27
This of course does not mean that crackers could
not cause significant damage to the operation of
the Internet. If the greatest centre of the network
is targeted huge damage can be caused. In summary: technical errors cannot damage scale-free
networks but their protection is poor against
attacks. Similar features were observed in living
systems. Paine (1969) introduced the term key
species. According to his idea, if in an ecological
system the species with most connections became
extinct in a food chain the system will collapse
astonishingly soon and a mass of species perish.
In nature protection therefore special care have to
be taken in studying such key species.
In networks
topology, fault tolerance and vulnerability cannot
be separated completely from each other. Every
complex system has its own Achilles’ heel
(Barabási 2003, p. 167).
It was proved, however, both in science and practice that centralised and even decentralised networks are vulnerable, best protection is shown by
divided networks (Fig. 2.9c). In the case of the
latter, being a grid like network, alternative routes
maintain connections in the remaining parts of a
network if some of the points were removed.
Networks—either random or scale-free—can
be classified as controlled and uncontrolled. In
controlled networks free “travel” (sending information) among network points is not possible.
For example, World Wide Web is a controlled network. According to Barabási (2003) controlled
networks can be divided naturally into several
“continents” (Fig. 2.10). In the core any points
can be reached from any other point located in the
core. It is possible to get into the core from IN
content but there is no way back. The points of
OUT continent can be reached only from the core
but there is no link towards the core. (Company
webpages are similar to the above.)
Apart from continents islands (isolated groups
of webpages connected to each other) “dead end”
tendrils connected to IN and OUT continents and
pipes connecting directly the IN and OUT continents also exist in controlled networks. According
to Barabási, around one-quarter of the total number of documents on the Internet are located on
either islands or tendrils. This structure of the
Internet results in the fact that around half of the
homepages cannot be reached even from the
core: the IN continent and the islands cannot be
reached from the core either.
2.5
System Models and Model
Making
Numerous characteristics of the systems are suitable for scientific study. A common feature of the
study methods is that the structure or the operation of the studied system is idealised, simplified
so that the model reflects the conditions of the
studied system as accurately as possible.
Depending on the aim of the study this simplification can be of high grade while in other cases a
detailed model is formed in order to illustrate
small details of the system as well.
Fig. 2.9 Basic types of
networks according to
Baran (1964) (A)
centralised, (B)
decentralised, (C)
distributed
2.5 System Models and Model Making
This of course does not mean that crackers could
not cause significant damage to the operation of
the Internet. If the greatest centre of the network
is targeted huge damage can be caused. In summary: technical errors cannot damage scale-free
networks but their protection is poor against
attacks. Similar features were observed in living
systems. Paine (1969) introduced the term key
species. According to his idea, if in an ecological
system the species with most connections became
extinct in a food chain the system will collapse
astonishingly soon and a mass of species perish.
In nature protection therefore special care have to
be taken in studying such key species.
In networks
topology, fault tolerance and vulnerability cannot
be separated completely from each other. Every
complex system has its own Achilles’ heel
(Barabási 2003, p. 167).
It was proved, however, both in science and practice that centralised and even decentralised networks are vulnerable, best protection is shown by
divided networks (Fig. 2.9c). In the case of the
latter, being a grid like network, alternative routes
maintain connections in the remaining parts of a
network if some of the points were removed.
Networks—either random or scale-free—can
be classified as controlled and uncontrolled. In
controlled networks free “travel” (sending information) among network points is not possible.
For example, World Wide Web is a controlled network. According to Barabási (2003) controlled
networks can be divided naturally into several
“continents” (Fig. 2.10). In the core any points
can be reached from any other point located in the
core. It is possible to get into the core from IN
content but there is no way back. The points of
OUT continent can be reached only from the core
but there is no link towards the core. (Company
webpages are similar to the above.)
Apart from continents islands (isolated groups
of webpages connected to each other) “dead end”
tendrils connected to IN and OUT continents and
pipes connecting directly the IN and OUT continents also exist in controlled networks. According
to Barabási, around one-quarter of the total number of documents on the Internet are located on
either islands or tendrils. This structure of the
Internet results in the fact that around half of the
homepages cannot be reached even from the
core: the IN continent and the islands cannot be
reached from the core either.
2.5
System Models and Model
Making
Numerous characteristics of the systems are suitable for scientific study. A common feature of the
study methods is that the structure or the operation of the studied system is idealised, simplified
so that the model reflects the conditions of the
studied system as accurately as possible.
Depending on the aim of the study this simplification can be of high grade while in other cases a
detailed model is formed in order to illustrate
small details of the system as well.
Fig. 2.9 Basic types of
networks according to
Baran (1964) (A)
centralised, (B)
decentralised, (C)
distributed
2.5 System Models and Model Making