• I can solve my performance problems by adding more machines.

The majority of performance issues are not hardware-related. For example, if a connection pool to the database is limited to 10, the number of simultaneous users accessing the database will never exceed 10, no matter how powerful the server might be.

• I can generate more load by decreasing the waiting time between pages.

The thinktime simulates the amount of time that a real user takes to pause between two web pages. Each user session consumes some server resources, but decreasing thinktime to artificially increase the hit rate does not generate a realistic load; it simply produces an inaccurate test.

• I don't need a web site load testing tool because I can react in case of problems.

When ‘live’ problems occur, you have little time to react; don't underestimate the time needed to identify and fix them. Once fixed, the only way to be sure that the same problem will not recur is to check your system with a Web Site Load Testing session before going live again.

Most problems have to do with the performance bottlenecks that arise when multiple users need to access a common resource. For example, if your site depends on a single, common database, then the speed of data access could become an issue if your database queries are not efficient. Incorrect use of multithreading in web applications can be another problem. Excessive disk access or memory leaks may also cause problems.

By identifying these problems early on, web developers, designers, testers and administrators can optimize applications, upgrade hardware, or employ caching strategies to get orders-of-magnitude performance improvements.