When presenting the different classes of networks, people often start by presenting the random model to

a network first, not because most networks in the world are random, quite the contrary, but it is first by
understanding what a network created without any specific rules. Looks like that we can then compare
the other networks we encounter around us to see if they differ from this. And if they do, then we can ask
about the rules that created them. A random network is more formally termed the Irda Freney random
Grauwe model, so named after two mathematicians who first introduced a set of models to random
graphs in the mid 20th century.

As the name implies, this type of network is generated by simply taking a set of nodes and randomly
placing links between them with some given probability. So we just take two nodes in the network and we
roll the dice to see if there will be a connection between them or not. The higher we set our probability,
the more likely there will be a connection and thus the more connected our overall graph will be.

So this is a simple system in that once you have decided how many nodes there will be, it is then really
just defined by a single parameter. That is the probability parameter for the likelihood that any two nodes
will form a connection. So if we looked at the degree distribution of this network, it would follow what is
called a normal distribution because it was randomly generated, there will be some difference in the
degree distribution of connectivity amongst the nodes. Some will have one degree, some five.

But there will be a well defined, normal or average degree in this distribution. There will be very few
nodes with a very large degree and very few with a very low degree. Most will tend towards the normal
amount of connections. Unlike real world networks, there is low clustering in random networks, therefore,
the resulting network very rarely contains highly connected nodes. Consequently, the random network is
not a good candidate model for the highly connected architecture that characterizes the many networks
we see around us.

Although a useful theoretical exercise, random networks in general do not represent networks in the real
world. They are considered far more random because real world networks are typically created to serve
some function and are constrained by some limiting resource that gives them a more distinct pattern. If
we look at some network, like the traditional trading routes across the Sahara Desert in Africa, it may look
somewhat random at first glance. But we know it is not because for the caravan of camels and traders
who created these networks, setting out across the Sahara in any random direction would, of course,
been fatal to them.

This helps to illustrate the two key factors to generating any given structure to a network that we've
previously only touched upon. Firstly, we have the context or environmental constraints that the network
is under different types of networks and the different types of environmental constraints. For example, this
may be the geological constraints placed upon the travelers in our example above. It may be the
physiological constraints placed upon the metabolic networks within our body, or it may be the financial
constraints placed upon a logistics network.
All of these represent resistance to network formation that the environment places upon the network, but
inversely, we could look at this the other way round, asking what methods the nodes in the network used
to overcome these constraints are. Travelers in the Sahara were using prior knowledge encoded in maps
as to where the water wells were in order to overcome the arid environmental conditions placed upon
them. A national airline, because of the limitations on finance, may not be able to run a direct route
between every city within a country.

But we'll get around this by creating a hub and spoke network so as to reach all locations. Again, this is a
method or strategy for overcoming resource constraints and out of the interaction between these
environmental constraints and the methods used by the nodes to overcome them, we will get a particular
overall structure to the network that makes these real world networks distinctly different from our random
network model. The first type of network model we'll discuss is called a distributed network. This type of
model is in many ways quite similar to our random network.

It is defined by low level to the degree distribution of the network, meaning all or most of the nodes have
the same degree of connectivity as there are no dominant nodes to provide global functions for the entire
network. Each node must contribute equally to the networks maintenance. And as there is no real global
coordination in a distributed network, nodes have a much higher degree of autonomy as they are largely
self-sufficient and independent from nodes outside of their neighborhood. An example of a distributed
network might be a community alert group, where each member to the community has equal
responsibility and authority to act.

When there is an event that others should know about, there is no hierarchy. And in this example, the
network is only actualised when needed, thus placing very limited constraints on its members within the
world of computing. Distributed networks are also called mesh networks. These distributed networks have
a number of advantages and disadvantages. On the positive side, as we'll be discussing later, they may
be very robust to failure as there is no critical or strategic nodes in the network.

Any node can theoretically be replaced by any other. And as we've already noted, nodes may have a high
degree of autonomy with little network maintenance tags placed upon them. But also this type of network
can be less efficient in many circumstances. Without centralized nodes, they cannot be any centralized
batch processing that leverages economic scale and diffusion across the network can be slow as there
are no central hubs with which to reach many nodes in a single hub. This can also create problems in
terms of coordinating the network as a whole.

In many ways, a distributed network represents a system in a fine balance and relatively stable state. And
this is often not what we see when we look at real world networks in the next module will be turning up
our degree distribution parameter to the point where small hubs emerge in what is called a decentralized
network.