On 24 June 2013 19:08, Robin D. Wilson <[email protected]> wrote:
> You also need to evaluate _all_ of the potential bottlenecks when running 
> multiple threads from the same machine. For example, the
> network interface - if you are requesting pages that return 1MB of data, and 
> you try to do 1000 threads - it is unlikely that your
> load-generating system can handle 1GB of data from each simultaneous page as 
> easily as say 10 machines can handle 100MB each. (Of
> course, then you have to look at the test servers - to make sure that they 
> are not similarly bottlenecked.
>
> Likewise, if your threads do a lot of pre and post processing (essentially 
> parsing the data that comes back from the servers) and
> have a lot listeners, can your client machine really handle that amount of 
> workload "times 1000"? It is unlikely unless you have the
> most basic of request/response values, and parsing rules.
>
> One example would be the 'tree listener'... If you are recording every 
> response in the tree listener, it will eventually run your
> machine out of memory. Not only that, but before it runs you out of memory - 
> it will dramatically slow down your test. So in our
> tests, we only use the 'tree listener' for recording errors. But even with 
> just that, there is still a measurable difference in
> performance when it is enabled versus disabled.

For maximum performance, don't use the tree listener.
Or, better, use non-GUI mode - the GUI parts of Listeners are deactivated.

> I run our load from 4 different load-generator machines (all running JMeter), 
> each with Core i7 processors, and 8GB of RAM. We never
> run more than 200 threads simultaneously (50 per box) - simply because we've 
> found that when we ramp up the testing clients beyond
> that level, the slowdown comes from the clients - not the servers. But then 
> our tests are pretty involved, with lots of pre/post
> processors and assertions.
>
> What you want to look for is how do you generate sufficient load on the 
> servers - and how many load-generating clients are required
> for that. You can try 1000 threads on a single client, but you look at the 
> load on the servers and see it nowhere near capacity, and
> then you look at the load on the load-generating client and see it maxed out 
> in CPU, Memory, I/O, (any single one will block it from
> generating more load), then you need to lower the number of simultaneous 
> threads on that client until you see the best mix of
> performance from the client vs. capacity on the server. Then you can add more 
> clients to meet your goal of 1000 simultaneous threads
> hitting the server.
>
> --
> Robin D. Wilson
> Sr. Director of Web Development
> KingsIsle Entertainment, Inc.
> VOICE: 512-777-1861
> http://www.kingsisle.com
>
>
> -----Original Message-----
> From: Deepak Shetty [mailto:[email protected]]
> Sent: Monday, June 24, 2013 12:05 PM
> To: JMeter Users List
> Subject: Re: System configuration
>
> you could probably get it to work , but I doubt your results would be 
> accurate. (i.e. you can run more threads but it might not
> simulate the same load as if you had as many machines as threads) Usually 
> with a windows client class machine , dont exceed 100-200
> threads unless your tests have large wait times In most cases you have to try 
> it out and see and measure - both your client machine
> and your server since answers to questions like yours are too heavily 
> dependent on environment and script.
>
>
> On Mon, Jun 24, 2013 at 2:36 AM, Saranya C 
> <[email protected]>wrote:
>
>> Hi All,
>>
>> I would like to know the system configuration details to run 1000
>> threads in jmeter.
>>
>> Physical memory: 2GB
>> OS: Windows XP
>> JVM: 1.7
>> Jmeter: 2.9
>>
>> Will the above specification workout?
>>
>> Thanks and regards,
>> Saranya C
>> ======================================================================
>> ======================================================================
>> ==================
>> Disclaimer: This message is intended only for the individual or entity
>> to which it is addressed. It may contain privileged, confidential
>> information which is exempt from disclosure under applicable laws. If
>> you are not the intended recipient, please note that you are strictly
>> prohibited from disseminating or distributing this information (other
>> than to the intended
>> recipient) or copying this information. If you have received this
>> communication in error, please notify us immediately by return email.
>> ======================================================================
>> ======================================================================
>> ==================
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to