The network topology gives a good sign on how well
a network will cope with traffic flows. As we have seen,
token ring provides excellent sharing and prioritization
on a network, whereas in a bus system, nodes contend to
get access to the network. Networks, though, are typically
inter-connected with other networks to give an overall network
architecture. This defines how the networks are interconnected,
and the routes that the data will take to get from a source
to a destination. It is important that the architecture
is modelled, as this can highlight problems with bottlenecks
in the data flow, and any potential problems for the future.
The Napier network provides an excellent example of this,
as it receives peaks in data throughout the day, which varies
from month to month. Thus, it is important to understand
the flow of the data, and identify if routes are over-burdened,
or even if they are under-burdened. An over-burdened route
can be eased by providing alternative paths around the route.
Another important factor of network modelling is to provide
an indication of how the network architecture responds to
faults, and if the network will cope with the fault.
It is possible to run a complex model of the network, but
the worst-case situation can often be modelled using simple
methods. One way to do this is to estimate the peak flows
that occur from sources to destinations, and estimate the
routes that they take. For example if we have networks A,
B, C, D, E and F, which are interconnected with data points
of 1, 2, 3, 4, 5, 6 and 7, then we can draw the network
architecture given in Figure Ch3.1.
Figure Ch3.1: Example network architecture
If we modelled the network we could estimate the data streams,
such as:
Source
Destination
Data
flow (Mbps)
1
7
10
2
5
1
1
4
5
5
7
4
3
6
10
Next we can model the data streams across each of the known
routes:
A
B
C
D
E
F
1®7
10
10
10
10
2®5
1
1
1
1®4
5
5
5
5®7
4
4
4
4
3®6
10
Total
15
16
10
5
24
14
It can be seen from this that the heaviest data flow is
on Network E, which has a flow of 24Mbps. If this flow was
too high for this network, we could create a new route in
parallel with Network E, and share the data flow. This is
known as load sharing, and an example is shown in Figure
Ch3.2. With this the router sends data through Network E
for half the time, and through Network G for the other half
of the time.
Figure Ch3.2: Example network
architecture with loading sharing
Thus in the network architecture in Figure
Ch3.1, loading sharing can be applied between Network E
and Network G, this will cause 12Mbps to flow through Network
E, and the same for Network G. This will thus reduce the
loading on Network E.
We can then determine the loading on
each network, but determining the maximum bandwidth of each
network, and then using:
Thus, using the previous example, if the bandwidth of Network
A, Network B and Network E is 40Mbps, and the bandwidth
of the other networks is 20Mbps then:
Network
Bandwidth
(Mbps)
Actual
Flow
Utilization
(%)
A
40
15
37.5
B
40
15
37.5
C
20
10
50.0
D
20
5
25.0
E
40
24
60.0
F
20
14
70.0
Now we can see that the most utilized
network route is actually Network F, with a 70% utilization,
thus it may be better to load share on this network, rather
that Network E, as it only has a 60% utilization.
Challenge 2
The network architecture in Figure Ch3.3
has the following data flows:
Source
Destination
Data
flow (Mbps)
1
4
5
1
3
3
2
5
7
2
6
2
3
7
3
3
4
1
4
5
10
6
7
2
Challenge A Determine the network(s)
with the largest data flow?
Challenge B If the maximum data flow for networks
A, C, E and F is 20Mbps, and Networks B and D are 40Mbps.
Determine the most utilized network and the least utilized
network route?
