Software should be chosen for its ability to support the high-availability needs of the business system. Not all software is compatible with clustering or load balancing. Not all software must be compatible, either. Instead of making an arbitrary decision, you should let the uptime needs of the application determine the level of availability required.

An availability goal of 99 percent uptime is usual for most noncritical business systems. If an application must have 99 percent uptime, the application might not need to support clustering or load balancing. To achieve 99 percent uptime means that the application can have about 88 hours of downtime in an entire year, or 100 minutes of downtime a week.

To have 99.9 percent uptime is the availability goal for highly available business systems. If an application must have 99.9 percent uptime, the application must support clustering or load balancing. To achieve 99.9 percent uptime means that the application has less than 9 hours of downtime in an entire year, or, put another way, less than 10 minutes of downtime a week.

Inside Out: Clustering support alone isn't enough

Applications that support clustering are said to be cluster-aware. Microsoft SQL Server and Microsoft Exchange Server are examples of applications that are cluster-aware. Although both applications can be configured to provide high availability in the enterprise, they don't achieve high availability through cluster support alone. High-availability applications must support online backups and be tested for compatibility with Windows Server 2003. Support for online backups ensures that you don't have to take the application offline to back up critical data. Compatibility testing ensures that the software has been thoroughly evaluated for operation with Windows Server 2003.

To evaluate the real-world environment prior to deployment, you should perform integration testing on applications that will be used together. The purpose of integration testing is to ensure that disparate applications interact as expected and to uncover problem areas if they don't. During integration testing, testers should look at system performance and overall system utilization, as well as compatibility. Testing should be repeated prior to releasing system or application changes to a production environment.

You should standardize the software components needed to provide system services. The goal of standardization is to set guidelines for software components and technologies that will be used in the enterprise. Standardization accomplishes the following:

Software standardization isn't meant to limit the organization to a single specification. Over the life of the datacenter, new application versions, software components, and technologies will be introduced, and the organization can implement new standards and specifications as necessary. The key to success lies in ensuring there is a standard process for deploying software updates and new technologies. The standard process must include the following:

The following checklist summarizes the recommendations for designing and planning software for high availability: