You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by kingt <ca...@yahoo.com> on 2007/11/09 16:37:02 UTC
JMeter Distributed (Remote) Testing vs. CSV Data Config
Dear all,
As I need to stress a cluster of Tomcat servers, using a single JMeter seems
insufficient.
I decide to try the JMeter remote testing.
Since I have to test the web application LRU cache behavior, the input
sequence of the requests (each would access a cached web object) is
important to me.
My input parameters carried on the requests are stored in a CSV file - like
a proxy trace log. So I used the CSV data config component to drive the
workload. I wish to ask - under distributed mode, will each client machine
feed alternate requests in coordinated manner or just feed redundant request
samples at about the same instant as if each client has the same replica of
the CSV file to the cluster?
Below example is to make it clear, suppose 3 remote clients are used and the
CSV list is
A
B
C
D
E
F
....
Which of the following patterns will the request injection be:
Client0 Client1 Client2 Client3
------ ------- ------ -------
A B C D
E F G H
...
or
Client0 Client1 Client2 Client3
------ ------- ------ -------
A A A A
B B B B
C C C C
...
I am also interested the answer to the same question for other components
like cookie manager.
Thanks a lot for your help!
--
View this message in context: http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13669189
Sent from the JMeter - User mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: JMeter Distributed (Remote) Testing vs. CSV Data Config
Posted by sebb <se...@gmail.com>.
On 09/11/2007, kingt <ca...@yahoo.com> wrote:
>
>
>
> sebb-2 wrote:
> >
> > On 09/11/2007, kingt <ca...@yahoo.com> wrote:
> >>
> >> Dear all,
> >>
> >> As I need to stress a cluster of Tomcat servers, using a single JMeter
> >> seems
> >> insufficient.
> >
> > Have you tried using a single JMeter instance?
> >
> > Yes, in the previous platform which I use a 2-way SMP P3 machine to drive
> > the workload. It seems okay to drive enough load to the cluster of 8
> > machines for some applications. But I will switch to a new platform of P4
> > machines, and the load driver is of same configuration as the web app
> > servers, except more RAM. Some applications are based on SOAP and so
> > running JMeter on single machine may not be enough to stress the cluster
> > of 8 Tomcat servers. Do you have that experience? How many clients are
> > required in such a case? Thanks.
> >
> >> I decide to try the JMeter remote testing.
> >>
> >> Since I have to test the web application LRU cache behavior, the input
> >> sequence of the requests (each would access a cached web object) is
> >> important to me.
> >>
> >> My input parameters carried on the requests are stored in a CSV file -
> >> like
> >> a proxy trace log. So I used the CSV data config component to drive the
> >> workload. I wish to ask - under distributed mode, will each client
> >> machine
> >> feed alternate requests in coordinated manner or just feed redundant
> >> request
> >> samples at about the same instant as if each client has the same replica
> >> of
> >> the CSV file to the cluster?
> >
> > The CSV file needs to be present on each server host; each server will
> > read its own copy.
> >
> > I suggest you partition the requests so each server exercises its own
> > set of URLs.
> >
> > OIC. Then I feel this "distributed testing" feature is really of no big
> > difference with manually having a shell script to start separate jmeter
> > instances on different machines. But it has an advantage - merge of
> > separate results. But will it give me the total/aggregated throughput and
> > average latency, and how (e.g. what listeners and controls I should add
> > but not affecting scalability)? Thanks.
> >
For minimum overhead, don't configure any Listeners in the test, and
use non-GUI client with the -l flag to start a simple listener; use
CSV output.
Perhaps add a Summariser (see jmeter.properties).
The CSV file can be read back in after the test has completed.
If this still overloads the client, the run separate non-GUI tests and
merge the output files - see:
http://wiki.apache.org/jakarta-jmeter/JtlFiles
> >> Below example is to make it clear, suppose 3 remote clients are used and
> >> the
> >> CSV list is
> >> A
> >> B
> >> C
> >> D
> >> E
> >> F
> >> ....
> >>
> >> Which of the following patterns will the request injection be:
> >> Client0 Client1 Client2 Client3
> >> ------ ------- ------ -------
> >> A B C D
> >> E F G H
> >> ...
> >
> > No
> >
> >>
> >> or
> >>
> >> Client0 Client1 Client2 Client3
> >> ------ ------- ------ -------
> >> A A A A
> >> B B B B
> >> C C C C
> >> ...
> >
> > Yes
> >
> >>
> >> I am also interested the answer to the same question for other components
> >> like cookie manager.
> >
> > Each server runs independently.
> >
> > The client JMeter collects and merges sample results from each server,
> > but there is no other co-operation between JMeter instances.
> >
> > It would not make sense for the cookie manager to be shared between
> > JMeter instances - indeed cookies are not even shared between threads
> > in the same instance, because threads represent separate users.
> >
> >
> >
> >> Thanks a lot for your help!
> >>
> >> --
> >> View this message in context:
> >> http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13669189
> >> Sent from the JMeter - User mailing list archive at Nabble.com.
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> >> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >>
> >>
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
> >
>
> --
> View this message in context: http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13671138
> Sent from the JMeter - User mailing list archive at Nabble.com.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: JMeter Distributed (Remote) Testing vs. CSV Data Config
Posted by kingt <ca...@yahoo.com>.
sebb-2 wrote:
>
> On 09/11/2007, kingt <ca...@yahoo.com> wrote:
>>
>> Dear all,
>>
>> As I need to stress a cluster of Tomcat servers, using a single JMeter
>> seems
>> insufficient.
>
> Have you tried using a single JMeter instance?
>
> Yes, in the previous platform which I use a 2-way SMP P3 machine to drive
> the workload. It seems okay to drive enough load to the cluster of 8
> machines for some applications. But I will switch to a new platform of P4
> machines, and the load driver is of same configuration as the web app
> servers, except more RAM. Some applications are based on SOAP and so
> running JMeter on single machine may not be enough to stress the cluster
> of 8 Tomcat servers. Do you have that experience? How many clients are
> required in such a case? Thanks.
>
>> I decide to try the JMeter remote testing.
>>
>> Since I have to test the web application LRU cache behavior, the input
>> sequence of the requests (each would access a cached web object) is
>> important to me.
>>
>> My input parameters carried on the requests are stored in a CSV file -
>> like
>> a proxy trace log. So I used the CSV data config component to drive the
>> workload. I wish to ask - under distributed mode, will each client
>> machine
>> feed alternate requests in coordinated manner or just feed redundant
>> request
>> samples at about the same instant as if each client has the same replica
>> of
>> the CSV file to the cluster?
>
> The CSV file needs to be present on each server host; each server will
> read its own copy.
>
> I suggest you partition the requests so each server exercises its own
> set of URLs.
>
> OIC. Then I feel this "distributed testing" feature is really of no big
> difference with manually having a shell script to start separate jmeter
> instances on different machines. But it has an advantage - merge of
> separate results. But will it give me the total/aggregated throughput and
> average latency, and how (e.g. what listeners and controls I should add
> but not affecting scalability)? Thanks.
>
>> Below example is to make it clear, suppose 3 remote clients are used and
>> the
>> CSV list is
>> A
>> B
>> C
>> D
>> E
>> F
>> ....
>>
>> Which of the following patterns will the request injection be:
>> Client0 Client1 Client2 Client3
>> ------ ------- ------ -------
>> A B C D
>> E F G H
>> ...
>
> No
>
>>
>> or
>>
>> Client0 Client1 Client2 Client3
>> ------ ------- ------ -------
>> A A A A
>> B B B B
>> C C C C
>> ...
>
> Yes
>
>>
>> I am also interested the answer to the same question for other components
>> like cookie manager.
>
> Each server runs independently.
>
> The client JMeter collects and merges sample results from each server,
> but there is no other co-operation between JMeter instances.
>
> It would not make sense for the cookie manager to be shared between
> JMeter instances - indeed cookies are not even shared between threads
> in the same instance, because threads represent separate users.
>
>
>
>> Thanks a lot for your help!
>>
>> --
>> View this message in context:
>> http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13669189
>> Sent from the JMeter - User mailing list archive at Nabble.com.
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
>> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
>
--
View this message in context: http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13671138
Sent from the JMeter - User mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: JMeter Distributed (Remote) Testing vs. CSV Data Config
Posted by sebb <se...@gmail.com>.
On 09/11/2007, kingt <ca...@yahoo.com> wrote:
>
> Dear all,
>
> As I need to stress a cluster of Tomcat servers, using a single JMeter seems
> insufficient.
Have you tried using a single JMeter instance?
> I decide to try the JMeter remote testing.
>
> Since I have to test the web application LRU cache behavior, the input
> sequence of the requests (each would access a cached web object) is
> important to me.
>
> My input parameters carried on the requests are stored in a CSV file - like
> a proxy trace log. So I used the CSV data config component to drive the
> workload. I wish to ask - under distributed mode, will each client machine
> feed alternate requests in coordinated manner or just feed redundant request
> samples at about the same instant as if each client has the same replica of
> the CSV file to the cluster?
The CSV file needs to be present on each server host; each server will
read its own copy.
I suggest you partition the requests so each server exercises its own
set of URLs.
> Below example is to make it clear, suppose 3 remote clients are used and the
> CSV list is
> A
> B
> C
> D
> E
> F
> ....
>
> Which of the following patterns will the request injection be:
> Client0 Client1 Client2 Client3
> ------ ------- ------ -------
> A B C D
> E F G H
> ...
No
>
> or
>
> Client0 Client1 Client2 Client3
> ------ ------- ------ -------
> A A A A
> B B B B
> C C C C
> ...
Yes
>
> I am also interested the answer to the same question for other components
> like cookie manager.
Each server runs independently.
The client JMeter collects and merges sample results from each server,
but there is no other co-operation between JMeter instances.
It would not make sense for the cookie manager to be shared between
JMeter instances - indeed cookies are not even shared between threads
in the same instance, because threads represent separate users.
> Thanks a lot for your help!
>
> --
> View this message in context: http://www.nabble.com/JMeter-Distributed-%28Remote%29-Testing-vs.-CSV-Data-Config-tf4778362.html#a13669189
> Sent from the JMeter - User mailing list archive at Nabble.com.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org