[AG-TECH] Re: Numerator vs Denominator

[Chris Greenhalgh]

> While true, I still say that the automatic scaling of beacon bandwidth is:
> 
>   - intractable in practice because not all sources hear each other
>     due to lack of uniform reachability

certainly the lack of uniform reachability makes life 
harder, requiring secondary 
control mechanisms, e.g. to cope with sites that are not 
receiving at all could use explicit unicast source quench 
messages from other beacon processes with some
implosion control mechanism, to cope with bad ttl imballances
some broader gossiping of observed bandwidths
 
>   - unnecessary in the context of O(100) beacons supporting Access Grid
>     (1000 pps is roughly equivalent to one Access Grid session; if this
>     is causing problems for networks somewhere then those networks need
>     to be repaired to support Access Grid nodes)

true as far as it goes; but how long will it stay O(100)? 
painful to be a victim of ones own success...

at the very least one could have a two-tier system
with a more restricted deployment of something like
the current beacon system in the 'core', and a less
demanding second tier liveness monitoring system,
e.g. to one or a hand full of 'nearby' 1st tier beacons -
it would seem useful to have some of the same
confidence/monitoring tools available for very small
sites, even if they will only be able to join very small
AG type sessions.

Chris