curl-loader-devel

WebLoad HTTP performance numbers

[Robert I. <cor...@gm...>]

FYI.

See the link about their performance numbers.

http://www.sqaforums.com/showflat.php?Cat=0&Number=381616&an=0&page=0#Post381616

People found, that performance of Webload is below 2000 virtual clients
and other numbers being also not so impressive (we have reached 60 000
and are limited mainly by the memory of the available HW).

To say the truth - my expectations from WebLoad were much higher.

We are talking, that 30-35 K for each virtual client is too much, whereas
WebLoad is taking an order more.

-- 
Sincerely,
Robert Iakobashvili,
coroberti %x40 gmail %x2e com
...........................................................
http://curl-loader.sourceforge.net
A web testing and traffic generation tool.

Re: WebLoad HTTP performance numbers

[Robert I. <cor...@gm...>]

Michael,

On 7/9/07, Robert Iakobashvili <cor...@gm...> wrote:
>>http://www.sqaforums.com/showflat.php?Cat=0&Number=381616&an=0&page=0#Post381616
>>
>> People found, that performance of Webload is below 2000 virtual clients
>> and other numbers being also not so impressive (we have reached 60 000
>> and are limited mainly by the memory of the available HW).
>>
>> We are talking, that 30-35 K for each virtual client is too much, whereas
>> WebLoad is taking an order more.


> what about loadrunner (former mercury, now hp) ?

Please, address always to the list:
cur...@li....

For example, I do not know, but Vlad seems to be in position to
comment about Mercury LoadRunner numbers.

Sincerely,
Robert Iakobashvili,
coroberti %x40 gmail %x2e com
...........................................................
http://curl-loader.sourceforge.net
A web testing and traffic generation tool.

Re: WebLoad HTTP performance numbers

[Vlad W. <wvl...@gm...>]

On 7/10/07, Robert Iakobashvili <cor...@gm...> wrote:

> Michael,
>
> On 7/9/07, Robert Iakobashvili <cor...@gm...> wrote:
> >>
> http://www.sqaforums.com/showflat.php?Cat=0&Number=381616&an=0&page=0#Post381616
> >>
> >> People found, that performance of Webload is below 2000 virtual clients
> >> and other numbers being also not so impressive (we have reached 60 000
> >> and are limited mainly by the memory of the available HW).
> >>
> >> We are talking, that 30-35 K for each virtual client is too much,
> whereas
> >> WebLoad is taking an order more.
>
>
> > what about loadrunner (former mercury, now hp) ?
>
> Please, address always to the list:
> cur...@li....
>
> For example, I do not know, but Vlad seems to be in position to
> comment about Mercury LoadRunner numbers.
>
> Sincerely,
> Robert Iakobashvili,
> coroberti %x40 gmail %x2e com
> ...........................................................
> http://curl-loader.sourceforge.net
> A web testing and traffic generation tool.
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> curl-loader-devel mailing list
> cur...@li...
> https://lists.sourceforge.net/lists/listinfo/curl-loader-devel
>


It was long time ago and I don't remember the numbers...

Roughly, LoadRunner performace depends of the choosed protocol and virtual
client complexity.
Since LR is positioned as business process oriented tool, it is usually used
running "GUI level" protocol which builds DOM and runs javascripts. Resulted
performance is relativelylow (hundreds concurrent virtual users for single
machine) but virtual user script is quite simple and easy to maintanace.

HTTP level protocol provides thousands concurrent virtual users on single
load generator machine, but business process implementation become be very
complicated task, providing good job for LR experts.

So, LR is definitely not the best stress testing tool. However, for business
process oriented load testing it is the best that I know. At the same time,
if performance is the issue, LR can use multiple load generator machines.
Usually, users are limited rather by the license than by performance issues.

Thanks,
Vlad.

Re: WebLoad HTTP performance numbers

[Robert I. <cor...@gm...>]

On 7/10/07, Vlad W. <wvl...@gm...> wrote:
>>http://www.sqaforums.com/showflat.php?Cat=0&Number=381616&an=0&page=0#Post381616
> > >>
> > >> People found, that performance of Webload is below 2000 virtual clients
> > >> and other numbers being also not so impressive (we have reached 60 000
> > >> and are limited mainly by the memory of the available HW).
> > >>
> > >> We are talking, that 30-35 K for each virtual client is too much,
> whereas
> > >> WebLoad is taking an order more.
> >
> >
> > > what about loadrunner (former mercury, now hp) ?
> >
> > Please, address always to the list:
> > cur...@li....
> >
> > For example, I do not know, but Vlad seems to be in position to
> > comment about Mercury LoadRunner numbers.

> It was long time ago and I don't remember the numbers...

Y are kidding ...
>
> Roughly, LoadRunner performace depends of the choosed protocol and virtual
> client complexity.
> Since LR is positioned as business process oriented tool, it is usually used
> running "GUI level" protocol which builds DOM and runs javascripts. Resulted
> performance is relativelylow (hundreds concurrent virtual users for single
> machine) but virtual user script is quite simple and easy to maintanace.
>
> HTTP level protocol provides thousands concurrent virtual users on single
> load generator machine, but business process implementation become be very
> complicated task, providing good job for LR experts.
>
> So, LR is definitely not the best stress testing tool. However, for business
> process oriented load testing it is the best that I know. At the same time,
> if performance is the issue, LR can use multiple load generator machines.
> Usually, users are limited rather by the license than by performance issues.

Thank you, Vlad.

To conclude, LoadRunner is also not in business of performance and stress
load testing, but rather in business process testing.

On the high-end we will be competing with IXIA IxLoad and Spirent Avalanche,
and by adding more context analysis features will fill the gap between
the Big-HW tools and business process testing (WebLoad, LoadRunner, OpenSTA),


Apache JMeter is quite another story.
JMeter is a really great open-source community project. We may
co-operate with them
by e.g. using the same Regex definitions, using the same proxy for fetching
user behavior for our batch files, etc and where is possible, assist to them.

> Thanks,
> Vlad.


-- 
Sincerely,
Robert Iakobashvili,
coroberti %x40 gmail %x2e com
...........................................................
http://curl-loader.sourceforge.net
A web testing and traffic generation tool.

Re: WebLoad HTTP performance numbers

[Vlad W. <wvl...@gm...>]

On 7/10/07, Robert Iakobashvili <cor...@gm...> wrote:
>
> On 7/10/07, Vlad W. <wvl...@gm...> wrote:
> > > For example, I do not know, but Vlad seems to be in position to
> > > comment about Mercury LoadRunner numbers.
>
> > It was long time ago and I don't remember the numbers...
>
> Y are kidding ...


Sometimes.

To conclude, LoadRunner is also not in business of performance and stress
> load testing, but rather in business process testing.


Not exactly. Its place is "application performace testing", including
business process simulation script creating, load testing of the application
this script and multiple virual users, and analysis of obtained statistics.

The business process testing itself, i.e. functional testing, is the area of
WinRunner/QuickTest.

On the high-end we will be competing with IXIA IxLoad and Spirent Avalanche,
> and by adding more context analysis features will fill the gap between
> the Big-HW tools and business process testing (WebLoad, LoadRunner,
> OpenSTA),


That's would be good position.

Vlad