Comments on: Performance Creep http://www.myloadtest.com/performance-creep/ Performance Testing with a LoadRunner focus Sun, 25 Jul 2010 20:27:35 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Danny http://www.myloadtest.com/performance-creep/#comment-25397 Danny Wed, 07 Jan 2009 13:24:27 +0000 /?p=7#comment-25397 Do you have any sample of CPC exam for 2009 Do you have any sample of CPC exam for 2009

]]> By: Simon Norris http://www.myloadtest.com/performance-creep/#comment-7021 Simon Norris Wed, 23 May 2007 12:57:40 +0000 /?p=7#comment-7021 I was brought into my current role as a permanent test analyst, in place of random consultants brought in on an infrequent basis 'just to test'. Because I have been here now for three years, I have the ability to compare not just two releases, but up to 12 releases (quarterly releases for three years). As you correctly say, the 'creep' isn't huge between releases, but going back historically can show graphs that look more like Vusers ramping up than actual performance graphs. To cover the 'like for like' comparison you describe, what we do is 'overlap' performance test profiles. We refresh the work profile on a yearly basis, so every fourth release is actually made up of two performance tests. One to compare the current release with previous releases, then the second is to run the current workload seen in the production environment. While this means that every year I get twice as much work to do, it does mean that we can follow the course of the application throughout its entire life. I was brought into my current role as a permanent test analyst, in place of random consultants brought in on an infrequent basis ‘just to test’. Because I have been here now for three years, I have the ability to compare not just two releases, but up to 12 releases (quarterly releases for three years). As you correctly say, the ‘creep’ isn’t huge between releases, but going back historically can show graphs that look more like Vusers ramping up than actual performance graphs.

To cover the ‘like for like’ comparison you describe, what we do is ‘overlap’ performance test profiles. We refresh the work profile on a yearly basis, so every fourth release is actually made up of two performance tests. One to compare the current release with previous releases, then the second is to run the current workload seen in the production environment.

While this means that every year I get twice as much work to do, it does mean that we can follow the course of the application throughout its entire life.

]]>