You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mina.apache.org by Edouard De Oliveira <do...@yahoo.fr> on 2011/09/20 16:03:51 UTC
Re : new benchmark results for mina1/mina2/grizzly/netty
First you could try enabling some socket options (like on the others frameworks) to the mina2 test :
server side add these before bind :
((NioSocketAcceptor) acceptor).setReuseAddress(true);
((NioSocketAcceptor) acceptor).getSessionConfig().setTcpNoDelay(true);
client side add after the setTcpnodelay line :
ioConnector.getSessionConfig().setReuseAddress(true);
what are the new figures with these settings ?
Cordialement, Regards,
-Edouard De Oliveira-
________________________________
De : BlueDavy Lin <bl...@gmail.com>
À : paliwalashish@gmail.com
Cc : dev@mina.apache.org
Envoyé le : Mardi 20 Septembre 2011 5h10
Objet : Re: new benchmark results for mina1/mina2/grizzly/netty
environment can be seen here: http://code.google.com/p/nfs-rpc
If u tuned,u can also give the code to me,or If u agree,I can add u to
nfs-rpc committer,so u can submit the code directly,after u tuned the
code,I'll run the benchmark test in the same environment again.
2011/9/20 Ashish <pa...@gmail.com>:
> Thanks for sharing :)
>
> Can you share a bit more about the environment that you are using to
> run the tests
>
> I shall try to run the benchmark at my end and see if I can tune the
> implementation.
>
> thanks
> ashish
>
> On Tue, Sep 20, 2011 at 7:08 AM, BlueDavy Lin <bl...@gmail.com> wrote:
>> hi!
>>
>> Alexey(the leader of grizzly) changed some grizzly &
>> nfs-rpc-grizzly code,now the new benchmark results are:
>> Mina: 123k
>> Mina2: 65k
>> Grizzly: 161k
>> Netty: 160k
>>
>> All're runned in 100 concurrents,100 request & response byte
>> size,single connection,protobuf serialize/deserialize.
>>
>> more details pls see: http://code.google.com/p/nfs-rpc
>>
>> --
>> =============================
>> | BlueDavy |
>> | http://www.bluedavy.com |
>> =============================
>>
>
>
>
> --
> thanks
> ashish
>
> Blog: http://www.ashishpaliwal.com/blog
> My Photo Galleries: http://www.pbase.com/ashishpaliwal
>
--
=============================
| BlueDavy |
| http://www.bluedavy.com |
=============================
Re: Re : Re : new benchmark results for mina1/mina2/grizzly/netty
Posted by Ashish <pa...@gmail.com>.
On Thu, Sep 22, 2011 at 8:16 PM, Edouard De Oliveira
<do...@yahoo.fr> wrote:
> @ashish he already did like said in a previous mail but it didn't change much as well
oops i missed that... i tried adding Executor filter as well, but it
made my results worse :(
>
> @bluedavy i'd like to know what are exactly the parameters used to start the client and the server ? (timeout , concurrents, byte size etc ...)
> also plz note that this link is broken : http://bluedavy.com/projects/nfs-rpc/benchmark-1.xslx
>
> Cordialement, Regards,
> -Edouard De Oliveira-
>
>
> ----- Mail original -----
> De : Ashish <pa...@gmail.com>
> À : dev@mina.apache.org; Edouard De Oliveira <do...@yahoo.fr>
> Cc :
> Envoyé le : Jeudi 22 Septembre 2011 15h24
> Objet : Re: Re : new benchmark results for mina1/mina2/grizzly/netty
>
> On Tue, Sep 20, 2011 at 7:33 PM, Edouard De Oliveira
> <do...@yahoo.fr> wrote:
>> First you could try enabling some socket options (like on the others frameworks) to the mina2 test :
>>
>> server side add these before bind :
>>
>> ((NioSocketAcceptor) acceptor).setReuseAddress(true);
>> ((NioSocketAcceptor) acceptor).getSessionConfig().setTcpNoDelay(true);
>>
>> client side add after the setTcpnodelay line :
>> ioConnector.getSessionConfig().setReuseAddress(true);
>>
>>
>> what are the new figures with these settings ?
>>
>> Cordialement, Regards,
>> -Edouard De Oliveira-
>
> @Edouard - updated the benchmark code with your suggestions. Did
> marginally increased TPS on my laptop :(
>
> @bluedavy - can you rerun the test again.
>
> thanks
> ashish
>
>
--
thanks
ashish
Blog: http://www.ashishpaliwal.com/blog
My Photo Galleries: http://www.pbase.com/ashishpaliwal
Re: Re : Re : new benchmark results for mina1/mina2/grizzly/netty
Posted by BlueDavy Lin <bl...@gmail.com>.
maybe the website have some problem...I attached the benchmark file...
client start args:
[serverIP] [serverPort] 100 500 3 100 120 1
server start args:
[listenPort] 300 100
2011/9/22 Edouard De Oliveira <do...@yahoo.fr>:
> @ashish he already did like said in a previous mail but it didn't change much as well
>
> @bluedavy i'd like to know what are exactly the parameters used to start the client and the server ? (timeout , concurrents, byte size etc ...)
> also plz note that this link is broken : http://bluedavy.com/projects/nfs-rpc/benchmark-1.xslx
>
> Cordialement, Regards,
> -Edouard De Oliveira-
>
>
> ----- Mail original -----
> De : Ashish <pa...@gmail.com>
> À : dev@mina.apache.org; Edouard De Oliveira <do...@yahoo.fr>
> Cc :
> Envoyé le : Jeudi 22 Septembre 2011 15h24
> Objet : Re: Re : new benchmark results for mina1/mina2/grizzly/netty
>
> On Tue, Sep 20, 2011 at 7:33 PM, Edouard De Oliveira
> <do...@yahoo.fr> wrote:
>> First you could try enabling some socket options (like on the others frameworks) to the mina2 test :
>>
>> server side add these before bind :
>>
>> ((NioSocketAcceptor) acceptor).setReuseAddress(true);
>> ((NioSocketAcceptor) acceptor).getSessionConfig().setTcpNoDelay(true);
>>
>> client side add after the setTcpnodelay line :
>> ioConnector.getSessionConfig().setReuseAddress(true);
>>
>>
>> what are the new figures with these settings ?
>>
>> Cordialement, Regards,
>> -Edouard De Oliveira-
>
> @Edouard - updated the benchmark code with your suggestions. Did
> marginally increased TPS on my laptop :(
>
> @bluedavy - can you rerun the test again.
>
> thanks
> ashish
>
>
--
=============================
| BlueDavy |
| http://www.bluedavy.com |
=============================
Re : Re : new benchmark results for mina1/mina2/grizzly/netty
Posted by Edouard De Oliveira <do...@yahoo.fr>.
@ashish he already did like said in a previous mail but it didn't change much as well
@bluedavy i'd like to know what are exactly the parameters used to start the client and the server ? (timeout , concurrents, byte size etc ...)
also plz note that this link is broken : http://bluedavy.com/projects/nfs-rpc/benchmark-1.xslx
Cordialement, Regards,
-Edouard De Oliveira-
----- Mail original -----
De : Ashish <pa...@gmail.com>
À : dev@mina.apache.org; Edouard De Oliveira <do...@yahoo.fr>
Cc :
Envoyé le : Jeudi 22 Septembre 2011 15h24
Objet : Re: Re : new benchmark results for mina1/mina2/grizzly/netty
On Tue, Sep 20, 2011 at 7:33 PM, Edouard De Oliveira
<do...@yahoo.fr> wrote:
> First you could try enabling some socket options (like on the others frameworks) to the mina2 test :
>
> server side add these before bind :
>
> ((NioSocketAcceptor) acceptor).setReuseAddress(true);
> ((NioSocketAcceptor) acceptor).getSessionConfig().setTcpNoDelay(true);
>
> client side add after the setTcpnodelay line :
> ioConnector.getSessionConfig().setReuseAddress(true);
>
>
> what are the new figures with these settings ?
>
> Cordialement, Regards,
> -Edouard De Oliveira-
@Edouard - updated the benchmark code with your suggestions. Did
marginally increased TPS on my laptop :(
@bluedavy - can you rerun the test again.
thanks
ashish
Re: Re : new benchmark results for mina1/mina2/grizzly/netty
Posted by Ashish <pa...@gmail.com>.
On Tue, Sep 20, 2011 at 7:33 PM, Edouard De Oliveira
<do...@yahoo.fr> wrote:
> First you could try enabling some socket options (like on the others frameworks) to the mina2 test :
>
> server side add these before bind :
>
> ((NioSocketAcceptor) acceptor).setReuseAddress(true);
> ((NioSocketAcceptor) acceptor).getSessionConfig().setTcpNoDelay(true);
>
> client side add after the setTcpnodelay line :
> ioConnector.getSessionConfig().setReuseAddress(true);
>
>
> what are the new figures with these settings ?
>
> Cordialement, Regards,
> -Edouard De Oliveira-
@Edouard - updated the benchmark code with your suggestions. Did
marginally increased TPS on my laptop :(
@bluedavy - can you rerun the test again.
thanks
ashish