You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Ravindra Gupta <ra...@gmail.com> on 2011/01/07 03:26:35 UTC
Single Jmeter instance produces better throughput than multiple
Hi All,
I am facing a problem where if i run a single jmeter instance it produces
much better throughput than when i am running 2 jmeter instance.
Details about the test
Script is very simple, it has just one sampler which post xml request to app
server. Each thread posts one xml request and finishes.
In one test,
At the same time I simulated 50 Threads with no 0 rampup time. Lets say it
produces x throughput.
In another test
I simulated 2 jmeter instance (not in master slave) from multiple machines
with 25 thread (0 rampup) on each instance. The combined throughput I am
getting in this model is much lower than in x.
What could be the case of such a behaviour.
Ideally the combined throughput should be equivalent to x.
All the jmeter clients are server class machines with CentOS. Jmeter is
running in non GUI mode.
wht could be the cause of this behaviour?
Thanks in advace,
Ravindra
Re: Single Jmeter instance produces better throughput than multiple
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
You need to give actual output details. Looks like the server and the
network has to do more work in serving 2 client instances than one!
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, Jan 7, 2011 at 7:56 AM, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi All,
>
> I am facing a problem where if i run a single jmeter instance it produces
> much better throughput than when i am running 2 jmeter instance.
>
> Details about the test
>
> Script is very simple, it has just one sampler which post xml request to
> app
> server. Each thread posts one xml request and finishes.
>
> In one test,
> At the same time I simulated 50 Threads with no 0 rampup time. Lets say it
> produces x throughput.
>
> In another test
> I simulated 2 jmeter instance (not in master slave) from multiple machines
> with 25 thread (0 rampup) on each instance. The combined throughput I am
> getting in this model is much lower than in x.
> What could be the case of such a behaviour.
>
> Ideally the combined throughput should be equivalent to x.
>
> All the jmeter clients are server class machines with CentOS. Jmeter is
> running in non GUI mode.
> wht could be the cause of this behaviour?
>
> Thanks in advace,
> Ravindra
>
Re: Single Jmeter instance produces better throughput than multiple
Posted by sebb <se...@gmail.com>.
On 7 January 2011 04:23, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi Sebb,
>
> I am using sampler called HTTP Request HTTPClient. Here we just post an xml
> file with the request.
OK.
> If I run 50 thread, throughput that I am getting is ~585 req/sec.
> But if I run two jmeter instances, its throughput as ~236 req/sec and 236
> req/sec repectively. These numbers. may differ but slightly.
Which means the total throughput is 472 req/sec.
> Using Jmerter version 2.3.2. test is running in a separate subnet, so tthere
> is no problem of network latency
Current version of JMeter ia 2.4 - 2.3.2 is quite old.
However this probably won't affect timings much unless you are using
certain Listeners, e.g. Aggregate.
Remove all Listeners except perhaps a Summary Listener.
> The interesting thing is that if I run only one Jmeter instance with 25
> threads , throughput is ~620 req/sec.
This suggests that 50 threads may be starting to stress the server.
What happens if you run two such on the same host?
How does this differ from two on different hosts?
>
> Hi Deepak,
>
> I can see from Jmeter.log file that all the threads gets activated , there
> are no failures. I think its able to make 50 connections because I am
> getting responses correctly.
>
> Two Jmeter instance with 25 threads each
> Avg request - 105 ms
> Min - 9 ms
> Max - 1305 req/sec
Huh?
Did you mean: 1305 ms?
>
> 50 threads -single jmeter instance
> Avg req - 82 ms
> Min - 9 ms
> Max - 1259 ms
>
> I have run multiple tests resulting in similar kind of data.
And what are the comparative results for 2 * 25 threads on same host
and different host?
Also worth trying smaller numbers of threads, e.g. 2 * 5 threads and 1
* 10 threads.
> Thanks,
> Ravindra
>
>
> On Thu, Jan 6, 2011 at 7:05 PM, sebb <se...@gmail.com> wrote:
>
>> On 7 January 2011 02:26, Ravindra Gupta <ra...@gmail.com> wrote:
>> > Hi All,
>> >
>> > I am facing a problem where if i run a single jmeter instance it produces
>> > much better throughput than when i am running 2 jmeter instance.
>> >
>> > Details about the test
>> >
>> > Script is very simple, it has just one sampler which post xml request to
>> app
>> > server. Each thread posts one xml request and finishes.
>> >
>> > In one test,
>> > At the same time I simulated 50 Threads with no 0 rampup time. Lets say
>> it
>> > produces x throughput.
>> >
>> > In another test
>> > I simulated 2 jmeter instance (not in master slave) from multiple
>> machines
>> > with 25 thread (0 rampup) on each instance. The combined throughput I am
>> > getting in this model is much lower than in x.
>> > What could be the case of such a behaviour.
>> >
>> > Ideally the combined throughput should be equivalent to x.
>> >
>> > All the jmeter clients are server class machines with CentOS. Jmeter is
>> > running in non GUI mode.
>> > wht could be the cause of this behaviour?
>>
>> It's very unlikely to be a JMeter problem, so you need to do some more
>> tests to try and establish what causes the difference when using 2
>> test hosts.
>>
>> What throughput do you get if you run the 50 thread test on the second
>> host?
>> Do you get similar throughput to a 50 thread run on the original host?
>>
>> What about running two 25 thread tests on the same host?
>> Do you get similar thoughput to a single test?
>>
>> What sampler are you using to post the XML?
>>
>> > Thanks in advace,
>> > Ravindra
>> >
>>
>> ---------------------------------------------------------------------
>> 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: Single Jmeter instance produces better throughput than multiple
Posted by Deepak Shetty <sh...@gmail.com>.
Hi
>I can see from Jmeter.log file that all the threads gets activated
Which doesnt mean that the server sees that many . if you could check on
your server the total number of concurrent connection you saw in either
case. if your average times are increasing as seen on the server then I
would guess your server is seeing a higher load in the second case.
regards
deepak
On Thu, Jan 6, 2011 at 8:23 PM, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi Sebb,
>
> I am using sampler called HTTP Request HTTPClient. Here we just post an xml
> file with the request.
> If I run 50 thread, throughput that I am getting is ~585 req/sec.
> But if I run two jmeter instances, its throughput as ~236 req/sec and 236
> req/sec repectively. These numbers. may differ but slightly.
>
> Using Jmerter version 2.3.2. test is running in a separate subnet, so
> tthere
> is no problem of network latency
>
> The interesting thing is that if I run only one Jmeter instance with 25
> threads , throughput is ~620 req/sec.
>
>
> Hi Deepak,
>
> I can see from Jmeter.log file that all the threads gets activated , there
> are no failures. I think its able to make 50 connections because I am
> getting responses correctly.
>
> Two Jmeter instance with 25 threads each
> Avg request - 105 ms
> Min - 9 ms
> Max - 1305 req/sec
>
> 50 threads -single jmeter instance
> Avg req - 82 ms
> Min - 9 ms
> Max - 1259 ms
>
> I have run multiple tests resulting in similar kind of data.
>
> Thanks,
> Ravindra
>
>
> On Thu, Jan 6, 2011 at 7:05 PM, sebb <se...@gmail.com> wrote:
>
> > On 7 January 2011 02:26, Ravindra Gupta <ra...@gmail.com> wrote:
> > > Hi All,
> > >
> > > I am facing a problem where if i run a single jmeter instance it
> produces
> > > much better throughput than when i am running 2 jmeter instance.
> > >
> > > Details about the test
> > >
> > > Script is very simple, it has just one sampler which post xml request
> to
> > app
> > > server. Each thread posts one xml request and finishes.
> > >
> > > In one test,
> > > At the same time I simulated 50 Threads with no 0 rampup time. Lets say
> > it
> > > produces x throughput.
> > >
> > > In another test
> > > I simulated 2 jmeter instance (not in master slave) from multiple
> > machines
> > > with 25 thread (0 rampup) on each instance. The combined throughput I
> am
> > > getting in this model is much lower than in x.
> > > What could be the case of such a behaviour.
> > >
> > > Ideally the combined throughput should be equivalent to x.
> > >
> > > All the jmeter clients are server class machines with CentOS. Jmeter is
> > > running in non GUI mode.
> > > wht could be the cause of this behaviour?
> >
> > It's very unlikely to be a JMeter problem, so you need to do some more
> > tests to try and establish what causes the difference when using 2
> > test hosts.
> >
> > What throughput do you get if you run the 50 thread test on the second
> > host?
> > Do you get similar throughput to a 50 thread run on the original host?
> >
> > What about running two 25 thread tests on the same host?
> > Do you get similar thoughput to a single test?
> >
> > What sampler are you using to post the XML?
> >
> > > Thanks in advace,
> > > Ravindra
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
>
Re: Single Jmeter instance produces better throughput than multiple
Posted by Ravindra Gupta <ra...@gmail.com>.
Hi Sebb,
I am using sampler called HTTP Request HTTPClient. Here we just post an xml
file with the request.
If I run 50 thread, throughput that I am getting is ~585 req/sec.
But if I run two jmeter instances, its throughput as ~236 req/sec and 236
req/sec repectively. These numbers. may differ but slightly.
Using Jmerter version 2.3.2. test is running in a separate subnet, so tthere
is no problem of network latency
The interesting thing is that if I run only one Jmeter instance with 25
threads , throughput is ~620 req/sec.
Hi Deepak,
I can see from Jmeter.log file that all the threads gets activated , there
are no failures. I think its able to make 50 connections because I am
getting responses correctly.
Two Jmeter instance with 25 threads each
Avg request - 105 ms
Min - 9 ms
Max - 1305 req/sec
50 threads -single jmeter instance
Avg req - 82 ms
Min - 9 ms
Max - 1259 ms
I have run multiple tests resulting in similar kind of data.
Thanks,
Ravindra
On Thu, Jan 6, 2011 at 7:05 PM, sebb <se...@gmail.com> wrote:
> On 7 January 2011 02:26, Ravindra Gupta <ra...@gmail.com> wrote:
> > Hi All,
> >
> > I am facing a problem where if i run a single jmeter instance it produces
> > much better throughput than when i am running 2 jmeter instance.
> >
> > Details about the test
> >
> > Script is very simple, it has just one sampler which post xml request to
> app
> > server. Each thread posts one xml request and finishes.
> >
> > In one test,
> > At the same time I simulated 50 Threads with no 0 rampup time. Lets say
> it
> > produces x throughput.
> >
> > In another test
> > I simulated 2 jmeter instance (not in master slave) from multiple
> machines
> > with 25 thread (0 rampup) on each instance. The combined throughput I am
> > getting in this model is much lower than in x.
> > What could be the case of such a behaviour.
> >
> > Ideally the combined throughput should be equivalent to x.
> >
> > All the jmeter clients are server class machines with CentOS. Jmeter is
> > running in non GUI mode.
> > wht could be the cause of this behaviour?
>
> It's very unlikely to be a JMeter problem, so you need to do some more
> tests to try and establish what causes the difference when using 2
> test hosts.
>
> What throughput do you get if you run the 50 thread test on the second
> host?
> Do you get similar throughput to a 50 thread run on the original host?
>
> What about running two 25 thread tests on the same host?
> Do you get similar thoughput to a single test?
>
> What sampler are you using to post the XML?
>
> > Thanks in advace,
> > Ravindra
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
Re: Single Jmeter instance produces better throughput than multiple
Posted by sebb <se...@gmail.com>.
On 7 January 2011 02:26, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi All,
>
> I am facing a problem where if i run a single jmeter instance it produces
> much better throughput than when i am running 2 jmeter instance.
>
> Details about the test
>
> Script is very simple, it has just one sampler which post xml request to app
> server. Each thread posts one xml request and finishes.
>
> In one test,
> At the same time I simulated 50 Threads with no 0 rampup time. Lets say it
> produces x throughput.
>
> In another test
> I simulated 2 jmeter instance (not in master slave) from multiple machines
> with 25 thread (0 rampup) on each instance. The combined throughput I am
> getting in this model is much lower than in x.
> What could be the case of such a behaviour.
>
> Ideally the combined throughput should be equivalent to x.
>
> All the jmeter clients are server class machines with CentOS. Jmeter is
> running in non GUI mode.
> wht could be the cause of this behaviour?
It's very unlikely to be a JMeter problem, so you need to do some more
tests to try and establish what causes the difference when using 2
test hosts.
What throughput do you get if you run the 50 thread test on the second host?
Do you get similar throughput to a 50 thread run on the original host?
What about running two 25 thread tests on the same host?
Do you get similar thoughput to a single test?
What sampler are you using to post the XML?
> Thanks in advace,
> Ravindra
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Single Jmeter instance produces better throughput than multiple
Posted by Deepak Goel <de...@gmail.com>.
Hey
Namaskara~Nalama~Guten Tag
It would be great if you can post results before and after one solves the
problem.
Something like that:
1. Throughput
2. Number of Concurrent Users
3. Response Time
4. Think Time
5. Utilization (Client, Server, Network)
6. System Configuration
Its easier to answer and learn.
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 Sat, Jan 8, 2011 at 5:56 AM, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi All,
>
> This problem got resolved. yesterday when I was running the test, there was
> some problem with the network. I had no clue what was going wrong.
> Today it got resolved when I ran the test in a separate subnet where there
> was no traffic.
>
> Sorry for the inconvenience guys.
>
> Cheers,
> Ravindra
>
> On Fri, Jan 7, 2011 at 12:46 PM, alwaysbmore <al...@yahoo.com> wrote:
>
> >
> > Ran into the same issue, I'm still trouble shooting on the cause.
> > --
> > View this message in context:
> >
> http://jmeter.512774.n5.nabble.com/Single-Jmeter-instance-produces-better-throughput-than-multiple-tp3331381p3332454.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: Single Jmeter instance produces better throughput than multiple
Posted by Ravindra Gupta <ra...@gmail.com>.
Hi All,
This problem got resolved. yesterday when I was running the test, there was
some problem with the network. I had no clue what was going wrong.
Today it got resolved when I ran the test in a separate subnet where there
was no traffic.
Sorry for the inconvenience guys.
Cheers,
Ravindra
On Fri, Jan 7, 2011 at 12:46 PM, alwaysbmore <al...@yahoo.com> wrote:
>
> Ran into the same issue, I'm still trouble shooting on the cause.
> --
> View this message in context:
> http://jmeter.512774.n5.nabble.com/Single-Jmeter-instance-produces-better-throughput-than-multiple-tp3331381p3332454.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: Single Jmeter instance produces better throughput than multiple
Posted by alwaysbmore <al...@yahoo.com>.
Ran into the same issue, I'm still trouble shooting on the cause.
--
View this message in context: http://jmeter.512774.n5.nabble.com/Single-Jmeter-instance-produces-better-throughput-than-multiple-tp3331381p3332454.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: Single Jmeter instance produces better throughput than multiple
Posted by Deepak Shetty <sh...@gmail.com>.
it's possible your 0 rampup combined with a single sampler , no iteration
might give you invalid results - Your server may not be seeing 50 concurrent
connections in the first case. How much do the min/avg/max response times
vary ?
On Thu, Jan 6, 2011 at 6:26 PM, Ravindra Gupta <ra...@gmail.com> wrote:
> Hi All,
>
> I am facing a problem where if i run a single jmeter instance it produces
> much better throughput than when i am running 2 jmeter instance.
>
> Details about the test
>
> Script is very simple, it has just one sampler which post xml request to
> app
> server. Each thread posts one xml request and finishes.
>
> In one test,
> At the same time I simulated 50 Threads with no 0 rampup time. Lets say it
> produces x throughput.
>
> In another test
> I simulated 2 jmeter instance (not in master slave) from multiple machines
> with 25 thread (0 rampup) on each instance. The combined throughput I am
> getting in this model is much lower than in x.
> What could be the case of such a behaviour.
>
> Ideally the combined throughput should be equivalent to x.
>
> All the jmeter clients are server class machines with CentOS. Jmeter is
> running in non GUI mode.
> wht could be the cause of this behaviour?
>
> Thanks in advace,
> Ravindra
>