'Re: [Grinder-use] Extending Grinder to test a'

[prev in list] [next in list] [prev in thread] [next in thread] 

List:       grinder-use
Subject:    Re: [Grinder-use] Extending Grinder to test a
From:       "Serra, Guido, VF-Group" <Guido.Serra () vodafone ! com>
Date:       2010-05-18 10:07:53
Message-ID: B1241385C30A7B4DAAED209D35E27B7703755661 () VF-MBX12 ! internal ! vodafone ! com
[Download RAW message or body]

Leave it, you only need to wrap the api method of the library you are going to use.
The timing is given as you receive the answer from the method call... 

-----Original Message-----
From: Galder Zamarreno [mailto:galder@zamarreno.com] 
Sent: 18 May 2010 12:00
To: grinder-use
Subject: Re: [Grinder-use] Extending Grinder to test a custombinaryprotocolserver

The thing is that I don't want to track the overall time. I want to track how long a \
particular request takes from the moment the request is sent until is received \
without taking in account the time on the client side. Imagine this scenario:

I want to compare the speed of 3 binary protocols and their corresponding servers. \
Each binary protocol has its own client but I don't want each client influencing the \
overall timing. I only want the server time to be computed. Hence, that's why I was \
interested in the TCPProxy. Maybe I can generate a test in jython and have a TCPProxy \
in the middle (with port forwarding enabled) so that it can generate a test script \
that's independent of the client for a particular test. Does this make sense?

I'm not sure I can do this cos TCPProxy only seem to generate HTTP scripts. Am I \
correct?

Cheers,

On 18 May 2010 09:48, Serra, Guido, VF-Group <Guido.Serra@vodafone.com> wrote:
> Hi Galder,
> In grinder you have to distiguish between HTTP and NORM data.
> HTTP is an extended subset of KPIs. You'll notice it as soon as you 
> have to report on the data collected. (i.e. you have more colums in 
> the data*log, related to the HTTP layer)
> 
> If you use an external library as a client, you get only the NORM subset of data.
> Which are basically: test successfulnes, time of execution, time to execute, and \
> some others. 
> Wrapping a method of an external library, if no exceptions are raised, the test \
> will be counted as successful. If you have a whole use case, and you want to track \
> the overall timing. The way I experienced was to use the "page" test way. If you \
> look on how the TCPProxy tools creates the script, you will notice that the page's \
> testcases are rewrapped in a particular manner. 
> Basically, you create a method, and you map is as a test. This is your scenario.
> Inside you create other tests, wrapping the methods of the library.
> 
> If a sub test fails, execution stops (you SHALL not use try/catch), exception \
> reaches the main scenario scope and let it fail as well... 
> Either else, all sub tests are executed, and the overall time of the scenario will \
> be given by the time of getting out of the piece of code of that scenario method. 
> Hope this helps ;-)
> G.
> 
> 
> -----Original Message-----
> From: Galder Zamarreno [mailto:galder@zamarreno.com]
> Sent: 18 May 2010 09:30
> To: grinder-use
> Subject: Re: [Grinder-use] Extending Grinder to test a custom 
> binaryprotocoerver
> 
> Hi Guido,
> 
> Thanks for your reply but I'm not sure it answers my question. If you add a library \
> execution to your jython script such as in the case of Java Messaging, I suppose \
> any performance numbers will be influenced by this client code and the work that it \
> does. Is that correct? If so, I'm trying to avoid that. If not, could you explain a \
> bit more correctly how the performance numbers are calculated? 
> Cheers,
> 
> On 13 May 2010 18:12, Serra, Guido, VF-Group <Guido.Serra@vodafone.com> wrote:
> > Hi,
> > you need a library which handles that protocol.
> > Than you just need to instantiate it with jython.
> > Have a look at the script gallery on the website: Java Messaging
> > 
> > Regards,
> > G.
> > 
> > Galder Zamarreno <galder@zamarreno.com> ha scritto:
> > 
> > Hi,
> > 
> > I have a question about grinder. I'm interested in extending Grinder 
> > to test a server with which I talk a particular binary protocol. I do 
> > also have a java client that can talk to this server but for my use 
> > of Grinder, I'd like to test the performance of the server. I don't 
> > want these tests results to be affected by what this client does. So, 
> > I wonder what's the best place to go about it. From reading the docu, 
> > seems like I can either:
> > 
> > - Create a Jython script that uses my java client but I'd assume that 
> > any work executed by the client would be counted in the overall 
> > request/response time to the server.
> > 
> > - With a TCPProxy, record requests by the java client and then use 
> > this recorded data to test the server. Seems to me that this would do 
> > what I want which is only test the performance of the server.
> > 
> > Am I right?
> > 
> > Thanks very much in advance
> > --
> > Galder Zamarreño
> > http://galder.zamarreno.com
> > 
> > ---------------------------------------------------------------------
> > -
> > ----
> > ----
> > 
> > _______________________________________________
> > grinder-use mailing list
> > grinder-use@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/grinder-use
> > 
> > 
> > 
> > ---------------------------------------------------------------------
> > -
> > --------
> > 
> > 
> > _______________________________________________
> > grinder-use mailing list
> > grinder-use@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/grinder-use
> > 
> > 
> 
> 
> 
> --
> Galder Zamarreño
> http://galder.zamarreno.com
> 
> ----------------------------------------------------------------------
> --------
> 
> _______________________________________________
> grinder-use mailing list
> grinder-use@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/grinder-use
> 
> ----------------------------------------------------------------------
> --------
> 
> _______________________________________________
> grinder-use mailing list
> grinder-use@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/grinder-use
> 



--
Galder Zamarreño
http://galder.zamarreno.com

------------------------------------------------------------------------------

_______________________________________________
grinder-use mailing list
grinder-use@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/grinder-use

------------------------------------------------------------------------------

_______________________________________________
grinder-use mailing list
grinder-use@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/grinder-use


[prev in list] [next in list] [prev in thread] [next in thread]
Configure | About | News | Add a list | Sponsored by KoreLogic