Four Approaches to Building Highly Available Systems Using Redundancy

To build high availability into applications and OS services, system architects can choose from four different approaches to adding redundancy: distributed services and applications, fault-tolerant servers, server farms, or server clusters.

Distributed services and applications. Sometimes Windows system services and other applications can be designed to run across multiple servers (distributed) and thereby provide redundancy and fault tolerance without the need for any OS support or special hardware considerations. Examples of distributed services and applications include Active Directory (AD), the Domain Name Service (DNS), and the combination of the Windows Distributed File System (DFS) and File Replication Service (FRS). This distributed approach works best when updates to shared data are relatively infrequent, and usually works well even when the servers are located in sites connected by a limited-bandwidth WAN. Some vendors, such as Oracle, even support high-availability transactional databases using the distributed approach. However, for the majority of applications, this approach is either not optimal or the applications were never built to work that way.