Application Performance Testing

W H I T E PA P E R

Application Performance Testing: From Conception to Gravestone

www.gomez.com

Performance testing of software systems is frequently discussed as an episodic activity that is accomplished once per production release as the final step in the development process for a given application. This is roughly equivalent to discussing a person’s life as existing only betweenentering high school and completing university. While this is certainly a particularly interesting period during many people’s lives, most folks would be sad to imagine that this is the only part of his or her life that mattered enough to discuss or to be remembered when they are gone. The same can be said about performance testing. In this paper, we will draw parallels between performance testing ofsoftware systems and a person’s growth, development, and eventual decline, from conception to gravestone.

Conception and PreNatal Care
Yes, you read that correctly, conception, not cradle. in much the same way that many of a person’s attributes, such as eye color and whether one’s earlobes are dangly or attached, are determined at the moment of conception, there are a huge number ofperformance attributes, positive and negative, that are established as soon as system is conceived. for example, conceiving a system that stores information in memory, even when it’s not being used, so that it will launch faster the next time it’s used is bound to have performance implications. in a manner of speaking, these are attributes of performance that are part of a system’s genetic code. lucky forus, research hospitals aren’t the only ones who can test genetic material. When applied to systems instead of people, genetic testing is relatively cheap and easy. most genetic testing of systems can be accomplished by asking performance questions during concept discussions, through peer reviews, and making use of your favorite search engine, documentation, or resident expert. often it’s not veryscientific, and some argue that these activities don’t count as “testing”, but just because the target of a test is something other than the software, doesn’t make it any less of a test. Whether or not you think it counts as testing, it’s hard to dispute that it’s far easier to achieve a well performing system when good performance is designed in from the onset, as opposed to trying to inject it inlater. it is equivalent to people who are born with a strong metabolism and a tendency toward above average musculature having an advantage in athletic endeavors. people without those advantages have to spend more time training and conditioning compared with those who have those advantages, in order to be competitive. some people even turn to illegal performance enhancing substances in an attemptto overcome their disadvantageous genes. While it is true that there are no laws against injecting performance enhancements into our systems such as there are in many sports, injecting those enhancements is still more time consuming, expensive and risky than simply ensuring that our system has the right genetic code to begin with. at this point you will want to be sure to consider how much loadthe system will need to handle not only at launch but accounting for growth.
G O M E Z — W h i t e pa p e r - a p p l i c at i o n p e r f o r m a n c e t e s t i n g 2

this analogy is applicable until the first line of code is written or the first piece of hardware is purchased. much like how an expectant mother has periodic medical exams and takes prenatal vitamins to ensure their child’sdevelopment stays on track, we must continue asking performance questions and doing our research throughout the initial design, or pre-natal, period to keep our system on track to becoming consistently well performing one. to illustrate, let’s examine the performance testing evolution of life insurance for testers (lift), inc. lift is a fictitious company, but is very closely based on a real...