High Availability Options for Windows

Whether you are configuring Windows 2000 Server or Windows Server 2003, you are referring to different forms of Windows Clustering. With any version of a stand-alone Windows server ”unless you employ RAID for the system disks ”clustering or using third-party offerings is the only way to ensure that another server will start to perform the work if the server and its applications fail. Even if you configure a duplicate server with the same software and operating system options, you would need a process to switch to that server. This is where a server cluster and Network Load Balancing might be able to assist you in increasing availability to your Windows installation.

Windows Reliability Features

In addition to clustering, there are other availability and reliability features built into Windows that are better covered in other documents. Here is a list of topics you can use to research more information.

Windows 2000 Server

The following were introduced with Windows 2000 Server to increase availability and reliability:

• Reboots minimized for some configuration tasks

• Slipstreaming of service packs and chaining of hotfixes

• Application failure recovery with IIS restart and out of process application protection

• Safe mode boot

• System Recovery Console

Windows Server 2003

The following were introduced with Windows Server 2003 to increase availability and reliability:

• Tools to measure and audit, including the already mentioned Shutdown Tracker, enhanced logging, and system tracing

• The overall hardware and software installation has been improved, including:

  • Driver install improvements

  • Driver rollback

  • New driver preference using .inf files

  • Hotplug Peripheral Component Interconnect (PCI)

  • Side-by-side assembly install

  • Hot add RAM

  • Signature support

• Volume Shadow Copy (VSS) Backups and enhanced backup and defragment APIs

• Fault tolerance system support through things like machine check architecture for 64-bit, multipath I/O and Load Balancing and Failover (LBFO)

• Better recovery through processes like autosystem recovery, last known good enhancements, up to an eight-node server cluster, geographically dispersed clusters, Majority Node Set (MNS) clusters, process cycling for IIS out-of-proc applications, COM+ object recycling, and service restarts

• Maintenance tasks have also been improved, such as hot patch operating system hotfixes, no reboot hotfixes, and update improvements

• Testing and qualification

  • Driver Verifier and App Verifier

  • Device, system, and domain controller Hardware Compatibility Tests (HCTs) and qualifications improved

  • Certification improvements

• Services decomposed

  More Info

  For basic information about server clusters and Network Load Balancing, see Windows Clustering in Chapter 3, An Introduction to Microsoft High Availability Technologies.

  Important

  You cannot combine a server cluster and Network Load Balancing on the same physical hardware. They can coexist on the same network or domain, but they cannot be configured to run together on Windows installation.