Performance Engineering , Scripting resources
I have a question/scenario for you but would first like to say thank you. I am a Performance Engineer for a small startup and your posts have been very enlightnening. They give me a since of comfort that I have been on the right track here for the past two years and that I am not the only one in this industry that comes up against the challenges I come up against.
Here's the question/scenario. We have a product that was primarily web based when I started and has now moved to Java client with a predictable user base but MUCH more complexity. The web testing piece was easy (BTW we use LoadRunner). Now we are testing using Java client technology. It is much harder to script against this application. The AUT is broken out into two peices (Back office functionality and Front Office) We have identified the most expensive processes with regard to DB, Middle tier impact and they are primarily located in the Front Office area. This is where I have chosen to spend the majority of my time and have identified several areas so for improvement. I am now becoming concerned about the Back Office portion but it is used in a much different (lighter) way. My concern is that something will go horribly wrong in Production and it will be percieved by management that this is so because I have not tested it as thouroughly as the Front Office piece. My logic is that 80% of user time is spent in the Front Office area among other factors such as looking at the impact that the Back Office processes have on the infrastucture. Due to the complexity of the Back Office aapplication and the time it would take to script against it using LoadRunner, I have decided to leverage the use of Winrunner scripts which a colleague of mine has spent the last three months scripting. These scripts will run the Back Office testing while Front Office LOAD is being applied.
I guess the bottom line in my situation is resources. I am the only Perfromance engineer here and would rather spend the majority of my time looking at performance than spending it scripting and maintaining scripts. An initial thought was to train the WinRunner person on LoadRunner scripting. That way he could script and maintain and I could focus on monitoring and analyzing the results and fixing the problems. Do you have any thoughts on this?
Re: Performance Engineering , Scripting resources
Thanks for the kind words!
(And for reposting)
It's all about risk and the 80/20 rule. You are correct, writing scripts is not Performance Engineering. I have been trying for a very long time to get 2 "scripter" on my team and leave them to do that work on 2 projects while I do the real engineering work on those (and probably a third), but the market fell out from under me before I got it firmly established.
You can't script everything. And not everything is worth scripting. The combination of WinRunner and LoadRunner scripts is perfect. Test this single user/process thing over here with WR while loading with LR. Elegant, yet simple - and better reproducable. Test the high risk areas and pray a low risk area doesn't surprize you in prod. <shrug>
Scott Barber, Sr. Performance Engineer