cxf-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel Kulp <daniel.k...@iona.com>
Subject Re: CXF & Celtix performance test result
Date Tue, 03 Oct 2006 14:25:47 GMT

On Monday October 02 2006 9:58 pm, Jiang, Ning (Willem) wrote:
> The two machines are connect with 100M switch, if I remember right.
> I think the test result will be limmited by the machines which run the
> client side, because when I used top to monitor the client side machine,
> its cpu was upto 90% when I started 10 client and the server side's cpu is
> only 40%.

That makes sense if you are using 10 Client VM's instead of 10 threads in a 
single VM.   The overhead (memory, cache misses, etc...) of that many VM's 
would definitely create a large hit on the client.    The server would be 
running in a single VM so it can better user the processor caches and such.

Dan



>
> Willem.
>
> -----Original Message-----
> From:	Kulp, John Daniel
> Sent:	2006-10-2 (星期一) 11:07
> To:	cxf-dev@incubator.apache.org
> Cc:
> Subject:	Re: CXF & Celtix performance test result
>
>
> Dan,
>
> I think the first number is the number of client threads running on the
> client machine.   Thus, in the echoString case, we're peaking around 8
> client threads.   Keep in mind, both client and server are only dual
> processor machines.
>
> Willem: can I ask how the two machines are connected?   I did an ethereal
> dump on the tcp stream and the echoString message involves 1568 bytes in
> the ethernet payloads.  At 2617/sec, that's 4.1MB/sec EACH WAY (if I did my
> math right), plus overhead for the ethernet and tcp headers and such.   
> That would be saturating a 100MB half-duplex link.   I'd like to make sure
> we aren't actually hitting a networking limit or similar.   I don't suppose
> there is anyway to get them connected via gigabit?
>
> Thanks!
> Dan
>
> On Saturday September 30 2006 10:25 pm, Dan Diephouse wrote:
> > How many client threads are being run here?
> > Thanks,
> > - Dan
> >
> > Jiang, Ning (Willem) wrote:
> > >I finally realized the mail list had remove my attached file.
> > >I had to copy the test result to this mail, sorry for the format.
> > >
> > >	Basic Type test (Echo String)
> > >
> > >	Celtix 1.0 	CXF 2.0M1
> > >Client 	Through put 	Response Time(ms) Through put 	Response Time (ms)
> > >1 	244.26  4.09 	599.09  	1.67
> > >2 	387.49 	5.16 	1119.08 	1.79
> > >3 	529.71 	5.66 	1541.06 	1.95
> > >4 	569.81 	7.02 	1904.97 	2.1
> > >5 	685.51 	7.29 	2175.04 	2.3
> > >6 	737.4 	8.14 	2377.22 	2.52
> > >7 	738.05 	9.48 	2539.41 	2.76
> > >8 	732.47 	10.92 	2617.16 	3.06
> > >9 	724.99 	12.41 	2572.39 	3.5
> > >10 	717.44 	13.94 	2521.87 	3.97
> > >
> > >
> > >	Complex Type test (Echo Struct)
> > >
> > >	Celtix 1.0 	CXF 2.0M1
> > >Client 	Through put 	Response Time(ms) Through put 	Response Time (ms)
> > >1 	184.4 	5.42 	359.13  	2.79
> > >2 	314.26 	6.36 	660.66  	3.03
> > >3 	430.01 	6.98 	885.91  	3.39
> > >4 	497.89 	8.03 	1095.78 	3.65
> > >5 	571.39 	8.75 	1256.33 	3.98
> > >6 	607.29 	9.88 	1401.40 	4.28
> > >7 	607.18 	11.53 	1482.88 	4.72
> > >8 	632.96 	12.64 	1511.44 	5.29
> > >9 	595.55 	15.11 	1513.82 	5.95
> > >10 	636.09 	15.72 	1473.38 	6.79
> > >
> > >-----Original Message-----
> > >From:	Jiang, Ning (Willem)
> > >Sent:	2006-9-30 (星期六) 20:34
> > >To:	cxf-dev@incubator.apache.org
> > >Cc:
> > >Subject:	FW: CXF & Celtix performance test result
> > >
> > >
> > >Sorry, I forgot to attache the result html.
> > >
> > >-----Original Message-----
> > >From:	Jiang, Ning (Willem)
> > >Sent:	2006-9-30 (星期六) 5:27
> > >To:	cxf-dev@incubator.apache.org
> > >Cc:
> > >Subject:	CXF & Celtix performance test result
> > >
> > >Hi ,
> > >
> > >I just finished the CXF performance test base on Dan Kulp's latest
> > >performance tuning.
> > >Here is the test result, attached as HTML.
> > >It is great news that the result shows CXF is 2~3 times fastter than
> > > Celtix.
> > >
> > >The test case is in the trunk/test/performance/.
> > >Basic_Type just test echo string for 1K message
> > >Complex_Type test  echo struct for about 2K complex type message.
> > >
> > >Test had been done with  2  DELL Workstation 670 , Linux ES3.0,  2 CPU
> > >2.8 GHZ,  memory 2G
> > >(one run as server  the  other run as  client)
> > >
> > >Cheers,
> > >Willem.

-- 
J. Daniel Kulp
Principal Engineer
IONA
P: 781-902-8727    C: 508-380-7194   F:781-902-8001
daniel.kulp@iona.com

Mime
View raw message