Fault Tolerance

Fault Tolerance is related to Clustering whether you are planning a high availability or load balanced implementation.  Fault tolerance has two components - Hardware and Software.

The hardware vendors who sell fault tolerant hardware would like you to think that that is all you need to think about.  It is a very big part of the solution to application crashes.  Fault tolerant hardware has redundant power supplies, CPU's, RAM, video cards, network cards, RAID adapters, etc - all plugged into a dual backplane system where almost any hardware fault has a backup.  Vendors that sell these products spend huge amounts on research and developement to make all of these pieces work together.  What you end up with is a server that costs about 5-10 times as much as an equivalent (in performance) non fault tolerant server.  If your application is a database that runs your world wide accounting database or your e-commerce site, it is probably OK to spend that much on the one hardware piece that you cannot have fail.  On all of these systems, the operating system is a limited version with special patches to minimise the chance of it crashing.  About the only software that companies run on such a platforms is Oracle which is also customized to run on that hardware to minimise the chance of it crashing.  Pretty good for your your Oracle database but not much use for any other application.

This is where software steps in to take over for hardware in the fault tolerence market.  Clustering software is one part of this and is talked about here .