jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: JMeter adding to response time
Date Thu, 02 Feb 2012 14:19:50 GMT
On 2 February 2012 14:11, Stanton W Derry <derry@cpt.com> wrote:
> As a matter of fact I have one Transaction Controller as a parent of all web
> services call.

Its samples will show the overall time for its children, but the
individual samples will show the actual time.

You can experiment with this by using a Java Sampler.

> Thread Group - 250 threads
>   Transaction Controller
>        HTTP Request
>        ....
>        HTTP Request
>    Test Action
>        Constant Throughput - 28.5
>
> I've spread the load across 5 engines, each engine is driving at 500tps per
> request.  Hosts average CPU is ~60%, two hosts one with 2 JMeter instances
> and one with 3 JMeter instances.  NIC bandwidth utilization is ~5% on both
> hosts.  Also, I'm not running in a Master/Slave configuration.

DO you see the same behaviour on all JMeter instances?
Are the elapsed times similar?

What happens if you run a single test with a single thread?

What happens if you use a third host with a single JMeter instance and
a single thread running at the same time as the load test?

> Stan
>
>
> On 2/2/12 4:45 AM, sebb wrote:
>>
>> On 2 February 2012 05:03, Stanton W Derry<derry@cpt.com>  wrote:
>>
>>>
>>> I'm performance testing a set of web services with an expected response
>>> time
>>> of ~20-25ms.   I've developed the script using HTTP Request (Http 3.1)
>>> samplers doing POSTs with a SOAP payload.  The samplers have Response
>>> Assertions, Regular Expression Extractors, and some have XPath
>>> Extractors.
>>>  The results have created a bit of a quandry. The average response times
>>> in
>>> JMeter are in the range of 50-60ms and the average request times on the
>>> web
>>> servers is ~15ms.  The network latency is<  1ms from the JMeter load
>>> generators to the web servers.
>>>
>>> Question:  Is JMeter processing adding in the additional  35-45ms?
>>>
>>
>> Very unlikely unless JMeter or the host is so busy that it cannot
>> service all its threads promptly.
>> But in that case I would expect to see wide variations in elapsed time
>> between threads .
>>
>> You could try using HttpClient 4.x instead of 3.1 and see if that
>> makes a difference.
>>
>>
>>>
>>> Does the response time include processing of the assertions and
>>> extractors?
>>>
>>
>> No, not unless you use a Transaction Controller.
>>
>>
>>>
>>> Stan
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
>>> For additional commands, e-mail: user-help@jmeter.apache.org
>>>
>>>
>>
>>
>>

---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
For additional commands, e-mail: user-help@jmeter.apache.org


Mime
View raw message