>> In a 10 year old test on old hardware (PIII, 100Mbps LAN), Spread

needed

>> 2-3 ms to totally order messages between 20 nodes. Note that this
>> includes dealing with failure cases, monitoring group membership

changes

>> and guaranteeing virtual synchrony.
>>
>> What throughputs are you achieving in the current builds of Nimbus? On
>> what hardware? LAN/WAN? Is that code also dealing with failure cases?
>
> I have to duck those questions. Sorry.

That's fine. Perhaps early next year we can compare notes on actual
performance.

Paul