You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by "Kohn, Rodolfo" <ro...@intel.com> on 2011/04/14 14:36:20 UTC
Distributed performance test question
Hello jmeter list,
Having used jmeter for a long time now, there are various questions I have about logging and summary in distributed (client/agent) performance test.
I googled and read the mail list and didn't find a fine answer to this. I wonder if anyone can help:
1- In a jmeter configuration client-server (or master-slave). Can the client become a bottleneck when receiving the log data from all jmeter servers/agents (say I have 100 agents with 200 threads each)?
2- If the client is located in a different network, say in another country, than the agents, instead of being in the same LAN, can latency and bandwidth affect performance results as measured by jmeter?
3- Is the throughput calculated at each agent and then sent back to the client which re-calculates the different throughputs measured by all agents? Or is all calculation done at the client based on the samples' timestamps.
4- Who sets the timestamps for each sample (in the log)? Is it the client or the agent (in this case can I'm interested in knowing how different clocks are correlated)?
If anybody could answer these questions, I would save a lot of time looking at the source code.
Thanks in advance,
Rodolfo
Re: Distributed performance test question
Posted by sebb <se...@gmail.com>.
On 14 April 2011 14:41, Kohn, Rodolfo <ro...@intel.com> wrote:
> Thank you sebb.
> A few follow up questions:
> 1-Different agents will measure different avg response time, depending on the latency. Does the client creates a super-avg from all avg. reponse times? I assume throughput doesn't change.
As I wrote, the client merely collects the responses.
> 2-If the timestamp is set at the agent then it is not possible to make any calculation at the client based on the timestamp - Is it possible to discern the agent that sent a single sample? (my objective is to know if it is possible to make calculations at the client based on timestamp)
Yes, you can add the hostname to the response data - check the
Listener sections in the manual.
> Thanks again,
> Rodolfo
>
>> -----Original Message-----
>> From: sebb [mailto:sebbaz@gmail.com]
>> Sent: Thursday, April 14, 2011 9:59 AM
>> To: JMeter Users List
>> Subject: Re: Distributed performance test question
>>
>> On 14 April 2011 13:36, Kohn, Rodolfo <ro...@intel.com> wrote:
>> > Hello jmeter list,
>> >
>> > Having used jmeter for a long time now, there are various questions I
>> have about logging and summary in distributed (client/agent)
>> performance test.
>> > I googled and read the mail list and didn't find a fine answer to
>> this. I wonder if anyone can help:
>> >
>> >
>> > 1- In a jmeter configuration client-server (or master-slave). Can
>> the client become a bottleneck when receiving the log data from all
>> jmeter servers/agents (say I have 100 agents with 200 threads each)?
>>
>> Yes.
>>
>> > 2- If the client is located in a different network, say in
>> another country, than the agents, instead of being in the same LAN, can
>> latency and bandwidth affect performance results as measured by jmeter?
>>
>> Possibly.
>>
>> Should not affect individual samples, but if sending the results back
>> slows down the agent, it may cause a reduction in maximum throughput.
>>
>> > 3- Is the throughput calculated at each agent and then sent back
>> to the client which re-calculates the different throughputs measured by
>> all agents? Or is all calculation done at the client based on the
>> samples' timestamps.
>>
>> At the agent.
>>
>> The client merely collects the samples.
>>
>> > 4- Who sets the timestamps for each sample (in the log)? Is it
>> the client or the agent (in this case can I'm interested in knowing how
>> different clocks are correlated)?
>>
>> The agent.
>>
>> > If anybody could answer these questions, I would save a lot of time
>> looking at the source code.
>> > Thanks in advance,
>> > Rodolfo
>> >
>> >
>>
>> ---------------------------------------------------------------------
>> 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: Distributed performance test question
Posted by "Kohn, Rodolfo" <ro...@intel.com>.
Thank you sebb.
A few follow up questions:
1-Different agents will measure different avg response time, depending on the latency. Does the client creates a super-avg from all avg. reponse times? I assume throughput doesn't change.
2-If the timestamp is set at the agent then it is not possible to make any calculation at the client based on the timestamp - Is it possible to discern the agent that sent a single sample? (my objective is to know if it is possible to make calculations at the client based on timestamp)
Thanks again,
Rodolfo
> -----Original Message-----
> From: sebb [mailto:sebbaz@gmail.com]
> Sent: Thursday, April 14, 2011 9:59 AM
> To: JMeter Users List
> Subject: Re: Distributed performance test question
>
> On 14 April 2011 13:36, Kohn, Rodolfo <ro...@intel.com> wrote:
> > Hello jmeter list,
> >
> > Having used jmeter for a long time now, there are various questions I
> have about logging and summary in distributed (client/agent)
> performance test.
> > I googled and read the mail list and didn't find a fine answer to
> this. I wonder if anyone can help:
> >
> >
> > 1- In a jmeter configuration client-server (or master-slave). Can
> the client become a bottleneck when receiving the log data from all
> jmeter servers/agents (say I have 100 agents with 200 threads each)?
>
> Yes.
>
> > 2- If the client is located in a different network, say in
> another country, than the agents, instead of being in the same LAN, can
> latency and bandwidth affect performance results as measured by jmeter?
>
> Possibly.
>
> Should not affect individual samples, but if sending the results back
> slows down the agent, it may cause a reduction in maximum throughput.
>
> > 3- Is the throughput calculated at each agent and then sent back
> to the client which re-calculates the different throughputs measured by
> all agents? Or is all calculation done at the client based on the
> samples' timestamps.
>
> At the agent.
>
> The client merely collects the samples.
>
> > 4- Who sets the timestamps for each sample (in the log)? Is it
> the client or the agent (in this case can I'm interested in knowing how
> different clocks are correlated)?
>
> The agent.
>
> > If anybody could answer these questions, I would save a lot of time
> looking at the source code.
> > Thanks in advance,
> > Rodolfo
> >
> >
>
> ---------------------------------------------------------------------
> 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: Distributed performance test question
Posted by sebb <se...@gmail.com>.
On 14 April 2011 13:36, Kohn, Rodolfo <ro...@intel.com> wrote:
> Hello jmeter list,
>
> Having used jmeter for a long time now, there are various questions I have about logging and summary in distributed (client/agent) performance test.
> I googled and read the mail list and didn't find a fine answer to this. I wonder if anyone can help:
>
>
> 1- In a jmeter configuration client-server (or master-slave). Can the client become a bottleneck when receiving the log data from all jmeter servers/agents (say I have 100 agents with 200 threads each)?
Yes.
> 2- If the client is located in a different network, say in another country, than the agents, instead of being in the same LAN, can latency and bandwidth affect performance results as measured by jmeter?
Possibly.
Should not affect individual samples, but if sending the results back
slows down the agent, it may cause a reduction in maximum throughput.
> 3- Is the throughput calculated at each agent and then sent back to the client which re-calculates the different throughputs measured by all agents? Or is all calculation done at the client based on the samples' timestamps.
At the agent.
The client merely collects the samples.
> 4- Who sets the timestamps for each sample (in the log)? Is it the client or the agent (in this case can I'm interested in knowing how different clocks are correlated)?
The agent.
> If anybody could answer these questions, I would save a lot of time looking at the source code.
> Thanks in advance,
> Rodolfo
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org