[javagroups-users] Re: Bug or a feature ?

[Bela B. <be...@ya...>]

Takoua ABDELLATIF wrote:

> Hi,
> I met in my tests very stange results. In many cases, the latency of a
> multicast (the time to multicast a message to n nodes) is better when node
> number increases!


are you measuring only IP multicasts, or JGroups end-to-end latency ? 
Because in JGroups you have many more things that might cause this.

> I am using the default protocol stack and I measure an average time after
> 10000 multicasts. Ethereal analysis of packets show no strange 
> behaviour of
> my network.
> Any explanation?


More details needed.

Can you run the tests described in 
http://www.jgroups.org/javagroupsnew/docs/performance.html to see 
whether you get the same results ?

-- 
Bela Ban
http://www.jgroups.org
Cell: (408) 316-4459