tag:blogger.com,1999:blog-6977755959349847093.post4566790937666635971..comments2024-03-01T21:53:15.921-08:00Comments on The Pith of Performance: Applying PDQ in R to Load TestingNeil Guntherhttp://www.blogger.com/profile/11441377418482735926noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-6977755959349847093.post-80511885488230445262011-05-20T08:42:37.418-07:002011-05-20T08:42:37.418-07:00re: 3 workload scenario
Check out GMantra: 2.11 &...re: 3 workload scenario<br /><br />Check out GMantra: 2.11 " No service, no queues"<br />http://www.perfdynamics.com/Manifesto/gcaprules.html#tth_sEc2.11 (a play on "No shoes, no service")<br /><br />Although the multi-class workload may be considered to better reflect reality, the show-stopper question is: Do you have the requisite service times for all these addition PDQ streams?<br /><br />A workload class is partly defined by it's individual service time at each resource (web, app, dbms). If, for whatever reason, you can't define those, then you are back to the aggregate model originally described.Neil Guntherhttps://www.blogger.com/profile/11441377418482735926noreply@blogger.comtag:blogger.com,1999:blog-6977755959349847093.post-39504072917377091832011-05-17T21:09:09.904-07:002011-05-17T21:09:09.904-07:00That's right. You can think of it as an aggreg...That's right. You can think of it as an aggregated workload with all the transactions (PDQ streams) lumped together. This is often a good starting point for building a performance model, anyway.<br /><br />Although the example does come from real measurements, it is intended to convey more about the convenience and flexibility of using PDQ within the context of R's statistical functions and programming language, than reflecting the details of any particular load test.<br /><br />It would be a straightforward matter to generalize this PDQ model to include three or more workload classes of the type mentioned in your e-commerce example.<br /><br />The different performance perspectives revealed by composite and component models are compared in Chapter 3, Section 3.7 of my Perl::PDQ book http://www.perfdynamics.com/iBook/ppa_new.htmlNeil Guntherhttps://www.blogger.com/profile/11441377418482735926noreply@blogger.comtag:blogger.com,1999:blog-6977755959349847093.post-26129556670932854542011-05-17T13:53:39.272-07:002011-05-17T13:53:39.272-07:00I am assuming that in the PDQ-R example in the blo...I am assuming that in the PDQ-R example in the blog above that all the virtual users perform the same actions?<br /><br />In a lot of the load testing that I have performed in the past we have individual types of vusers that are made into a scenario.<br /><br />Simple e-Commerce example:<br /><br />10 vusers that search for product.<br />5 vusers that browse for product.<br />3 vusers that purchase product.icrushservershttps://www.blogger.com/profile/13349512853701490365noreply@blogger.com