have turned. Microsoft’s approach to clustering has differed from other operating system vendors. As organizations begin to take Windows Server 2000/2003 technology more seriously and deploy more and more mission-critical applications on this platform, clustering will become increasingly important.

For Exchange 2003, clustering technology offers significant reliability benefits for organizations willing to invest in this technology. As the Exchange Server platform becomes a basis for more application solutions and business-critical processes, the benefits of clustering will become quite necessary. In addition, as Microsoft continues to strive to position Exchange 2000/2003 as a platform for Internet Service Providers (ISPs) and Application Service Providers (ASPs)—Hosted Messaging Services, high-availability technologies such as clustering will be a high-priority requirement. As such, it is necessary for us to spend some time discussing the clustering technology available in Windows Server and how this technology can be leveraged to provide mission-critical capabilities for Exchange deployments.

8.1 High availability is a key concern

Organizations deploying messaging and collaboration applications face many challenges. Very often, implementers focus on performance and scalability. Among the most often-heard questions, How many users per server can I deploy? How should I design storage for Exchange Clusters? How do we do performance monitoring and tuning? Equally important are the administrative and management aspects of deployment. Issues such as user administration, directory synchronization, database maintenance, and server management are common concerns. However, disaster recovery and high availability are at the forefront of the minds of implementers of Microsoft Exchange Server. Due to the mission-critical nature of messaging and collaborative applications, minimizing downtime and providing for rapid recovery are paramount. With the increasing size of information stores combined with the limits of current backup and restore technologies, disaster-recovery issues can be the limiting factor when you are deciding on the number of users per server to deploy. Scheduled downtime is also an issue, since most organizations have much more scheduled than unscheduled downtime. Microsoft estimates that its own Exchange deployment’s scheduled-downtime to unscheduled downtime ratio can be as high as 6:1. With tolerance for downtime being low, keeping an Exchange server constantly available is a key concern for IT organizations. The issue of high availability (reliability) is among the most important issues for Exchange deployments. When many look to increase availability and reliability for Exchange Server, cluster technology immediately comes to mind.

In this chapter, I will try to avoid an in-depth focus on Windows clustering technology itself. My focus will be mostly on support and implementation of clustering in Exchange (from versions 5.5 to 2003). Since this book is geared toward Exchange 2003, I will spend the majority of my effort on Exchange 2000/2003 (which are fundamentally identical other than some service pack and minor feature improvements as far as clustering support is concerned), other than to discuss the origins of clustering support for Exchange Server that were first supported in Exchange Server 5.5 Enterprise Edition. I will give a brief overview of the history and technology surrounding clustering technology as a foundation for later discussions. Next, I will spend some time focusing on support of clustering within the Windows Server 2000/2003 operating system implemented as Microsoft Cluster Service (MSCS). I will follow this with a close examination of how clustering support for Exchange 2000/2003 is designed, implemented, and managed and discuss deployment best practices. I will take special care here to call out any notable differences in clustering between Exchange 2000 and Exchange 2003. Finally, I will also look at some of the business and decision criteria that impact one’s choice to implement cluster technology for an Exchange deployment. My goal here will be to give you a look at clustering technology for Exchange 2000/2003 and how this technology can be leveraged to increase reliability for your Exchange deployment.