[Spread-users] spread's performance: solaris vs. linux

[Yair Amir]

Hi,

George beats me on that :)

Your Solaris numbers are respectable for out-of-the-box Spread running
on 100Mbits/sec. (I think you get a bit over 60Mbits/sec clean data).

The Linux numbers are really bad - about 1/6 of that. Sparcs and
Intels on the same configuration will cause a bit of reduction in
speed due to conversions, but that should not be more than a few
percent difference.

It was not clear from your tests which daemons were participating.
If all 4 daemons participated in all of the tests then the problem is
not with the linux spread daemons because they kept up during the
solaris tests. If not all daemons participated in all tests then you
should specify exactly which daemons are running for which tests.

One thought I have is maybe your multicast address does not work well
on your original-sender linux box and its messages are always lost by
the solaris guys. Monitor reports will show that. You can also check
whether your network is set correctly by running the s and r programs
to see if a multicast address sent from each of the participating
machines is actually received by each of them.

I am sure this is solvable - we regularly run configurations with both
Linux and Solaris with no performance problems.

      Cheers,

      :) Yair.
      
George> Yair will ask you for a couple loops of output from the monitor program, 
George> but can you also send your spread.conf file over, as well as netstat -rn 
George> output from your hosts?

George> Best,

George> George

George> On Thursday, June 27, 2002, at 09:03 AM, Jan Rickes wrote:

>> hello spread people,
>>
>> i'm evaluating spread's performance (spread-src-3.16.2.tar)
>> and was encountering dramatic performance losts,
>> if there was any LINUX host in my segment.
>> does anyone have a clue what's the reason of this phenomenon?
>>
>> as well i was testing Flooder.java and i've got several times this 
>> exception:
>>
>> java.lang.NegativeArraySizeException
>>         at 
>> spread.SpreadConnection.internal_receive(SpreadConnection.java:1184)
>>         at spread.SpreadConnection.access$2(SpreadConnection.java)
>>         at 
>> spread.SpreadConnection$Listener.run(SpreadConnection.java:1652)
>>
>> what's happening here?
>>
>>
>> any comments would be much appreciated :)
>> regards, jan
>>
>> -------------------------------------------------------------------------------------
>> performance results:
>>
>> ** linux **
>> ren_linux> time ./spflooder -m 10000 -b 100000 -s 6824 -wo
>> flooder: connecting to 6824
>> flooder: starting  multicast of 10000 messages, 100000 bytes each (self 
>> discarding).
>> flooder: completed  10000 messages of 100000 bytes
>> flooder: completed multicast of 10000 messages, 100000 bytes each.
>> real    13m32.138s
>> user    0m0.080s
>> sys     0m6.320s
>>
>> stimpy_linux> time ./spflooder -s 6824 -ro
>> flooder: connecting to 6824
>> flooder: Only receiving messages
>> flooder: completed  10000 messages of 100000 bytes
>> flooder: completed multicast of 10000 messages, 1000 bytes each.
>> real    14m16.372s
>> user    0m0.130s
>> sys     0m6.200s
>>
>>
>> ** solaris **
>> ren_solaris> time ./spflooder -m 10000 -b 100000 -s 6824 -wo
>> flooder: connecting to 6824
>> flooder: starting  multicast of 10000 messages, 100000 bytes each (self 
>> discarding).
>> flooder: completed multicast of 10000 messages, 100000 bytes each.
>> real    2m12.567s
>> user    0m0.190s
>> sys     0m9.250s
>>
>> stimpy_solaris> time ./spflooder -ro -s 6824
>> flooder: connecting to 6824
>> flooder: Only receiving messages
>> flooder: completed multicast of 10000 messages, 100000 bytes each.
>> real    2m19.112s
>> user    0m0.560s
>> sys     0m10.520s
>>
>>
>> ren_solaris:    SunOS 5.6 Generic_105181-20 sun4u sparc 
>> SUNW,UltraSPARC-IIi-Engine
>> stimpy_solaris: SunOS 5.8 Generic_108528-14 sun4u sparc SUNW,UltraAX-MP
>> ren_linux:      Linux 2.4.10-4GB #1 Tue Sep 25 12:33:54 GMT 2001 i686 
>> unknown
>> stimpy_linux:   Linux 2.4.10-64GB-SMP #3 SMP Thu Dec 6 15:35:10 CET 
>> 2001 i686 unknown
>>
>>
>> Spread_Segment 224.100.100.224:6824 {
>>         ren_solaris     172.27.27.21
>>         stimpy_solaris  172.27.27.28
>>         ren_linux       172.27.27.19
>>         stimpy_linux    172.27.100.237
>> }
>>
>>
>> _______________________________________________
>> Spread-users mailing list
>> Spread-users at lists.spread.org
>> http://lists.spread.org/mailman/listinfo/spread-users
>>
>>
George> // George Schlossnagle
George> // Principal Consultant
George> // OmniTI, Inc          http://www.omniti.com
George> // (c) 301.343.6422   (e) george at omniti.com
George> // 1024D/1100A5A0  1370 F70A 9365 96C9 2F5E 56C2 B2B9 262F 1100 A5A0



George> _______________________________________________
George> Spread-users mailing list
George> Spread-users at lists.spread.org
George> http://lists.spread.org/mailman/listinfo/spread-users