Not all applications require the same level of fault tolerance or reliability, Stoltzfus says. He suggests companies conduct a business case study on the impact of Web downtime or degradation in response time on their business. "There are very few environments that need zero tolerance and zero impact on the user," he says.

Next, companies should map their tolerance for downtime against the cost of implementing a total solution. On a graph with the Y axis representing cost and the X axis system availability, the cost curve goes up exponentially as you approach zero tolerance, Stoltzfus says.

A word of advice: Make sure you have business users looking at the cost vs. availability charts. "If you leave it to the engineers, they'll always build it bigger than you need," he says.

After this process, companies can define exactly what they mean by the terms "high availability" and "good response time." After they perform a cost/benefit analysis, companies define their systems requirements around these parameters.

Finally, Stoltzfus advises, companies must make sure they have the right engineers to build the applications and infrastructure.