[Spread-users] groups and daemons

[John Schultz]

I'm not sure what the bottleneck here is.  

It could be the flow control parameters of your configuration are too low to hit cpu and/or bandwidth limits.  You can change the flow control parameters manually through the spmonitor program if you have dangerous monitor turned on in your configuration file.  Out of the box the window is 60 and the personal window is 15.  You could try experimentally increasing these parameters by a multiplicative factor and see if it improves your throughput (e.g. - 120 and 30).

Cheers!

-----
John Lane Schultz
Spread Concepts LLC
Phn: 301 830 8100
Cell: 443 838 2200

On Nov 22, 2011, at 3:39 PM, Parisa wrote:

Hi ,

I was testing the spread with the following setting:

one segment with 4 daemons. which are deployed on 4 different machines.
then each daemon has two flooders connected to it. one is a writer and  
the other is a reader. these flooders are also on different machines.

Then I have 4 groups. each pair of the flooders connected to that same  
daemon, form a group. Each writer only sends to its own group.

and I am testing with SAFE_MESS type.

The behavior I am observing looks a little strange to me. I am  
monitoring the cpu and bandwidth usage at the daemons during the  
execution. After a certain load which is close to 50 Mbps for each  
writer (which in aggregate we can say 200 Mbps) throughput does not  
increase any further. Though neither cpu nor the bandwidth at daemons  
and flooders is not saturated.


Do you think I am doing sth wrong? Any suggestions why this should be  
the case?
Thanks.



Parisa

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

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3805 bytes
Desc: not available
Url : http://lists.spread.org/pipermail/spread-users/attachments/20111122/9b4b3448/attachment.bin