Better Software - April 2009 - (Page 35) ADVERTORIAL Performance Testing: No one knows it better than Performance Testing is one of the many tests that are often rushed or mis-scoped to accommodate the true traffic an application must endure. TCT Computing Group is a leading performance and functional test consulting and training organization. With over 15 years as a Mercury/HP partner, as well as a Borland® and AutomatedQA reseller, TCT has extensive experience in the best practices in testing. Here are a few mistakes that we see people make and how to fix them. If you want more information on how to best use automation testing tools or best practices on testing in general, join our upcoming FREE TRAINING SNEAK PEEK or schedule your free consultation by going to www.TCTcomputing.com. FREE JUNE WEBINAR: June 4, 2009 at 1 PM EST Sign up now for our Performance Testing Done Right. Visit www.TCTcomputing.com/ webinars/performance to register! Can't make this time? Go online to view previous webinars or see future topics and dates. Visit www.TCTcomputing.com/webinars 10 TOP 1 Incorrect Load Levels 6 Poor Transaction Naming Convention Automated Performance Testing Mistakes Fix: For easier analysis, maintainability, future growth, and possible server consolidation, provide a clear naming convention. Performance Tester Stands Alone 7 Fix: Let the experts do what they do best. Involve all members of the team: the DBA, Web Server Expert, Application Expert, Developer, etc. Executing a performance test is best done as a team. Not Validating the Execution Transaction Levels Fix: Improve and focus communications by conducting a one-on-one interview to determine expected “real world” load levels. If the application is already in production, obtain live usage data. Performance Tester Identifying Key Business Processes 8 Fix: Validating transaction levels against the test plan after executing a full load test ensures the test presented the desired load to the application. 2 9 Server Experts Become Bouncy Fix: Conduct a one-on-one interview to identify key business processes: mission critical, heavy throughput, dynamic content, and any other business process that causes concern. Fix: If possible, don't bounce the server between test executions. Bouncing the server prior to every test causes the need to rebuild the memory and cache. Memory leaks and other reliability and availability issues will be more easily identified if the server is left untouched. on Results Reporting 3 Too Many Business Processes Fix: Concentrate on a few, perhaps 6-8, key business processes for performance testing - this is not a functional test. Altering Load Test Objectives 10 Over-technical 4 Fix: Develop a complete custom plan with goal load levels, business process steps, transaction names, and acceptable response times. Analyze any missed objectives - stay consistent with the original goals unless there is a compelling reason for change. Late Focus on Performance Testing Fix: Performance testing creates mountains of data that are usually very interesting to technical experts. Ultimately, however, the owners of the application are looking for key performance indicators to make a sound business decision for their application go-live. Create a one-minute overview for managers that concisely conveys the critical data related to application performance. 5 Fix: Engage the performance testing team as early as possible in the development life cycle to allow time between executions for a good analysis of the results from the performance tester. Detecting and correcting performance issues early reduces the repair cost. Having issues with your performance testing? Let the experts of TCT help you with your next performance test. TCT has a refined discipline and methodology with a proven track record. TCT COMPUTING GROUP • 410.893.5800 or toll free at 866.828.6372 • www.TCTcomputing.com http://www.TCTcomputing.com http://www.TCTcomputing.com/webinars/performance http://www.TCTcomputing.com http://www.TCTcomputing.com/webinars http://www.TCTcomputing.com
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.