Software Performance Testing

On the website of testers recently appeared an article that describes one of the approaches to software testing. I believe that it is the most correct and developers need to take it into service when testing their own products.

In particular, the moments of preparation for testing, then, in fact, testing, and analysis of the results are considered.

Along the way, I want to ask a question to developers who read Habr: do you test your software products for performance? What algorithm for this use? Tools?
')

Quote:

1. The first tests should preferably be carried out on the lowest load profile, since the behavior of the system is still unclear, there may be serious obstacles to normal / expected performance.
   Usually, an increase in the load, expressed in an increase in the intensity of operations (increase in the number of users), is practiced in order to obtain the dependences of response times on various loads.
   Simultaneously with the test, it is necessary to remove the performance metrics of the server hardware, since the testing of the Application is usually done relative to the hardware configurations (number of CPU x Memory). The most important of these are:
   
   1. Percentages for processors:
      
      1. CPU user use of processors for the application,
         CPU wio waiting i / o processors,
         CPU idle "simple" processor
         CPU queues
         Memory usage
         Queuing on wheels
         At the same time, the disk subsystem and the network should not be “bottlenecks” since in this case it will be unclear how effective the application is in terms of performance.
         
         PS source and full text: software-testing.ru