LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [lvs-users] Ldirector Performance Testing

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: [lvs-users] Ldirector Performance Testing
From: Anil Pillai <rcamphor@xxxxxxxxx>
Date: Fri, 29 Oct 2010 22:57:13 +0530
Initially i have tried with same 100 concurrency and the results are almost
identical. That is where my doubt started. So later when hitting the
director i tried with higer concurrency and there is marginal increase in
the TPS.
I am not sure how much %age of increase in TPS we can expect via Load
balancer.


On Fri, Oct 29, 2010 at 7:05 PM, Henrique Fernandes <sf.rique@xxxxxxxxx>wrote:

> Why are you raising concurrency when test diretor ?  Like when you do this,
> i guessi you increse by 3 you number of "clients". So i guess the director
> is ok right ?
>
> Try use same ab comand, just change the ip.
>
> If i got it right, you are using 100 client with 1000 request. At one
> server. it gives 150 RPS . And when you give 300 clietns with 1000 request
> at diretord it gives 165 RPS i guess its better...
>
> Or explain to me why you inscrease the number of clients...
>
>
> []'sf.rique
>
>
> On Fri, Oct 29, 2010 at 11:13 AM, Anil Pillai <rcamphor@xxxxxxxxx> wrote:
>
> > Most of the time i am testing from one server only and while testing from
> > one server i am geting around 155 TPS on Cluster.
> >
> > When tried from 2 server the TPS is getting splitted like 75 TPS on one
> > server and 80 TPS on other sever.
> >
> > In actual setup, Jboss (Port 8080) is running on Real servers. Web
> Services
> > are deployed on these Jboss.
> >
> >    <Connector port="8080" address="${jboss.bind.address}"
> >         maxThreads="250" maxHttpHeaderSize="8192"
> >         emptySessionPath="true" protocol="HTTP/1.1"
> >         enableLookups="false" redirectPort="8443" acceptCount="100"
> >         connectionTimeout="20000" disableUploadTimeout="true" />
> >
> > When i do the ab test on one Real server directly i am geting a TPS of
> 150.
> >
> > Below is the ab test result for Real server directly.
> >
> > [root@om-01 tmp]$ ab -n1000 -c100 -p i.xml '
> > http://real1:8080/chargingManager/services/ChargingService'
> > Completed 100 requests
> > Completed 200 requests
> > Completed 300 requests
> > Completed 400 requests
> > Completed 500 requests
> > Completed 600 requests
> > Completed 700 requests
> > Completed 800 requests
> > Completed 900 requests
> > Finished 1000 requests
> >
> > Server Software:        Apache-Coyote/1.1
> > Server Hostname:        192.168.16.176
> > Server Port:            8080
> > Document Path:          /chargingManager/services/ChargingService
> > Document Length:        511 bytes
> > Concurrency Level:      100
> > Time taken for tests:   6.613239 seconds
> > Complete requests:      1000
> > Failed requests:        0
> > Write errors:           0
> > Total transferred:      894000 bytes
> > Total POSTed:           825825
> > HTML transferred:       511000 bytes
> > Requests per second:    151.21 [#/sec] (mean)
> > Time per request:       661.324 [ms] (mean)
> > Time per request:       6.613 [ms] (mean, across all concurrent requests)
> > Transfer rate:          132.01 [Kbytes/sec] received
> >                        121.95 kb/s sent
> >                        253.96 kb/s total
> > Connection Times (ms)
> >              min  mean[+/-sd] median   max
> > Connect:        0    0   2.4      0      13
> > Processing:   222  642 302.7    551    1769
> > Waiting:      221  641 302.7    550    1769
> > Total:        222  643 304.4    551    1771
> > Percentage of the requests served within a certain time (ms)
> >  50%    551
> >  66%    596
> >  75%    627
> >  80%    657
> >  90%   1017
> >  95%   1597
> >  98%   1618
> >  99%   1691
> >  100%   1771 (longest request)
> >
> >
> >  Below is the ab test result for Director.
> >
> > [root@om-01 tmp]$ ab -n1000 -c300 -p i.xml '
> > http://Director:8080/chargingManager/services/ChargingService<http://director:8080/chargingManager/services/ChargingService>
> '
> > Completed 100 requests
> > Completed 200 requests
> > Completed 300 requests
> > Completed 400 requests
> > Completed 500 requests
> > Completed 600 requests
> > Completed 700 requests
> > Completed 800 requests
> > Completed 900 requests
> > Finished 1000 requests
> >
> > Server Software:        Apache-Coyote/1.1
> > Server Hostname:        192.168.16.183
> > Server Port:            8080
> > Document Path:          /chargingManager/services/ChargingService
> > Document Length:        511 bytes
> > Concurrency Level:      300
> > Time taken for tests:   6.46081 seconds
> > Complete requests:      1000
> > Failed requests:        0
> > Write errors:           0
> > Total transferred:      894892 bytes
> > Total POSTed:           826650
> > HTML transferred:       511509 bytes
> > Requests per second:    165.40 [#/sec] (mean)
> > Time per request:       1813.824 [ms] (mean)
> > Time per request:       6.046 [ms] (mean, across all concurrent requests)
> > Transfer rate:          144.39 [Kbytes/sec] received
> >                        133.52 kb/s sent
> >                        278.06 kb/s total
> > Connection Times (ms)
> >              min  mean[+/-sd] median   max
> > Connect:        0   11  95.7      1    2999
> > Processing:   359 1706 731.1   1579    3464
> > Waiting:      359 1705 731.1   1578    3463
> > Total:        360 1718 746.9   1579    4806
> > Percentage of the requests served within a certain time (ms)
> >  50%   1579
> >  66%   2086
> >  75%   2302
> >  80%   2388
> >  90%   2616
> >  95%   2896
> >  98%   3439
> >  99%   3441
> >  100%   4806 (longest request)
> >
> >
> > On Fri, Oct 29, 2010 at 5:34 PM, Henrique Fernandes <sf.rique@xxxxxxxxx
> > >wrote:
> >
> > > Can you post ab results as well ?
> > >
> > > And how about apache, worker or prefork ? how is the, serverlimit,
> > > startservers, maxclients?
> > >
> > > []'sf.rique
> > >
> > >
> > > On Fri, Oct 29, 2010 at 9:59 AM, Henrique Fernandes <
> sf.rique@xxxxxxxxx
> > > >wrote:
> > >
> > > > But why are you testing with two serves ?
> > > >
> > > > Why not just one ab ?
> > > >
> > > >
> > > > []'sf.rique
> > > >
> > > >
> > > > On Fri, Oct 29, 2010 at 9:35 AM, Anil Pillai <rcamphor@xxxxxxxxx>
> > wrote:
> > > >
> > > >> I have tried with
> > > >>
> > > >> ab -n10000 -c100
> > > >> ab -n10000 -c200
> > > >> ..
> > > >> ..
> > > >> ab -n10000 -c600
> > > >> Even i tried -k option, but results are still the same.
> > > >>
> > > >>
> > > >>
> > > >> On Fri, Oct 29, 2010 at 4:51 PM, Graeme Fowler <graeme@xxxxxxxxxxx>
> > > >> wrote:
> > > >>
> > > >> > On Fri, 2010-10-29 at 16:38 +0530, Anil Pillai wrote:
> > > >> > > One observation.
> > > >> >
> > > >> > One more observation - ApacheBench needs to have parameters
> tweaked
> > to
> > > >> > increase the concurrency. The default is one request at a time.
> > > >> >
> > > >> > If you push the concurrency up, you should see a corresponding
> > > increase
> > > >> > in requests/sec.
> > > >> >
> > > >> > ab -c100 -n10000 http://target_url/
> > > >> >
> > > >> > For more fun, try
> > > >> >
> > > >> > ab -c100 -n10000 -k http://target_url/
> > > >> >
> > > >> > Graeme
> > > >> >
> > > >> >
> > > >> > _______________________________________________
> > > >> > Please read the documentation before posting - it's available at:
> > > >> > http://www.linuxvirtualserver.org/
> > > >> >
> > > >> > LinuxVirtualServer.org mailing list -
> > lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> > > >> > Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> > > >> > or go to http://lists.graemef.net/mailman/listinfo/lvs-users
> > > >> >
> > > >> _______________________________________________
> > > >> Please read the documentation before posting - it's available at:
> > > >> http://www.linuxvirtualserver.org/
> > > >>
> > > >> LinuxVirtualServer.org mailing list -
> lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> > > >> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> > > >> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
> > > >>
> > > >
> > > >
> > > _______________________________________________
> > > Please read the documentation before posting - it's available at:
> > > http://www.linuxvirtualserver.org/
> > >
> > > LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> > > Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> > > or go to http://lists.graemef.net/mailman/listinfo/lvs-users
> > >
> > _______________________________________________
> > Please read the documentation before posting - it's available at:
> > http://www.linuxvirtualserver.org/
> >
> > LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> > Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> > or go to http://lists.graemef.net/mailman/listinfo/lvs-users
> >
> _______________________________________________
> Please read the documentation before posting - it's available at:
> http://www.linuxvirtualserver.org/
>
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

<Prev in Thread] Current Thread [Next in Thread>