Ask a networking company such as SCO or Microsoft Corp. about clustering, and you'll be told that a cluster is the ability to have a secondary computer standing by to pick up the work load of a primary machine when it crashes. To a network administrator, clustering means the ability to use the computing power of multiple computer systems that are linked together.

With few exceptions, today's server clustering products do nothing more than provide system redundancy. Moreover, they require a high degree of software customization to take full advantage of network operating system clustering features.

As clustering technology becomes a standard part of operating systems, we'll start to see a much greater cohesion between applications and clustering services.

Ideally, clustering is a combination of network operating system technology with highly optimized third-party applications designed to run modularly, spreading their work loads across multiple machines.

There are a few applications of this type around in the form of sophisticated applications (usually databases) that are designed to use transaction processors to distribute their work loads across multiple machines.

For example, Oracle Corp. offers a clustered version of its software on several platforms, including Solaris and Windows NT, that can run on up to four servers.

As clustering technology becomes more widespread and is integrated as a standard part of network operating systems (as both Sun Microsystems Inc. and Microsoft are planning), we can expect to see more apps written to take advantage of these features.

Some clustering products, such as Microsoft's Cluster Server on Windows NT and SCO's ReliantHA on UnixWare, allow a basic level of clustering with many (but not all) existing applications on their operating systems, by tailoring special scripts or description files for the software. But this can be a tricky business, as demonstrated by SCO's insistence on selling ReliantHA exclusively through specialized technicians.

On the hardware side, some vendors, such as Tandem Computers Inc., offer consulting services and preconfigured clusters for specific applications as part of their customer support.

Reference Corner: Clustering technologies

The Benefits of Clustering in Shared Address Space Multiprocessors: An Applications-Driven Investigation