Software Testing - What is the measurable parameters of performance test in browser

Author

What is the measurable parameters of performance test in browser

Johnny Lee

2005-08-31, 9:57 pm

Hi, all:
I'm going to carry out a performance test on a browser, but I
haven't tried any performance test before. I was told firstly to find
out the measurable parameters of a browser. What would it be? Is that
means the normal parameters in PT such as CPU usage, PF usage? Or there
is something special for a browser?

Regards,

Johnny

Mark Ferguson

2005-09-03, 7:57 am

Johnny Lee wrote:
> Hi, all:
> I'm going to carry out a performance test on a browser, but I
> haven't tried any performance test before. I was told firstly to find
> out the measurable parameters of a browser. What would it be? Is that
> means the normal parameters in PT such as CPU usage, PF usage? Or there
> is something special for a browser?
>
> Regards,
>
> Johnny

A performance test is a test of some parameter of the SUT that is
causally related to 'how' the SUT implements its functionality.
Generally performance testing is measured in the time domain i.e. how
much time does the SUT take to perform some specified piece of
functionality, although measurements can also be taken in the space and
energy domains. Other measurements are, in some way, derivatives of the
time,space,enegy(work) tuple. Take for example CPU usage, this is
really an averaged measurement of energy/time where the energy is the
total number of CPU cycles required to complete a function of the SUT.
Page faulting is related to the space domain i.e. the smaller the
memory footprint of the SUT, the less likely it is to trigger a page
fault. Of course CPU usage and PF rates are not specifically
performance attributes of a browser, they apply to all processes
running on a CPU, which is why a bigger, faster CPU and more physical
memory are generally the first routes taken when attempting to get
improved performance.

So, browser specific performance tests must relate to how a browser
performs the functions that make it a browser rather than, say, a word
processor. I'd have a go at the 'browsing' bits, as they seem pretty
central to the concept of a browser.

So, what, essentially, happens when you browse? The browser takes a URL
and loads the object at that URL into its memory (and it might save a
copy on the local disk as well) and then renders a visual
representation of the object for the user to view and interact with. If
the object is an html page, the browser will parse the html and load
other resources referenced in the html. There may be javascript that is
executed by the browser as well. (... and lots of other things too, but
I'm not a browser designer!)

This gives me some things that I might be interested in from a
performance perspective:

- How long pages take to load and render (I may not be able to
decompose the timings and I might have to take just the complete time
from starting to load until rendering completes). I might want to do
this for both cached and uncached objects and also for both HTTP and
HTTPS traffic.

- How much additional memory is required to hold browsed objects.

- How long javascript functions take to run.

Finally, as this is supposed to be testing, you'll need test criteria
from which to determine a pass/fail result. If you don't have them
then, strictly speaking, you aren't doing testing!

Mark.

Bernard Homès

2005-09-03, 6:58 pm

Hello Johnny,

Let's see if I understand correctly : are you tasked to check the
performances of a browser, or are you goind to check performances using a
browser ?
In the second case I would suggest you buy a good (browser based)
performance tool. I would suggest eValid from Software Research
(www.soft.com/evalid). I used it and it always gave me results that checked
out with real life.
In the first case, I would suggest that you check the different aspects that
are covered in terms of performances :
- reliability
- impact on the machine (number of threads/processes, impact in term of
memory and/or CPU usage, ...)
- impact in the long run (memory leaks, ...)
- ability to display any/all data with adequate detail (graphical
performances)
- ability to be used on different platforms, and comparison of the impact of
the different OS (Linux, XP, ...) and processor types (Intel, Amd,...)
although this can be considered as configuration testing, not performances.
....
Regards
Bernard Homes
President CFTL French Software Testing Board

"Johnny Lee" <johnnyandfiona@hotmail.com> a écrit dans le message de
news:1125541991.726368.4910@z14g2000cwz.googlegroups.com...
> Hi, all:
> I'm going to carry out a performance test on a browser, but I
> haven't tried any performance test before. I was told firstly to find
> out the measurable parameters of a browser. What would it be? Is that
> means the normal parameters in PT such as CPU usage, PF usage? Or there
> is something special for a browser?
>
> Regards,
>
> Johnny
>

Sponsored Links