If you do not know how to engage in manual correlation then you should not be using automated correlation. For when automated correlation fails then you will not have the skills required to use the tool to construct a proper virtual user.
Agree. I suppose there are times when automatic correlation should be used (a script with many correlations or a very quick script, maybe) but manual is the way to go, imho. The automatic correlations tend to be more difficult to read at a glance and as you say automatic correlation is very fallible.
If one does not fully understand manual correlation and the logic and process behind using it, something is lacking in fundamental knowledge. That deficiency is best addressed through working with manual correlation. One passes the "test" when they can manually correlate whatever is necessary.
Of course deadlines, time constraints, pressure from the boss, etc might dictate a novice simply roll the dice and see if it works but that's obviously not ideal and pretty risky.
manual correlation should be the first learning before writing script.
Automatic correlation only be use after getting comfortable with manual correlation so problem solving should be easy. But sales people sale the idea in reverse order.
Every once and a while you run into an app for which you can build (notice I said build - not just assume the defaults will work) a set of correlation rules that may help you to get some percentage of the correlations, but there is always manual work to be done. In these rare cases, then auto correlation is a great tool to speed up the script development. However, if you can't manually correlate, then your overall success rate is going to be very low.
IMO - learn how to read and understand the full logs (this is a key step in learning to correlate). I've assisted many Perf Engineers with correlation issues, and typically find that they are running with minimal logging and can't figure out why their correlation doesn't work. I have them turn on full logging and with a little research, you can often find the issue quickly.
That having been said, there are always going to be correlation nightmares! But having a good foundation and understanding of how it all works will get you far.
If they don't know manual correlation wouldn't that indicate a really low skillset? Manual correlation seems pretty basic to me and if someone didn't understand it they are likely inexperienced to the point of where they probably shouldn't be doing the job yet. They need more schooling. If they don't know manual correlation what other fairly basic problems won't they know how to deal with? So yeah I agree they shouldn't use the auto if they can't do the manual.
Last edited by LoadRunner421; 05-09-2013 at 10:25 AM.
Totally Agreed. If you do not have any idea of solutions why to engage with the problem. And correlation is the very basic step every Performance Test Engineer is expected to be expert at; if someone reviews your script and tells "Man it is failing because of correlation" its just like saying "Dude You are not performing your job well and this is because your do not have that level of expertise or skillset."
The only thing to say is DO THE THINGS THE WAY THEY NEED TO BE DONE, NOT THE WAY YOU CAN DO THEM.
I always recommend doing manual correlation as automated one not all time pick correct LB / RB. Learning manual correlation is first mantra to be known by performance testers for building effective test scripts. This is applicable for all testers irrespective of the tool you use for performance testing.
Completely agree. One should be familiar with the manual correlation before using automated correlation as many times the automated correlations doesn't find any correlations. I generally use automated correlation first as it requires much less time & effort & use manual correlation only when automated correlation doesn't catch correlation or shows some problem.