Software testing is more than validating the software simply does what it should do (functional testing) we also have to make sure it does what it should do in a certain way (non-functional testing). Performance testing is one of these non-functional testing technique with a huge array of tools and techniques available to testers.
At a basic level setting up simple performance measures on the machines under test can offer an amazing insight into the performance of a solution.
The following is a very simple introduction to gathering simple performance metrics from the machines under test.
This guide is using Windows Performance monitoring and was also created on a Windows 2003 server. Other operating systems may be slightly different.
Here goes:
Log on to machine you want to gather the metrics from.
Start > Control Panel > Administrative Tools > Performance
Click on the Performance Logs and Alerts > Counter Logs
Performance window
Right click in right hand pane
Choose New Log Setting
New Log dialogue
Enter a name (this name will become the name of the log file)
Log dialogue
Choose ‘Add Counters’
Add counters window
Choose the performance object and then the relevant counters required (i.e. CPU, Memory, etc). When finished click close.
Counters summary window
Choose suitable Sample data rate interval for the testing being done.
Sampling too often may mean you have too much data to waft through, yet sampling too infrequently may mean you miss vital clues to performance. This value may have a slight impact on the performance of the machine also as it is grabbing data and writing it to a file so this is a consideration when analysing the results.
Click apply when you have finished.
Configuring the log files
Set the location of the log file. This file will contain all of the data used for analysis. Choose a log file type. I always use Excel as powerful charts and graphs can be produced off the back of the metrics gathered.
Schedule Window
Move on to the Schedule Tab and choose a suitable schedule. Starting the file can be done automatically or manually. For greater control it is often worth starting them automatically. A batch file or script could be written to control this.
Performance window with configured log
The new log will show in the right hand pane of the performance window.
Right clicking can stop/start the counter. Once the log has run it will output a file to the selected location.
Opening the created log file from the destination set up before looks like this in Excel.
More visual metrics can be produced such as graphs and charts off the back of this data.
This is a very quick and easy way for manual testers to see basic performance figures of the machines under test.
Tying these metrics in with common user actions/scenarios and gathering the metrics through automation tools is the next logical step. Ensuring that the machines and network configurations are as near to Live as possible is essential if performance testing is a formal part of your test strategy.
A performance test tool such as Visual Studio Load Test can offer all of this functionality as do many other testing tools. For those with no access to these sorts of tools this is the next best thing.