Network Survivability
Yishi Li, Matt Willis
(Mentor: Svetlana Poroseva)
Summer 2005 Research Experience for
Undergraduates at Florida State University
School of Computational Science, Florida State
University, Tallahassee, FL, USA
Definitions of Network Survivability
Network Topology – This is a set up in which a given node has
one or more links to others, and they can appear in a variety of
different shapes. Topologies consist of generators (a computer,
for example), vertical edges (possibly a wire connecting to the
rest of the network), and horizontal edges that serve to carry
information and reinforce the structure of the topology.
Definitions
Survivability. The goal of this
project is to determine how
network topologies react while
undergoing multiple failures
simultaneously. Systems
respond differently when
different faulty scenarios
occur. The below figure
illustrates three primary
responses that a network
would experience.
Probability of Selected Scenarios
Occurs (Denotations)
Denotations: m represents the number of faults in a topology
S is the number of possible fault scenarios
SN is the number of N-scenarios,
SF of F-scenarios,
SR1 of R-scenarios in which at least one generator is destroyed
SR2 is the number of R-scenarios with all generators intact.
SN
P( SN )
S
SR
P( SR )
S
SF
P( SF )
S
Few Simple Topologies I
The Ring:
m S SN SR1 SR2 SF
1 6 4 2 0 0
2 15 5 6 0 4
3 20 0 6 2 12
4 15 0 2 0 13
5 6 0 0 0 6
Few Simple Topologies II
The Single Bus:
m S SN SR1 SR2
1 7 3 2 2
2 21 0 8 8
3 35 0 8 2
4 35 0 2 0
5 21 0 0 0
6 7 0 0 0
The Baseball Diamond
blue represents the chance of
Number of faults occurring simultaneously
failure P(F), green represents
P(R), and red P(N). The x-axis
represents the number of
faults occurring at a given
time.
The Double Bus
blue represents the chance of
failure P(F), green represents
P(R), and red P(N). The x-axis Number of faults occurring simultaneously
represents the number of
faults occurring at a given
time.
Results Analysis
This graph shows the chance of failure in different topologies
with two generators, two vertical edges, and a varying number
of horizontal edges while undergoing two faults.
Similar improvements can
be seen when vertical edges
are added. The next slide
shows the chance of failure
at two faults in topologies
with two generators and two
horizontal edges.
Results Analysis Continued
By adding just one vertical
edge, the chance of failure
is less than half of what it
was previously!!
Computational Network
Survivability
Computational Network Survivability is the use of
computers to generate results sufficiently to predict the
outcomes of network survivability at any given scenario.
Program Structure
Computational Results (Double Bus)
3 Generators
Computational Results (Double Bus)
4 Generators
Findings and Discoveries
P(N) vs. M ( 3 P(N) vs. M (4
G) G)
By looking at two graphs, one can see the similarity
between them. Let’s look at the next one.
Findings and Discoveries
The plot illustrates how P(N)
varies depending on the number
of generators. It is seen that the
difference in P(N) is small for a
small number of faults. As M
grows, the effect of the number of
generators on P(N) becomes
more pronounced.
Conclusions
Potential of this Research. In the future, network will
become more and more essential to our lives. In order
to design a reliable network, it is extremely important
to understand the network performance under various
conditions. Since networks are a very complex
system, by combining computational science and
network survivability theory, we will significantly
improve our capability for analysing various forms of
network topologies.
Final Discussion
New Generation of Network Design. Due to a high
demand for reliable communication and power
network systems, it is crucial to develop a highly
survivable network that can sustain catastrophic
events. Continuous research on this topic would
enhance our understanding on survivability and
reliability performance in different configuration. It
provides essential reference for network engineers to
develop systematic schemes in designing a highly
reliable network.
