Presenter: Andre B. Bondi
Building load testing scripts that generate a wide variety of sequences of web page visits can be arduous and error prone. The code can be difficult to modify and maintain. We use compiler design methods and techniques from discrete event simulation to transform a customer behavior graph model of transitions from web page to web page to randomly generate valid, representative sequences of page visits during a load test. A simple transformation of a Markov chain model allows one to predict the relative frequencies with which pages will be hit during a load test before the test is actually run. We describe a number of advantages of our method. These include the ability to build a representative workload model, the ability to do what-if analysis and experiment with different page transition patterns without altering the program logic, and ease of extensibility. Our method has been successfully used to conduct load tests on different websites at a corporation.
About the Presenter
Andre B. Bondi is an independent consultant and the founder of Software Performance and Scalability Consulting LLC. In 2016, he was a visiting professor at the University of L’Aquila and received the Computer Measurement Group’s A. A. Michelson Award. Dr. Bondi worked for prolonged periods at Siemens Corp., Corporate Technologies and at AT&T (Bell) Labs. He has held senior performance positions at two startups. His book, Foundations of Software and Systems Performance Engineering, was published by Addison-Wesley in 2014. He has worked on performance issues in many domains. He taught performance, simulation, operating systems, and architecture at UCSB for three years. He holds a Ph.D. in computer science from Purdue University, and an M.Sc. in statistics from University College London. He has nine US patents.