-I- First, a basic packet (BP) with a size of maximum transfer unit (MTU) is used a
benchmark. The BP is a packet which has an IP header, with one source and one
destination, and its payload.
-I- Secondly, Multicast packets (MPs) are always of a bigger size than a BP due to the
fact that there might be a multicast group address or a protocol header in the multicast
packets. However, the control messages (CMs) (in SEM and SReM) are usually of a
size of less than MTU (for SEM) or far less than MTU (for SReM). For simplicity,
we treat them as multicast packets with a size of MTU.
-I- Thirdly, assuming that the processing cost to be 1 ( unit ) if a packet with a size of
BP is needed to be processed at a node and the delivering cost of this packet to be 1
as well if it is delivered via a link with a cost of 1.
Depending on the previous assumptions, the processing cost at node k is given by
S
Pcos t (Nk ) = MkMP × — + M , (1)
S BP
where M and M are the number of destinations in multicast packets and that in
M k , MP k,CM
control messages at node k, S MP and S BP are the size of a multicast packet and the size
of a BP (i.e. MTU), respectively. In this simulation, the values of MTU fixed at 576 octets,
i.e., the minimum size of packets required for routers. When the size of multicast packets is
over MTU, a fragment processing for packets is needed, which might happen for Xcast
multicast packets, but not being considered in our simulation.
Correspondingly, the delivering cost for one multicast packet over a link i is given by
S
D cos t ( Li) = Li × ( -M^ + J1Cm ), (2)
S BP
WhereJ i,CM and L i are the number of destinations in control messages and the cost over
a link I, respectively.
Based on previous discussion, the average processing cost ( p cos t ) at a node and average
delivering cost ( Dra t ) over a link are given by,
P c0s t = ɪ ∑ Pcost ( Nt ), (3)
cost t
N t∈T
84