| description |
In this paper, an analytical bandwidth evaluation of generic
tree-based reliable multicast protocols is presented. Our analysis
is based on a realistic system model, including data packet and
control packet loss, asynchronous local clocks and imperfect
scope-limited local groups. Two of the four considered protocol
classes use aggregated acknowledgments (AAKs). With AAKs they are
able to provide reliability even in case of node failures. All
tree-based protocols provide good scalability for large receiver
groups. Relating to protocols with aggregated acknowledgments, the
analysis shows only little additional bandwidth overhead and
therefore high throughput rates. Finally, we have analyzed the
influence of the branching factor on a protocol's performance.
Our results show that the optimal branching factor depends mainly on
the probability for receiving messages from other local groups. If
local groups are assigned to a separate multicast address, the
optimal branching factor is 2. On the other hand, if TTL scoping is
used as in RMTP or TMTP and therefore the probability for receiving
messages from other local groups is greater than zero, larger local
groups provide better performance.
|