performance questions

[Charlie]

Hi folks,
I am working on the Web application testing these days, and I have
some questions need to clear.

Basically the application that I am testing is using the
Apache/cgi-bin, the back end database is mysql. From user point of
view, it is a website where you can place several "business" Orders.

The way that I am doing the testing is also by using the
Apache/cgi-bin( a seperate Apache running on different platform),
basides, I have used a list of modules get from www.cpan.org, such as
WWW.Mechanize, Threads, etc. I have created my perl file to simulate
realtime users' interaction, filling the text area, picking the radio
box, etc, till at last to push the submit button to have the order
submitted.

What I really interested is once the final submition button is pushed,
when the confirmation page is desplayed to the uses, (in
Perl/Mechanize, when "mech->submit()" is finished), while there are
tens, houdreds users doing the same thing concurrently.

So far I have created some testcases to test two kinds of orders. The
first one takes about 20 mi-seconds for that period of time, while the
second one takes about 20,000 mi-seconds. To create those two kinds of
orders, it may needs some extra informations for the second one than
the first one.

So based on the above description, can I necessarily say that the
performance of the second one is not as good as the first one ???

Do you see any overhead on this test module? the time consumed by the
Apache, perl/cgi-bin, not consumed by the real transaction ?


thanks a lot for any tips



CJ

 

[gnari]

[application testing]

So far I have created some testcases to test two kinds of orders. The
first one takes about 20 mi-seconds for that period of time, while the
second one takes about 20,000 mi-seconds. To create those two kinds of
orders, it may needs some extra informations for the second one than
the first one.

So based on the above description, can I necessarily say that the
performance of the second one is not as good as the first one ???

I am not sure I understand what you are asking.

whether 20 seconds is too long depends on:
a) how much processing has to take place
b) how much processing the user imagines has to take place

what you should be concerned about is how does your
system scale with load. does the 20 sec request take 2 secs under
low load? in that case you are suffering from a bottleneck

also, you should make sure that the same request sequences return
the same results under high and inder low load

also, you should find out if there is some load shreshold where
your application fails miserably, and figure out if that
level is acceptable, and make plans to monitor when you are
nearing that shreshold.

gnari