Re: Re: Jetty 4.1.0 load test

To service 1500 simultaneous requests, you will need max threads
at around 1800 and min threads at around 300.  Jetty is run on
big irons with such configurations.

With 300 simultaneous CONNECTIONS, you should not run out of threads
if you have max threads at 350.

But you have to work out what your test client is doing and is it
realistic -  It sounds like it may be have 300 simultaneous REQUESTS
but maybe keep more than 300 connections open - or cycling the
connections very quickly.

So to really answer these questions, you need to find out what your
test client is doing with connections and how many requests per connection
it is likely to send etc.    You then need to consider is that similar
to what your real clients will do?

You can turn on Jetty stats via the jmx interface if you want to
collect stats about how a real client interacts with the service.
You then need to adjust your test client to give similar stats - else
you are not measuring real load.

cheers




Stallone wrote:> Hi ,
> 
> We did test for 300 concurrent requests and here is the results and 
> issues...
> 
>  1. If we fire all the 300 users at the same time then Jetty gives 
> out a lot of warning messages as mentioned below and the initial 
> login will be delayed. 
> 16:20:46.825 WARN!! OUT OF THREADS: SocketListener@xxxxxxx:80
> This messages come up even if the MAx threads is set to 350 which is 
> more than concurrent requests.
> 
>  2. If we specify a delay time for login , that is login 300 users in 
> a span of 1 minute then the warn messages appear in the later part of 
> our application (all distributed ) and that delays requests for 
> nearly 5-6 secs which otherwise takes 1 sec to complete.
>  
> My question is why are we getting thread warn message even if thread 
> max is set to a number more than concurrent-requests also  our target 
> is to run 1500 concurrent requests with a reasonable response time... 
> what is the thread value we have to specify if this needs to be 
> achieved?
> 
> is there something we have to set for efficient re-use of threads.....
> 
> Please advice... 
> 
> 
> 
> 
> 
> --- In jetty-support@xxxxxxxxxxxxxxx, Greg Wilkins <gregw@xxxx> wrote:
> 
>>
>>Stallone wrote:> Hi ,
>>
>>
>>>        <Set name="MinThreads">5</Set>
>>
>>This is how many threads you want allocated when you are idle.
>>Effects how well you respond to sudden bursts.
>>
>>Also used when threads>maxThreads-minThreads to determine
>>low resources.
>>
>>
>>>        <Set name="MaxThreads">255</Set>
>>
>>How many threads - and thus how many simultaneous connections you
>>want.
>>
>>
>>>        <Set name="MaxIdleTimeMs">30000</Set>
>>
>>How long to leave a connection connected if it has no requests
>>on it.  Higher values give better individual client performance
>>lower values will allow more clients to cycle through your server.
>>
>>
>>>        <Set name="MaxReadTimeMs">10000</Set>
>>
>>THIS SHOULD ALWAYS BE SET THE SAME AS MAXIDLETIMEMS!!!!!!!!!!!
>>You get an extra 15% throughput in the server this way.
>>This param is now deprecated.
>>
>>
>>>        <Set name="MaxStopTimeMs">10000</Set>
>>
>>Ignorable and now deprecated.
>>
>>
>>>        <Set name="LowResourcePersistTimeMs">2000</Set>
>>
>>If threads>maxThreads-minThreads then this timeout is used
>>instead of maxIdleTimeMs.  Allows you to quickly discard idle
>>connections if you are running low on threads.  Can be reduced
>>to almost nothing if you have a bursty load.
>>
>>
>>>      </New>
>>>    </Arg>
>>>  </Call>
>>>
>>>The above settings runs ok with 200 concurrent requests (approx 3-
> 
> 5 
> 
>>>secs/ request) , with lots  "out of thread" warnings in log . 
> 
> After  
> 
>>>increaseing the max threads to 500 and min to 100 the warnings 
> 
> have 
> 
>>>gone but still slow but better than the previous run. 
>>>
>>>
>>>1.Can anyone please point me to a detailed document which 
> 
> explains 
> 
>>>addlistner tag like what are MaxIdleTimeMs, MaxReadTimeMs,
>>>MaxReadTimeMs etc and the effects of changing them.
>>>
>>>2. If I set 500 threads for 200 concurrent requests then for 2000 
>>>concurrent requests we have to specify a very huge value of 
>>>threads... I am sure that this will inturn de-grade performance 
> 
> So is 
> 
>>>it better to live with the warnings  or am I am missing something 
> 
> in 
> 
>>>the settings....
>>>
>>>3. some documentation on running jetty as a cluster.......
>>
>>See mod_jk stuff in FAQ.
>>
>>
>>
>>>Thanks in advance...
>>>
>>>Darshan
>>>
>>>
>>>
>>>
>>>
>>>For the latest information about Jetty, please see 
> 
> http://jetty.mortbay.org
> 
>>> 
>>>
>>>Your use of Yahoo! Groups is subject to 
> 
> http://docs.yahoo.com/info/terms/ 
> 
>>
>>
>>-- 
>>Greg Wilkins<gregw@xxxx>             Phone/fax: +44 7092063462
>>Mort Bay Consulting Australia and UK.          
> 
> http://www.mortbay.com
> 
> 
> 
> For the latest information about Jetty, please see http://jetty.mortbay.org
> 
>  
> 
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 
> 



-- 
Greg Wilkins<gregw@xxxxxxxxxxx>             Phone/fax: +44 7092063462
Mort Bay Consulting Australia and UK.          http://www.mortbay.com


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get 128 Bit SSL Encryption!
http://us.click.yahoo.com/CBxunD/vN2EAA/xGHJAA/CefplB/TM
---------------------------------------------------------------------~->

For the latest information about Jetty, please see http://jetty.mortbay.org

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/