There are many things that can be done to help manage application performance. In the best cast, someone will review system design early in the development process. Modeling (either formal or informal) can help point out issues or potential performance problems very early on. Performance and load tests can help point up issues prior to implementation. SPE best-practices (guidelines) and worst-practices (anti-patterns) can be used make systems more responsive, efficient, and effective.

The question is: What are you doing? What works for you? Can you find time and support in your shop to review applications and develop models? Does your organization perform performance or load testing, and is that testing successful? Is your organization using SPE design guidelines or anti-patterns as a way to ensure that applications deliver optimal service for your customers?

We really want to know what you are doing. Mary Hesselgrave, who has done a lot of work with SPE offers a series of questions and challenges to start us off. If you would like to offer any ideas or suggestions or even just tell us we are crazy for even suggesting that we do SPE, please click on the link below and send us your thoughts. Whether you agree or disagree, whether you have a lot to say or just a little, we would love to hear from you. We plan to compile your responses and share them in next month’s issue. There we also plan to share some of Mary’s ideas of how she has seen SPE successfully applied.

Your thoughts and opinions are invited, regardless how brief or lengthy they may be.

Questions and challenges from Mary Hesselgrave:

Thinking about what SPE activities can add value to this particular system leads to some general SPE and performance measurement questions:

• How do you tailor an SPE process to the environment in which you work?
• What performance-related questions are important to the business?
• How long is your development cycle?
• What SPE activities add business value?
• What SPE approaches minimize business costs?
• How do you meet management expectations?
• What can you accomplish in an uncontrolled measurement environment?

• What is the problem you are trying to solve? Do we always understand the question that led to the idea of making a measurement experiment? Do we always focus on how to answer the underlying question rather than on measuring?

• How can you deal with "dirty" data that may reflect resource use from unrelated workloads or measurements taken on a different class of machine?

• Consider a mature complex distributed system with monthly releases that handles its current workload with reasonable response times and resource use. There is no time in the release cycle for classic front-end SPE activities or for performance regression testing, and there is no controlled measurement environment. What SPE activities have economic benefit?