Information Technology Reference
In-Depth Information
5.
N
UMERICAL
R
ESULTS
In this section, we compare the performance of the multi-passive-combiner hub with the
AWG hub under different traffic patterns and loads (we remind that the AWG hub has the
best performance in the literature). We consider a metro WDM network which connects three
local networks to the backbone network. Node 4 has a heavier traffic to/from other nodes
because it transports the traffic between all the three local networks and the backbone
network (e.g., see Figure 1 with
N
=4). Let the packet arrival process be Poisson and the
packet arrival rate
i
α
from node
i
to node
j
be:
0
i
=
j
⎪
(9)
αα
α
=
i
≠ ≠
j
4
and (
⎨
⎪
ij
f
×
i
≠
j
i
=
4 or
j
=
4)
⎩
where the parameters
α
and
f
determine the traffic load and traffic pattern respectively. If
f
=1, the traffic is uniform; if
f
is larger, the traffic is more non-uniform. Let the packet size be
exponentially distributed, the mean packet transmission time be 1 μs, the packet delay
requirement
*
Delay
for the multi-passive-combiner hub be 2 μs, and
W
=4.
Figure 9(a) shows the mean packet delay under uniform traffic. We see that the multi-
passive-combiner hub gives smaller mean delay especially when the traffic load is heavy. It is
because the multi-passive-combiner hub can be suitably scaled to provide enough channels to
fit the traffic requirements. More specifically, when the traffic load increases, the mean delay
of the multi-passive-combiner hub increases until it reaches the given delay requirement
*
Delay
=
2
μ
. At this point, when the traffic load further increases, one additional passive
*
=
so that the mean delay decreases.
As a result, the mean packet delay is always smaller than the given requirement
*
Delay
2
combiner will be used to avoid exceeding
Delay
=
2s
μ
.
(a) (b)
Figure 9. (continued)
(c)