You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by iyerbalaji <ba...@mindtree.com> on 2010/09/24 13:33:43 UTC
Scalability and Stability of JMeter’s Distributed Mode
We are planning to load test a large financial web application with JMeter.
We intend to utilize JMeter in distributed/master-slave mode with 15 load
generators (JMeter-server) + 1 Master (JMeter GUI Client).
Based on our initial experiments with JMeter in a single instance GUI mode,
we could generate between 500-700 virtual users per machine (2.8GHz, 2GB
RAM, Windows XP) against the web application. The final objective is to
generate a concurrent load of 10000 virtual users. Hence, the need to use 15
load generators.
Can JMeter successfully scale up to 15 nodes (jmeter-servers) in distributed
mode? What has been your experience in using JMeter for generating such high
load levels?
Couple of additional questions that we have includes:
- What is maximum number of load generators that can be used per Master?
- What is the impact of distributed mode on network utilization?
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2852466.html
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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
The question is "- What is the impact of distributed mode on network
utilization?"
So i am guessing he means that instead of a very very large server (mammoth
- 64 processors or something ) he is gonna run JMeter in a distributed mode
on 15 servers with a .capacity of 4 processors each. So there is gonna be
some communication between the master and the other 14 clients which will
not happen if it was one server.
So then the question would be - what would be the master and the other 14
clients communicating. Mostly what kind of request to send
and consolidating the results. This network load still would be very very
less as compared to the actual network load (download and upload) between
the clients (15 of them) and the server.
And this should certainly not be over 5% - otherwise it would defeat the
whole purpose of spreading the load on a cluster (it would choke the
network) rather than a single server.
Regards
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Fri, Sep 24, 2010 at 8:06 PM, sebb <se...@gmail.com> wrote:
> On 24 September 2010 15:02, Deepak Goel <de...@gmail.com> wrote:
> > Hey
> >
> > Namaskara~Nalama~Guten Tag
> >
> > To answer your last question it would have a max overhead of 5% on the
> > network.
>
> Huh?
>
> How do you arrive at that figure?
>
> The network traffic from server to client depends on the test plan
> listeners and the server configuration.
>
> And the percentage overhead surely depends on the current network traffic?
>
>
> > Deepak
> > --
> > Keigu
> >
> > Deepak
> > +91-9765089593
> > deicool@gmail.com
> > http://www.simtree.net
> >
> > Skype: thumsupdeicool
> > Google talk: deicool
> > Blog: http://loveandfearless.wordpress.com
> > Facebook: http://www.facebook.com/deicool
> >
> > "Contribute to the world, environment and more :
> http://www.gridrepublic.org
> > "
> >
> >
> >
> > On Fri, Sep 24, 2010 at 5:03 PM, iyerbalaji <balaji_iyer@mindtree.com
> >wrote:
> >
> >>
> >> We are planning to load test a large financial web application with
> JMeter.
> >> We intend to utilize JMeter in distributed/master-slave mode with 15
> load
> >> generators (JMeter-server) + 1 Master (JMeter GUI Client).
> >>
> >> Based on our initial experiments with JMeter in a single instance GUI
> mode,
> >> we could generate between 500-700 virtual users per machine (2.8GHz, 2GB
> >> RAM, Windows XP) against the web application. The final objective is to
> >> generate a concurrent load of 10000 virtual users. Hence, the need to
> use
> >> 15
> >> load generators.
> >>
> >> Can JMeter successfully scale up to 15 nodes (jmeter-servers) in
> >> distributed
> >> mode? What has been your experience in using JMeter for generating such
> >> high
> >> load levels?
> >>
> >> Couple of additional questions that we have includes:
> >> - What is maximum number of load generators that can be used per
> >> Master?
> >> - What is the impact of distributed mode on network utilization?
> >>
> >> --
> >> View this message in context:
> >>
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2852466.html
> >> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Shetty <sh...@gmail.com>.
Ill just add one thing , your volumes (10,000) users would probably need you
to load your results in a DB , so you could actually run everything
independently(as separate JMeter instances instead of master slave just kick
all the plans off at the same time using a cron on each PC or something) and
do away with any network issues between master/slave .
On Wed, Sep 29, 2010 at 8:57 AM, iyerbalaji <ba...@mindtree.com>wrote:
>
> Thanks everyone for your suggestions.Probably i would give a try and ask
> for
> more inputs as and when required.
> --
> View this message in context:
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p3030273.html
> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by iyerbalaji <ba...@mindtree.com>.
Thanks everyone for your suggestions.Probably i would give a try and ask for
more inputs as and when required.
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p3030273.html
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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 28 September 2010 11:29, Deepak Goel <de...@gmail.com> wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> The number of master-slave is independent of the hardware configuration.
Sorry, but that is just not the case.
The slave systems also need to be sufficiently powerful to run JMeter
without running out of resources.
The master system and its network throughput can be a limiting factor.
> You
> should be able to run any number of nodes until unless there is some
> software configuration mentioned in Jmeter which limits the number of slaves
> to be run.
There is no fixed limitation on the number of slaves that can be
defined to JMeter.
However, if there are hundreds, there might be problems with the line
length in the properties file or certainly there will be command-line
length restrictions.
Java has a maximum string length. However none of these are likely to
be the limiting factor.
> Deepak
> --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
> http://www.simtree.net
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
> Facebook: http://www.facebook.com/deicool
>
> "Contribute to the world, environment and more : http://www.gridrepublic.org
> "
>
>
>
> On Tue, Sep 28, 2010 at 3:36 PM, sebb <se...@gmail.com> wrote:
>
>> On 28 September 2010 11:01, Felix Frank <ff...@mpexnet.de> wrote:
>> > On 09/28/2010 11:57 AM, iyerbalaji wrote:
>> >>
>> >> To be specific, the configurations are:
>> >>
>> >> Load generators(agent machines) : 2.8GHz, 1GB RAM, Windows XP/2K3
>> machines
>> >> and
>>
>> 1 GB is a bit on the low side for WinXP.
>>
>> >> Master machine : 2.8 Ghz,8GB RAM,Windows XP/2K3 with a dedicated LAN
>> setup.
>>
>> 8GB is excessive for WinXP (not sure it can use it all effectively).
>>
>> >>
>> >> Can JMeter successfully scale up to 15 nodes (jmeter-servers/load
>> >> generators) with the configurations mentioned above?
>> >>
>> >> (Again,let me know if this data is not sufficient to determine the
>> number of
>> >> load generators that can
>> >> be used per Master.)
>> >
>> > I don't think anyone here can give you a definite response to that. As
>> > Sebb pointed out, it depends on other factors as well.
>> >
>> > You should probably just give it a shot.
>>
>> Agreed.
>>
>> > Personally, I have great faith in the Jmeter master. I usually run two
>> > load generators, one of them sharing a physical machine with the master
>> > process, and the master has not so far eaten up my resources.
>> > I'm not a habitual users of Listeners though, except the implied Simple
>> > Data Writer when running the master with -l <logfile>.
>> >
>> > HTH,
>> > Felix
>> >
>> > ---------------------------------------------------------------------
>> > 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
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
The number of master-slave is independent of the hardware configuration. You
should be able to run any number of nodes until unless there is some
software configuration mentioned in Jmeter which limits the number of slaves
to be run.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Tue, Sep 28, 2010 at 3:36 PM, sebb <se...@gmail.com> wrote:
> On 28 September 2010 11:01, Felix Frank <ff...@mpexnet.de> wrote:
> > On 09/28/2010 11:57 AM, iyerbalaji wrote:
> >>
> >> To be specific, the configurations are:
> >>
> >> Load generators(agent machines) : 2.8GHz, 1GB RAM, Windows XP/2K3
> machines
> >> and
>
> 1 GB is a bit on the low side for WinXP.
>
> >> Master machine : 2.8 Ghz,8GB RAM,Windows XP/2K3 with a dedicated LAN
> setup.
>
> 8GB is excessive for WinXP (not sure it can use it all effectively).
>
> >>
> >> Can JMeter successfully scale up to 15 nodes (jmeter-servers/load
> >> generators) with the configurations mentioned above?
> >>
> >> (Again,let me know if this data is not sufficient to determine the
> number of
> >> load generators that can
> >> be used per Master.)
> >
> > I don't think anyone here can give you a definite response to that. As
> > Sebb pointed out, it depends on other factors as well.
> >
> > You should probably just give it a shot.
>
> Agreed.
>
> > Personally, I have great faith in the Jmeter master. I usually run two
> > load generators, one of them sharing a physical machine with the master
> > process, and the master has not so far eaten up my resources.
> > I'm not a habitual users of Listeners though, except the implied Simple
> > Data Writer when running the master with -l <logfile>.
> >
> > HTH,
> > Felix
> >
> > ---------------------------------------------------------------------
> > 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 28 September 2010 11:01, Felix Frank <ff...@mpexnet.de> wrote:
> On 09/28/2010 11:57 AM, iyerbalaji wrote:
>>
>> To be specific, the configurations are:
>>
>> Load generators(agent machines) : 2.8GHz, 1GB RAM, Windows XP/2K3 machines
>> and
1 GB is a bit on the low side for WinXP.
>> Master machine : 2.8 Ghz,8GB RAM,Windows XP/2K3 with a dedicated LAN setup.
8GB is excessive for WinXP (not sure it can use it all effectively).
>>
>> Can JMeter successfully scale up to 15 nodes (jmeter-servers/load
>> generators) with the configurations mentioned above?
>>
>> (Again,let me know if this data is not sufficient to determine the number of
>> load generators that can
>> be used per Master.)
>
> I don't think anyone here can give you a definite response to that. As
> Sebb pointed out, it depends on other factors as well.
>
> You should probably just give it a shot.
Agreed.
> Personally, I have great faith in the Jmeter master. I usually run two
> load generators, one of them sharing a physical machine with the master
> process, and the master has not so far eaten up my resources.
> I'm not a habitual users of Listeners though, except the implied Simple
> Data Writer when running the master with -l <logfile>.
>
> HTH,
> Felix
>
> ---------------------------------------------------------------------
> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Felix Frank <ff...@mpexnet.de>.
On 09/28/2010 11:57 AM, iyerbalaji wrote:
>
> To be specific, the configurations are:
>
> Load generators(agent machines) : 2.8GHz, 1GB RAM, Windows XP/2K3 machines
> and
> Master machine : 2.8 Ghz,8GB RAM,Windows XP/2K3 with a dedicated LAN setup.
>
> Can JMeter successfully scale up to 15 nodes (jmeter-servers/load
> generators) with the configurations mentioned above?
>
> (Again,let me know if this data is not sufficient to determine the number of
> load generators that can
> be used per Master.)
I don't think anyone here can give you a definite response to that. As
Sebb pointed out, it depends on other factors as well.
You should probably just give it a shot.
Personally, I have great faith in the Jmeter master. I usually run two
load generators, one of them sharing a physical machine with the master
process, and the master has not so far eaten up my resources.
I'm not a habitual users of Listeners though, except the implied Simple
Data Writer when running the master with -l <logfile>.
HTH,
Felix
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by iyerbalaji <ba...@mindtree.com>.
To be specific, the configurations are:
Load generators(agent machines) : 2.8GHz, 1GB RAM, Windows XP/2K3 machines
and
Master machine : 2.8 Ghz,8GB RAM,Windows XP/2K3 with a dedicated LAN setup.
Can JMeter successfully scale up to 15 nodes (jmeter-servers/load
generators) with the configurations mentioned above?
(Again,let me know if this data is not sufficient to determine the number of
load generators that can
be used per Master.)
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2856450.html
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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
I doubt if there would be a constraint on the load generators if you are
looking at the accuracy. Any constraint in resources like network, server
resources (cpu, memory, disk) would only by the true representative of your
environment and hence reflect in your results.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Mon, Sep 27, 2010 at 9:19 PM, sebb <se...@gmail.com> wrote:
> On 27 September 2010 16:20, iyerbalaji <ba...@mindtree.com> wrote:
> >
> > Thanks everyone for your valuable inputs.
> > But is there a constraint/limit on the number of load generators that
> can
> > be used per Master to maintain the accuracy of the results?
>
> Of course, but the number will vary a lot depending on the test plan,
> network and other resources.
>
> > --
> > View this message in context:
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2855455.html
> > 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 27 September 2010 16:20, iyerbalaji <ba...@mindtree.com> wrote:
>
> Thanks everyone for your valuable inputs.
> But is there a constraint/limit on the number of load generators that can
> be used per Master to maintain the accuracy of the results?
Of course, but the number will vary a lot depending on the test plan,
network and other resources.
> --
> View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2855455.html
> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by iyerbalaji <ba...@mindtree.com>.
Thanks everyone for your valuable inputs.
But is there a constraint/limit on the number of load generators that can
be used per Master to maintain the accuracy of the results?
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2855455.html
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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 24 September 2010 15:52, Felix Frank <ff...@mpexnet.de> wrote:
> On 09/24/2010 04:36 PM, sebb wrote:
>> On 24 September 2010 15:02, Deepak Goel <de...@gmail.com> wrote:
>>> Hey
>>>
>>> Namaskara~Nalama~Guten Tag
>>>
>>> To answer your last question it would have a max overhead of 5% on the
>>> network.
>>
>> Huh?
>>
>> How do you arrive at that figure?
>>
>> The network traffic from server to client depends on the test plan
>> listeners and the server configuration.
>>
>> And the percentage overhead surely depends on the current network traffic?
>
> Agreed. The actual overhead should be carefully measured and (if
> necessary) extrapolated for the specific use case.
>
> If it threatens to be a problem, you can use the system properties
> mode=Statistical and num_sample_threshold=X where X is the number of
> sampler results that are aggregated into a single value that is transferred.
And use as few listeners as possible, and don't try to return the sample data.
It's easy enough to create a simple test plan using the Java Test sampler.
Or you can use the HTTP sampler with a file: URL.
This can be used to test the client-server communications without
needing to contact the real servers.
If the additional load does prove too much, then just run the test
plan using non-GUI mode on all the servers independently, and combine
the results files.
This has been discussed many times on the mailing list.
> HTH,
> Felix
>
> ---------------------------------------------------------------------
> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Felix Frank <ff...@mpexnet.de>.
On 09/24/2010 04:36 PM, sebb wrote:
> On 24 September 2010 15:02, Deepak Goel <de...@gmail.com> wrote:
>> Hey
>>
>> Namaskara~Nalama~Guten Tag
>>
>> To answer your last question it would have a max overhead of 5% on the
>> network.
>
> Huh?
>
> How do you arrive at that figure?
>
> The network traffic from server to client depends on the test plan
> listeners and the server configuration.
>
> And the percentage overhead surely depends on the current network traffic?
Agreed. The actual overhead should be carefully measured and (if
necessary) extrapolated for the specific use case.
If it threatens to be a problem, you can use the system properties
mode=Statistical and num_sample_threshold=X where X is the number of
sampler results that are aggregated into a single value that is transferred.
HTH,
Felix
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Philippe Mouawad <p....@ubik-ingenierie.com>.
Hello,
Provided you configure it correctly it scales very well.
Read this for important tips and tuning options:
-
http://www.ubik-ingenierie.com/ubikwiki/index.php?title=JMeter_Performance_and_Tuning_Tips
Regards
Philippe M.
-----
Philippe M.
http://www.ubik-ingenierie.com
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p5715651.html
Sent from the JMeter - User mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
I speak only from my experience of performance testing of over 100
applications. I am yet to come across a situation where the master-slave
traffic would be so significant that it would cause a network bottleneck and
would be greater than the client-server & server-server traffic
More so, it is simple logic & would defeat common sense if the master slave
traffic is more than the other network traffic in the system.
Anyways, as i mentioned earlier, it would be great if practitioner would
post actual scenario results. We would then be able to catch the snake from
the scruff of the neck.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Thu, Sep 30, 2010 at 12:59 PM, Felix Frank <ff...@mpexnet.de> wrote:
> Hi Deepak,
>
> while I respect your opinion and value your repeated contributions here,
> I don't see why you keep so insistent that the master-slave traffic is
> negligible.
>
> Without knowledge of the application under test, you cannot assume
> anything about the payload and thus the amount of traffic coming from
> the web servers.
> Without knowledge of the Test Plan and, specifically, the number and
> nature of Listeners used, you cannot know the amount of traffic
> necessary for the master to operate.
>
> Granted, it is most common to stress test large-ish documents (i.e.,
> bodies several orders of magnitude larger than headers) and making
> scarce use of Listeners to prevent network bottlenecks, but your blind
> insistence that it cannot ever be the dominating factor wrt. network use
> is clearly not justified.
>
> No personal offense intended.
>
> Sincerely,
> Felix
>
> On 09/30/2010 12:41 AM, Deepak Goel wrote:
> > Hey
> >
> > Namaskara~Nalama~Guten Tag
> >
> > I forgot to add, i am unsure about the server configuration, but if there
> > are multiple web servers-app servers-db servers-file servers and there
> are
> > lot of sub transactions within a transaction, there would be
> significantly
> > more traffic on the network which would make the master<=>15 client
> traffic
> > look like a needle in the haystack.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Felix Frank <ff...@mpexnet.de>.
Hi Deepak,
while I respect your opinion and value your repeated contributions here,
I don't see why you keep so insistent that the master-slave traffic is
negligible.
Without knowledge of the application under test, you cannot assume
anything about the payload and thus the amount of traffic coming from
the web servers.
Without knowledge of the Test Plan and, specifically, the number and
nature of Listeners used, you cannot know the amount of traffic
necessary for the master to operate.
Granted, it is most common to stress test large-ish documents (i.e.,
bodies several orders of magnitude larger than headers) and making
scarce use of Listeners to prevent network bottlenecks, but your blind
insistence that it cannot ever be the dominating factor wrt. network use
is clearly not justified.
No personal offense intended.
Sincerely,
Felix
On 09/30/2010 12:41 AM, Deepak Goel wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> I forgot to add, i am unsure about the server configuration, but if there
> are multiple web servers-app servers-db servers-file servers and there are
> lot of sub transactions within a transaction, there would be significantly
> more traffic on the network which would make the master<=>15 client traffic
> look like a needle in the haystack.
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
I think they are trying load testing and it is not the actual production
system. I doubt if your hypothesis is true. Anyways it is best for the
person who posted the question (and any other person who is in a similar
situation) to post their observations.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Thu, Sep 30, 2010 at 5:43 AM, sebb <se...@gmail.com> wrote:
> On 29 September 2010 23:41, Deepak Goel <de...@gmail.com> wrote:
> > Hey
> >
> > Namaskara~Nalama~Guten Tag
> >
> > I forgot to add, i am unsure about the server configuration, but if there
> > are multiple web servers-app servers-db servers-file servers and there
> are
> > lot of sub transactions within a transaction, there would be
> significantly
> > more traffic on the network which would make the master<=>15 client
> traffic
> > look like a needle in the haystack.
>
> Again, this depends on the network configuration.
>
> For a high performance system it is likely that the client-facing
> servers will be on one network, and the application and database
> servers on another network.
>
> >
> > Deepak
> > --
> > Keigu
> >
> > Deepak
> > +91-9765089593
> > deicool@gmail.com
> > http://www.simtree.net
> >
> > Skype: thumsupdeicool
> > Google talk: deicool
> > Blog: http://loveandfearless.wordpress.com
> > Facebook: http://www.facebook.com/deicool
> >
> > "Contribute to the world, environment and more :
> http://www.gridrepublic.org
> > "
> >
> >
> >
> > On Thu, Sep 30, 2010 at 12:41 AM, sebb <se...@gmail.com> wrote:
> >
> >> On 29 September 2010 19:04, Deepak Goel <de...@gmail.com> wrote:
> >> > Hey
> >> >
> >> > Namaskara~Nalama~Guten Tag
> >> >
> >> > Well, the network traffic between client and server would be for 15
> >> slaves
> >> > <=> Server. For most applications this network traffic would be
> >> > significantly higher than the Master <=>15 Slave network traffic.
> >> >
> >> > If you find a system which the case is otherwise, please let me know.
> >> That
> >> > would be a discovery which we can send it out to the Guinness Book of
> >> World
> >> > Records!
> >>
> >> Oops - I was thinking of a different posting where each slave sent to
> >> a different server.
> >>
> >> However, it is still true that the sample data that is returned to the
> >> master *can* be bigger than the response from the (web) server, even
> >> if only one (web) server is involved.
> >>
> >> If the listener is configured to return the sample data (and
> >> statistical sampling is not used), then the data sent back to the
> >> master will include the headers and the body and the additional JMeter
> >> data fields, and the overhead for serialisation. Furthermore, if the
> >> body content was gzipped, it will be returned in unzipped form.
> >>
> >> If there are multiple listeners, it may well be that the same data is
> >> sent to each of them.
> >>
> >> That is why I wrote that the data traffic depends on the test plan.
> >>
> >> > Regards
> >> > Deepak
> >> > --
> >> > Keigu
> >> >
> >> > Deepak
> >> > +91-9765089593
> >> > deicool@gmail.com
> >> > http://www.simtree.net
> >> >
> >> > Skype: thumsupdeicool
> >> > Google talk: deicool
> >> > Blog: http://loveandfearless.wordpress.com
> >> > Facebook: http://www.facebook.com/deicool
> >> >
> >> > "Contribute to the world, environment and more :
> >> http://www.gridrepublic.org
> >> > "
> >> >
> >> >
> >> >
> >> > On Wed, Sep 29, 2010 at 11:29 PM, sebb <se...@gmail.com> wrote:
> >> >
> >> >> On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
> >> >> > Hey
> >> >> >
> >> >> > Namaskara~Nalama~Guten Tag
> >> >> >
> >> >> > The network traffic from client to server depends on the
> >> >> request-response.
> >> >> > Whereas the network traffic between master and slaves is only for
> >> sending
> >> >> > what type of request to send to server and the result statistics
> >> (which
> >> >> > would be very very less)
> >> >>
> >> >> The master sends the test plan to the slave(s) at the start of the
> test
> >> >> run.
> >> >> The slaves return test samples back to the listeners.
> >> >>
> >> >> How much data is sent back to the master depends on the batching mode
> >> >> and the listener configuration.
> >> >> The data volume could be bigger than the response from the server.
> >> >>
> >> >> If one uses statistical batching, then yes, it will in general be
> less
> >> >> than the sample responses.
> >> >>
> >> >> Remember that the JMeter master node has to receive data from all the
> >> >> slaves, so the total traffic volume could still be greater than that
> >> >> from a single JMeter slave <=> server under test.
> >> >>
> >> >> > Deepak
> >> >> > --
> >> >> > Keigu
> >> >> >
> >> >> > Deepak
> >> >> > +91-9765089593
> >> >> > deicool@gmail.com
> >> >> > http://www.simtree.net
> >> >> >
> >> >> > Skype: thumsupdeicool
> >> >> > Google talk: deicool
> >> >> > Blog: http://loveandfearless.wordpress.com
> >> >> > Facebook: http://www.facebook.com/deicool
> >> >> >
> >> >> > "Contribute to the world, environment and more :
> >> >> http://www.gridrepublic.org
> >> >> > "
> >> >> >
> >> >> >
> >> >> >
> >> >> > On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com>
> wrote:
> >> >> >
> >> >> >>
> >> >> >> How do you arrive at that figure?
> >> >> >>
> >> >> >> The network traffic from server to client depends on the test plan
> >> >> >> listeners and the server configuration.
> >> >> >>
> >> >> >> And the percentage overhead surely depends on the current network
> >> >> traffic?
> >> >> >>
> >> >> >> -----
> >> >> >> The voice input and output is very good in this
> >> >> >>
> >> >> >>
> >> >>
> >>
> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
> >> >> >> Wireless PS3 Headset . It is compatible with all PS3 games.Buy
> from
> >> >> >> Reliable
> >> >> >>
> >> >> >>
> >> >>
> >>
> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
> >> >> >> Google Android PC apad Wholesalers.
> >> >> >> --
> >> >> >> View this message in context:
> >> >> >>
> >> >>
> >>
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
> >> >> >> 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
> >> >>
> >> >>
> >> >
> >>
> >> ---------------------------------------------------------------------
> >> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 29 September 2010 23:41, Deepak Goel <de...@gmail.com> wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> I forgot to add, i am unsure about the server configuration, but if there
> are multiple web servers-app servers-db servers-file servers and there are
> lot of sub transactions within a transaction, there would be significantly
> more traffic on the network which would make the master<=>15 client traffic
> look like a needle in the haystack.
Again, this depends on the network configuration.
For a high performance system it is likely that the client-facing
servers will be on one network, and the application and database
servers on another network.
>
> Deepak
> --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
> http://www.simtree.net
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
> Facebook: http://www.facebook.com/deicool
>
> "Contribute to the world, environment and more : http://www.gridrepublic.org
> "
>
>
>
> On Thu, Sep 30, 2010 at 12:41 AM, sebb <se...@gmail.com> wrote:
>
>> On 29 September 2010 19:04, Deepak Goel <de...@gmail.com> wrote:
>> > Hey
>> >
>> > Namaskara~Nalama~Guten Tag
>> >
>> > Well, the network traffic between client and server would be for 15
>> slaves
>> > <=> Server. For most applications this network traffic would be
>> > significantly higher than the Master <=>15 Slave network traffic.
>> >
>> > If you find a system which the case is otherwise, please let me know.
>> That
>> > would be a discovery which we can send it out to the Guinness Book of
>> World
>> > Records!
>>
>> Oops - I was thinking of a different posting where each slave sent to
>> a different server.
>>
>> However, it is still true that the sample data that is returned to the
>> master *can* be bigger than the response from the (web) server, even
>> if only one (web) server is involved.
>>
>> If the listener is configured to return the sample data (and
>> statistical sampling is not used), then the data sent back to the
>> master will include the headers and the body and the additional JMeter
>> data fields, and the overhead for serialisation. Furthermore, if the
>> body content was gzipped, it will be returned in unzipped form.
>>
>> If there are multiple listeners, it may well be that the same data is
>> sent to each of them.
>>
>> That is why I wrote that the data traffic depends on the test plan.
>>
>> > Regards
>> > Deepak
>> > --
>> > Keigu
>> >
>> > Deepak
>> > +91-9765089593
>> > deicool@gmail.com
>> > http://www.simtree.net
>> >
>> > Skype: thumsupdeicool
>> > Google talk: deicool
>> > Blog: http://loveandfearless.wordpress.com
>> > Facebook: http://www.facebook.com/deicool
>> >
>> > "Contribute to the world, environment and more :
>> http://www.gridrepublic.org
>> > "
>> >
>> >
>> >
>> > On Wed, Sep 29, 2010 at 11:29 PM, sebb <se...@gmail.com> wrote:
>> >
>> >> On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
>> >> > Hey
>> >> >
>> >> > Namaskara~Nalama~Guten Tag
>> >> >
>> >> > The network traffic from client to server depends on the
>> >> request-response.
>> >> > Whereas the network traffic between master and slaves is only for
>> sending
>> >> > what type of request to send to server and the result statistics
>> (which
>> >> > would be very very less)
>> >>
>> >> The master sends the test plan to the slave(s) at the start of the test
>> >> run.
>> >> The slaves return test samples back to the listeners.
>> >>
>> >> How much data is sent back to the master depends on the batching mode
>> >> and the listener configuration.
>> >> The data volume could be bigger than the response from the server.
>> >>
>> >> If one uses statistical batching, then yes, it will in general be less
>> >> than the sample responses.
>> >>
>> >> Remember that the JMeter master node has to receive data from all the
>> >> slaves, so the total traffic volume could still be greater than that
>> >> from a single JMeter slave <=> server under test.
>> >>
>> >> > Deepak
>> >> > --
>> >> > Keigu
>> >> >
>> >> > Deepak
>> >> > +91-9765089593
>> >> > deicool@gmail.com
>> >> > http://www.simtree.net
>> >> >
>> >> > Skype: thumsupdeicool
>> >> > Google talk: deicool
>> >> > Blog: http://loveandfearless.wordpress.com
>> >> > Facebook: http://www.facebook.com/deicool
>> >> >
>> >> > "Contribute to the world, environment and more :
>> >> http://www.gridrepublic.org
>> >> > "
>> >> >
>> >> >
>> >> >
>> >> > On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
>> >> >
>> >> >>
>> >> >> How do you arrive at that figure?
>> >> >>
>> >> >> The network traffic from server to client depends on the test plan
>> >> >> listeners and the server configuration.
>> >> >>
>> >> >> And the percentage overhead surely depends on the current network
>> >> traffic?
>> >> >>
>> >> >> -----
>> >> >> The voice input and output is very good in this
>> >> >>
>> >> >>
>> >>
>> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
>> >> >> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
>> >> >> Reliable
>> >> >>
>> >> >>
>> >>
>> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
>> >> >> Google Android PC apad Wholesalers.
>> >> >> --
>> >> >> View this message in context:
>> >> >>
>> >>
>> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
>> >> >> 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
>> >>
>> >>
>> >
>>
>> ---------------------------------------------------------------------
>> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
I forgot to add, i am unsure about the server configuration, but if there
are multiple web servers-app servers-db servers-file servers and there are
lot of sub transactions within a transaction, there would be significantly
more traffic on the network which would make the master<=>15 client traffic
look like a needle in the haystack.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Thu, Sep 30, 2010 at 12:41 AM, sebb <se...@gmail.com> wrote:
> On 29 September 2010 19:04, Deepak Goel <de...@gmail.com> wrote:
> > Hey
> >
> > Namaskara~Nalama~Guten Tag
> >
> > Well, the network traffic between client and server would be for 15
> slaves
> > <=> Server. For most applications this network traffic would be
> > significantly higher than the Master <=>15 Slave network traffic.
> >
> > If you find a system which the case is otherwise, please let me know.
> That
> > would be a discovery which we can send it out to the Guinness Book of
> World
> > Records!
>
> Oops - I was thinking of a different posting where each slave sent to
> a different server.
>
> However, it is still true that the sample data that is returned to the
> master *can* be bigger than the response from the (web) server, even
> if only one (web) server is involved.
>
> If the listener is configured to return the sample data (and
> statistical sampling is not used), then the data sent back to the
> master will include the headers and the body and the additional JMeter
> data fields, and the overhead for serialisation. Furthermore, if the
> body content was gzipped, it will be returned in unzipped form.
>
> If there are multiple listeners, it may well be that the same data is
> sent to each of them.
>
> That is why I wrote that the data traffic depends on the test plan.
>
> > Regards
> > Deepak
> > --
> > Keigu
> >
> > Deepak
> > +91-9765089593
> > deicool@gmail.com
> > http://www.simtree.net
> >
> > Skype: thumsupdeicool
> > Google talk: deicool
> > Blog: http://loveandfearless.wordpress.com
> > Facebook: http://www.facebook.com/deicool
> >
> > "Contribute to the world, environment and more :
> http://www.gridrepublic.org
> > "
> >
> >
> >
> > On Wed, Sep 29, 2010 at 11:29 PM, sebb <se...@gmail.com> wrote:
> >
> >> On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
> >> > Hey
> >> >
> >> > Namaskara~Nalama~Guten Tag
> >> >
> >> > The network traffic from client to server depends on the
> >> request-response.
> >> > Whereas the network traffic between master and slaves is only for
> sending
> >> > what type of request to send to server and the result statistics
> (which
> >> > would be very very less)
> >>
> >> The master sends the test plan to the slave(s) at the start of the test
> >> run.
> >> The slaves return test samples back to the listeners.
> >>
> >> How much data is sent back to the master depends on the batching mode
> >> and the listener configuration.
> >> The data volume could be bigger than the response from the server.
> >>
> >> If one uses statistical batching, then yes, it will in general be less
> >> than the sample responses.
> >>
> >> Remember that the JMeter master node has to receive data from all the
> >> slaves, so the total traffic volume could still be greater than that
> >> from a single JMeter slave <=> server under test.
> >>
> >> > Deepak
> >> > --
> >> > Keigu
> >> >
> >> > Deepak
> >> > +91-9765089593
> >> > deicool@gmail.com
> >> > http://www.simtree.net
> >> >
> >> > Skype: thumsupdeicool
> >> > Google talk: deicool
> >> > Blog: http://loveandfearless.wordpress.com
> >> > Facebook: http://www.facebook.com/deicool
> >> >
> >> > "Contribute to the world, environment and more :
> >> http://www.gridrepublic.org
> >> > "
> >> >
> >> >
> >> >
> >> > On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
> >> >
> >> >>
> >> >> How do you arrive at that figure?
> >> >>
> >> >> The network traffic from server to client depends on the test plan
> >> >> listeners and the server configuration.
> >> >>
> >> >> And the percentage overhead surely depends on the current network
> >> traffic?
> >> >>
> >> >> -----
> >> >> The voice input and output is very good in this
> >> >>
> >> >>
> >>
> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
> >> >> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
> >> >> Reliable
> >> >>
> >> >>
> >>
> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
> >> >> Google Android PC apad Wholesalers.
> >> >> --
> >> >> View this message in context:
> >> >>
> >>
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
> >> >> 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
> >>
> >>
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 29 September 2010 19:04, Deepak Goel <de...@gmail.com> wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> Well, the network traffic between client and server would be for 15 slaves
> <=> Server. For most applications this network traffic would be
> significantly higher than the Master <=>15 Slave network traffic.
>
> If you find a system which the case is otherwise, please let me know. That
> would be a discovery which we can send it out to the Guinness Book of World
> Records!
Oops - I was thinking of a different posting where each slave sent to
a different server.
However, it is still true that the sample data that is returned to the
master *can* be bigger than the response from the (web) server, even
if only one (web) server is involved.
If the listener is configured to return the sample data (and
statistical sampling is not used), then the data sent back to the
master will include the headers and the body and the additional JMeter
data fields, and the overhead for serialisation. Furthermore, if the
body content was gzipped, it will be returned in unzipped form.
If there are multiple listeners, it may well be that the same data is
sent to each of them.
That is why I wrote that the data traffic depends on the test plan.
> Regards
> Deepak
> --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
> http://www.simtree.net
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
> Facebook: http://www.facebook.com/deicool
>
> "Contribute to the world, environment and more : http://www.gridrepublic.org
> "
>
>
>
> On Wed, Sep 29, 2010 at 11:29 PM, sebb <se...@gmail.com> wrote:
>
>> On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
>> > Hey
>> >
>> > Namaskara~Nalama~Guten Tag
>> >
>> > The network traffic from client to server depends on the
>> request-response.
>> > Whereas the network traffic between master and slaves is only for sending
>> > what type of request to send to server and the result statistics (which
>> > would be very very less)
>>
>> The master sends the test plan to the slave(s) at the start of the test
>> run.
>> The slaves return test samples back to the listeners.
>>
>> How much data is sent back to the master depends on the batching mode
>> and the listener configuration.
>> The data volume could be bigger than the response from the server.
>>
>> If one uses statistical batching, then yes, it will in general be less
>> than the sample responses.
>>
>> Remember that the JMeter master node has to receive data from all the
>> slaves, so the total traffic volume could still be greater than that
>> from a single JMeter slave <=> server under test.
>>
>> > Deepak
>> > --
>> > Keigu
>> >
>> > Deepak
>> > +91-9765089593
>> > deicool@gmail.com
>> > http://www.simtree.net
>> >
>> > Skype: thumsupdeicool
>> > Google talk: deicool
>> > Blog: http://loveandfearless.wordpress.com
>> > Facebook: http://www.facebook.com/deicool
>> >
>> > "Contribute to the world, environment and more :
>> http://www.gridrepublic.org
>> > "
>> >
>> >
>> >
>> > On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
>> >
>> >>
>> >> How do you arrive at that figure?
>> >>
>> >> The network traffic from server to client depends on the test plan
>> >> listeners and the server configuration.
>> >>
>> >> And the percentage overhead surely depends on the current network
>> traffic?
>> >>
>> >> -----
>> >> The voice input and output is very good in this
>> >>
>> >>
>> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
>> >> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
>> >> Reliable
>> >>
>> >>
>> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
>> >> Google Android PC apad Wholesalers.
>> >> --
>> >> View this message in context:
>> >>
>> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
>> >> 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
>>
>>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
Well, the network traffic between client and server would be for 15 slaves
<=> Server. For most applications this network traffic would be
significantly higher than the Master <=>15 Slave network traffic.
If you find a system which the case is otherwise, please let me know. That
would be a discovery which we can send it out to the Guinness Book of World
Records!
Regards
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Wed, Sep 29, 2010 at 11:29 PM, sebb <se...@gmail.com> wrote:
> On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
> > Hey
> >
> > Namaskara~Nalama~Guten Tag
> >
> > The network traffic from client to server depends on the
> request-response.
> > Whereas the network traffic between master and slaves is only for sending
> > what type of request to send to server and the result statistics (which
> > would be very very less)
>
> The master sends the test plan to the slave(s) at the start of the test
> run.
> The slaves return test samples back to the listeners.
>
> How much data is sent back to the master depends on the batching mode
> and the listener configuration.
> The data volume could be bigger than the response from the server.
>
> If one uses statistical batching, then yes, it will in general be less
> than the sample responses.
>
> Remember that the JMeter master node has to receive data from all the
> slaves, so the total traffic volume could still be greater than that
> from a single JMeter slave <=> server under test.
>
> > Deepak
> > --
> > Keigu
> >
> > Deepak
> > +91-9765089593
> > deicool@gmail.com
> > http://www.simtree.net
> >
> > Skype: thumsupdeicool
> > Google talk: deicool
> > Blog: http://loveandfearless.wordpress.com
> > Facebook: http://www.facebook.com/deicool
> >
> > "Contribute to the world, environment and more :
> http://www.gridrepublic.org
> > "
> >
> >
> >
> > On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
> >
> >>
> >> How do you arrive at that figure?
> >>
> >> The network traffic from server to client depends on the test plan
> >> listeners and the server configuration.
> >>
> >> And the percentage overhead surely depends on the current network
> traffic?
> >>
> >> -----
> >> The voice input and output is very good in this
> >>
> >>
> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
> >> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
> >> Reliable
> >>
> >>
> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
> >> Google Android PC apad Wholesalers.
> >> --
> >> View this message in context:
> >>
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
> >> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 29 September 2010 18:40, Deepak Goel <de...@gmail.com> wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> The network traffic from client to server depends on the request-response.
> Whereas the network traffic between master and slaves is only for sending
> what type of request to send to server and the result statistics (which
> would be very very less)
The master sends the test plan to the slave(s) at the start of the test run.
The slaves return test samples back to the listeners.
How much data is sent back to the master depends on the batching mode
and the listener configuration.
The data volume could be bigger than the response from the server.
If one uses statistical batching, then yes, it will in general be less
than the sample responses.
Remember that the JMeter master node has to receive data from all the
slaves, so the total traffic volume could still be greater than that
from a single JMeter slave <=> server under test.
> Deepak
> --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
> http://www.simtree.net
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
> Facebook: http://www.facebook.com/deicool
>
> "Contribute to the world, environment and more : http://www.gridrepublic.org
> "
>
>
>
> On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
>
>>
>> How do you arrive at that figure?
>>
>> The network traffic from server to client depends on the test plan
>> listeners and the server configuration.
>>
>> And the percentage overhead surely depends on the current network traffic?
>>
>> -----
>> The voice input and output is very good in this
>>
>> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
>> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
>> Reliable
>>
>> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
>> Google Android PC apad Wholesalers.
>> --
>> View this message in context:
>> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
>> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
The network traffic from client to server depends on the request-response.
Whereas the network traffic between master and slaves is only for sending
what type of request to send to server and the result statistics (which
would be very very less)
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Wed, Sep 29, 2010 at 11:44 AM, yj2133011 <27...@qq.com> wrote:
>
> How do you arrive at that figure?
>
> The network traffic from server to client depends on the test plan
> listeners and the server configuration.
>
> And the percentage overhead surely depends on the current network traffic?
>
> -----
> The voice input and output is very good in this
>
> http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
> Wireless PS3 Headset . It is compatible with all PS3 games.Buy from
> Reliable
>
> http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
> Google Android PC apad Wholesalers.
> --
> View this message in context:
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by yj2133011 <27...@qq.com>.
How do you arrive at that figure?
The network traffic from server to client depends on the test plan
listeners and the server configuration.
And the percentage overhead surely depends on the current network traffic?
-----
The voice input and output is very good in this
http://www.tomtop.com/black-ps3-wireless-bluetooth-headset-for-playstation-3.html?aid=z
Wireless PS3 Headset . It is compatible with all PS3 games.Buy from Reliable
http://www.tomtop.com/google-android-7-notebook-3g-tablet-pc-umpc-wifi-mid-pda.html?aid=z
Google Android PC apad Wholesalers.
--
View this message in context: http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2965678.html
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: Scalability and Stability of JMeter’s Distributed Mode
Posted by sebb <se...@gmail.com>.
On 24 September 2010 15:02, Deepak Goel <de...@gmail.com> wrote:
> Hey
>
> Namaskara~Nalama~Guten Tag
>
> To answer your last question it would have a max overhead of 5% on the
> network.
Huh?
How do you arrive at that figure?
The network traffic from server to client depends on the test plan
listeners and the server configuration.
And the percentage overhead surely depends on the current network traffic?
> Deepak
> --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
> http://www.simtree.net
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
> Facebook: http://www.facebook.com/deicool
>
> "Contribute to the world, environment and more : http://www.gridrepublic.org
> "
>
>
>
> On Fri, Sep 24, 2010 at 5:03 PM, iyerbalaji <ba...@mindtree.com>wrote:
>
>>
>> We are planning to load test a large financial web application with JMeter.
>> We intend to utilize JMeter in distributed/master-slave mode with 15 load
>> generators (JMeter-server) + 1 Master (JMeter GUI Client).
>>
>> Based on our initial experiments with JMeter in a single instance GUI mode,
>> we could generate between 500-700 virtual users per machine (2.8GHz, 2GB
>> RAM, Windows XP) against the web application. The final objective is to
>> generate a concurrent load of 10000 virtual users. Hence, the need to use
>> 15
>> load generators.
>>
>> Can JMeter successfully scale up to 15 nodes (jmeter-servers) in
>> distributed
>> mode? What has been your experience in using JMeter for generating such
>> high
>> load levels?
>>
>> Couple of additional questions that we have includes:
>> - What is maximum number of load generators that can be used per
>> Master?
>> - What is the impact of distributed mode on network utilization?
>>
>> --
>> View this message in context:
>> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2852466.html
>> 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: Scalability and Stability of JMeter’s Distributed Mode
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
To answer your last question it would have a max overhead of 5% on the
network.
Deepak
--
Keigu
Deepak
+91-9765089593
deicool@gmail.com
http://www.simtree.net
Skype: thumsupdeicool
Google talk: deicool
Blog: http://loveandfearless.wordpress.com
Facebook: http://www.facebook.com/deicool
"Contribute to the world, environment and more : http://www.gridrepublic.org
"
On Fri, Sep 24, 2010 at 5:03 PM, iyerbalaji <ba...@mindtree.com>wrote:
>
> We are planning to load test a large financial web application with JMeter.
> We intend to utilize JMeter in distributed/master-slave mode with 15 load
> generators (JMeter-server) + 1 Master (JMeter GUI Client).
>
> Based on our initial experiments with JMeter in a single instance GUI mode,
> we could generate between 500-700 virtual users per machine (2.8GHz, 2GB
> RAM, Windows XP) against the web application. The final objective is to
> generate a concurrent load of 10000 virtual users. Hence, the need to use
> 15
> load generators.
>
> Can JMeter successfully scale up to 15 nodes (jmeter-servers) in
> distributed
> mode? What has been your experience in using JMeter for generating such
> high
> load levels?
>
> Couple of additional questions that we have includes:
> - What is maximum number of load generators that can be used per
> Master?
> - What is the impact of distributed mode on network utilization?
>
> --
> View this message in context:
> http://jmeter.512774.n5.nabble.com/Scalability-and-Stability-of-JMeter-s-Distributed-Mode-tp2852466p2852466.html
> 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
>
>