In this instalment of the series of posts on the architectural considerations for designing a predominantly-Microsoft IT infrastructure, based on the MCS Talks: Enterprise Infrastructure series, I’ll look at some of the architecture considerations relating to providing high availability through redundancy in the infrastructure.

The whole point of high availability is ensuring that there is no single point of failure. In addition to hardware redundancy (RAID on storage, multiple power supplies, redundant NICs, etc.) consideration should be given to operating system or application-level redundancy.

For some applications, redundancy is inherent:

• Active Directory uses a multiple-master replicated database.
• Exchange Server 2007 offers various replication options (local, clustered or standby continuous replication).
• SQL Server 2008 has enhanced database mirroring.

Other applications may be more suited to the provision of redundancy in the infrastructure – either using failover clusters (e.g. for SQL Server 2005, file and print servers, virtualisation hosts, etc.) or with network load balancing (NLB) clusters (e.g. ISA Server, Internet Information Services, Windows SharePoint Services, Office Communications Server, read-only SQL Server, etc.) – in many cases the choice is made by the application vendor as some applications (e.g. ISA Server, SCOM and SCCM) are not cluster-friendly.

Failover clustering (the new name Microsoft cluster services) is greatly improved in Windows Server 2008, with simplified support (no more cluster hardware compatibility list – replaced by a cluster validation tool, although the hardware is still required to be certified for Windows Server 2008), support for more nodes (the maximum is up from 8 to 16), support for multiple-subnet geoclusters and IPv6 as well as new management tools and enhanced security.

In the final post in this series, I’ll take a look at how to build an infrastructure for data centre consolidation.