Re: Simple Multicast - building a case for a BOF or WG

[Radia Perlman - Boston Center for Networking <Radia.Perlman@East.Sun.COM>]

>>	From: amc@cs.berkeley.edu (Adam M. Costello)

>>	> You can do multisender with unidirectional trees, by building multipl
e
>>	> trees (one per source) or unicasting to the Root, which creates a
>>	> hotspot at the root...
	
>>	What exactly is meant by "hotspot"?  The number of packets arriving
>>	at the root is the same as the number arriving at every other node in
>>	the tree.  Is there much more work to be done per packet at the root?
>>	If so, is it fundamentally more work, or just more work in current
>>	implementations?
	
In a bidirectional tree you're right. The Root gets exactly the same
number of packets as everyone else. The case I was talking about is a
unidirectional tree, where a sender other than the Root transmits
packets by unicasting to the Root, which must
then decapsulate and send it on the tree. So for example if the Root had
only a single link, then that link would have twice the traffic, since
each packet has to be unicast to the Root on that link, and then multicast
out onto the same link. Also, it's more work to decapsulate and forward
than simply to forward along the tree, so the Root in a unidirectional
tree (with non-Root senders) has more computation burden also.

Radia