Monitoring Performance

We have established that performance monitoring does not just begin in a vacuum. It begins once the developers and system architects have released the first version of an application and have collected initial performance measurements.

Presumably the team has also established its performance expectations for the Web application and already has a sense of whether additional optimization will be required. The initial baseline is a collection of measurements on specific performance factors. These factors combine to produce the overall responsiveness of the application on a specific hardware platform, and under a specific load.

Performance profiling involves three important steps:

• Monitoring: Monitoring includes setting up the counters and traces to collect performance data and picking a sampling period.

• Analysis: The monitoring data must be collected, analyzed for problems, and compared against the baseline data.

• Loading, or stress testing: This involves a forceful ramping up of load on the Web application to observe how the performance metrics change in response. Monitoring by itself can be a passive activity where measurements are collected and analyzed, but the system is allowed to operate without interference by the development team.

Once performance profile reports have been generated and analyzed, the technical team may choose to optimize a specific area of the application or the system. Once the optimization has been implemented, the iterative cycle of profiling and optimization begins again for as long as it takes to bring the application performance within an acceptable range.

Optimization is never really completed. It continues for as long as necessary, or for as long as the team thinks that further performance improvements are possible. Ultimately, the goal is to bring performance to a level where the users are satisfied and to have the technical team feel satisfied they have delivered an application whose performance meets everyone's expectations.

Quoted.