Choosing A Clustering Implementation

As clustering implementations vary so much from implementation to implementation, I can’t recommend a single product that will fulfill everyone’s requirements. Instead, I would like to point out some items to help you determine a method to pick the right implementation for you. Consider the following:

•  Does the vendor have an evaluation copy of the software for you to use? These are the products I would choose to look at first, as you can quickly determine if the software meets your minimum requirements. You can also determine if the software works as advertised and if the administration tools provided are easy to use.

•  Does the implementation require specific hardware or will it work using your existing hardware? If you are planning to migrate to a cluster, an implementation that functions on your existing hardware is a big plus. Generally speaking, you can use your existing server as the master and install another server to become the slave. This can significantly lower the amount of downtime for your server.

•  Does the implementation use a redundant or fault tolerant system methodology? If it uses a redundant system methodology, then, when one server fails, the other server will automatically assume the duties of the failed server. If the implementation uses a fault tolerant methodology, then, when any server fails, the protection server (as in the Octopus HA+ method) will assume the duties of the failed server.

•  Does the implementation provide load balancing similar to the Microsoft Wolfpack implementation? This type of implementation can be used to build a virtual server that outperforms any single server currently available.

•  Does the implementation provide redundancy support for your services? These services may include SQL Server, Oracle, Notes, Internet Information Server, Exchange Server, or any other applications that operate as a Windows NT service. If the implementation doesn’t support your specific service, does it provide a generic method to implement a script to initiate a rollover? This is an important question because the clustering implementation will not do you much good if it cannot detect a failure of a service that you need.

•  Finally, Does the implementation support your clients? Not every clustering implementation will support MS-DOS, Windows 3.x, Windows 95, and Windows NT clients. You should make sure that the clustering implementation you use supports your existing client base. You should consider as well whether the rollover requires your clients to manually connect to another server or not. The best option is an automatic rollover so that your clients never know that a failure occurred.

Summary

In this chapter, we examined the basics behind clustering technology as it is implemented in the real world. We took a brief glance into some of the different methods employed by various clustering implementations. This information, along with basic guidelines, can help you choose a clustering implementation that is right for your specific configuration.