Integrated Reliability Analysis Tool

Submitted by:

Mike Sjulin, Cindy Phillips, Elmer Collins, Bob Hutchinson 

Sandia National Laboratories
M. S. 0451
P. O. Box 5800
Albuquerque, NM 87185-0451

(505) 844-5012 voice (Mike Sjulin)
(505) 844-9641 fax

mrsjuli@sandia.gov


---------------------------------------------------------------------------- ------

As the size and complexity of communication networks grow, designers of 
these
networks are increasingly dependent on tools which aid in the design. Some 
significant
factors that influence the design of these networks include: functionality, 
performance,
reliability, and cost. A designer must balance these factors in a sound 
design. Current
network design tools focus on functionality and performance. The design 
engineer does
not have a direct indication of reliability or long term cost through these 
tools.

The proposal is to develop an automated tool that assists the network 
designer in
evaluating the system reliability of a communications network. This tool 
will be
integrated into an existing network design package. Primitive data 
structures of the
existing tool will be transformed into a reliability model. This reliability 
model in
concert with a reliability data base will be used to evaluate the expected 
system
reliability based on the topology and the content of the network 

The automated tool is envisioned to consist of three parts: a translator, a 
reliability data
base, and an analysis engine. The translator will be capable of extracting 
information
about network topology, component attributes, and communication 
protocols directly
from the design tool . This information and a precise definition of system 
success entered
by the user will be used to generate a reliability representation of the 
network. The
reliability data base will contain field failure information, number of 
failures per million
hours of operational usage, and the applicable probability distribution 
function for most
commercially-available communication network components. Users will 
have the option
to use this data or enter their own failure data. If a component is not in the 
data base, the
user will be provided guidance in determining a reliability metric, i. e., 
Mean Time
Between Failure (MTBF), for that component. The analysis engine uses 
the reliability representation from the translator and the reliability data 
base to evaluate
the model and provide a reliability assessment of the system. 

A communication network can be abstracted as a graph where the nodes 
represent hubs or
termination points (routers, gateways, switches, users, etc.) an edge between 
two nodes
exists if there is a direct communication channel between them 

The analysis engine is envisioned to be developed in different stages. The 
first stage will
consider all-terminal reliability. The input is a graph where the nodes are 
assumed to be
completely reliable and each edge e is assigned a value p_e, which 
represents the
probability that edge e will fail in a given time interval. Given that all 
edges fail
independently with their given probability, the challenge is to compute the 
probability
that viable (unfailed) paths exist between every pair of nodes at the end of 
the time
interval. The goal will be to provide the network designer with real time 
reliability data
for interactive design. The second stage will add channel capacity and node 
failure. A
multi-commodity flow model will be incorporated into a Monte Carlo 
simulation to
estimate the reliability versus channel capacity distribution. 

Analysis results should be integrated into the existing network design tool 
display. This
will provide the user with an integrated design tool. Reliability data along 
with repair
cost data can be used to estimate long term system cost.