Peter,
Well, it's been 2-3 years since I wrote that stuff, so I don't
remember it all myself.
Bandwidth: RMI is not very efficient. I think it's more of a latency
than a bandwidth issue, but at 192K it might also be a bandwidth
issue. I don't have good figures on this, but if you really want to
know, there is a Sample MBean which accepts a byte[] and echoes it
back; this could be used to measure bandwidth and latency, in fact
one of the AMX unit tests does so in a basic way.
Lloyd
On Nov 10, 2008, at 12:19 PM, Peter Williams wrote:
> What I came up with on Friday is following this chain (which seems
> the same as what you describe).
>
> DomainRoot -> J2EEDomain -> J2EECluster (map) -> J2EEServer (map)
>
> It does work, FWIW.
>
> One question -- do you happen to know how much bandwidth a typical
> JMX call consumes? I noticed a substantial delay when accessing
> from a remote workstation where one channel is 192Kb (upstream
> DSL). I will probably just have to run some tests, but if you have
> any advice on minimizing the costs of a call or call aggregation, it
> would be most welcome.