Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262728960

The Research on the Reliability Evaluation Systems of Communication


Networks

Conference Paper · August 2001

CITATION READS

1 229

2 authors, including:

Wei Li
CitiGroup
14 PUBLICATIONS   400 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Wei Li on 01 June 2014.

The user has requested enhancement of the downloaded file.


The RESEARCH ON THE RELIABILITY EVALUATION
SYSTEMS OF COMMUNICATION NETWORKS

Li Wei Zeng Sheng Kui


(Systems Engineering Department of Beijing University of Aeronautics and Astronautics)

Keywords: network reliability, reliability evaluation system, performability, performance reliability

Abstract: Firstly this paper summarizes the current reliability evaluation methods of communication network, and
sets forth a new method for evaluating network reliability — evaluation of network performance reliability ,
then a system for network reliability evaluation is set forth also. Secondly, the evaluation system has been used to
analyze the predecessor to Internet, Arpanet. This paper also expounds on the mathematical analysis of network
performability and performance reliability. Though the reliability evaluation system of communication network is
formed in this paper , it is still not a closed form .The future work is about the whole network’s performability
and reliability.

Foreword
Since the 1960’s there has been a rapid rise in both network technology and consumerism,
and there has also been a continuous need for better network performance and reliability.
Negative influence of network failure on society and telecommunication companies is greater
and greater. According to research data,a break in communications for one hour could cause the
loss of 2.5 million dollars to an insurance company, 6 million dollars to a commercial bank. If
communication is stopped for 2 days, a bank could go bankrupt. Whether from the perspective of
satisfying the needs of the society or corporate interests, the issue of network reliability has
become extremely crucial.
The evaluation of network reliability is very special, conventional evaluation systems of
network reliability are no longer adequate to fully describe all the characteristics of network
systems .The need of the consumer is not only to complete a set function(preserve connection)
in a given time, but it is to do the best to complete function(multimedia,large capacity,timely
and so on).Methods that evaluate performance and reliability are no longer adequate to meet the
current demands because of network development .
In the development of the research of network reliability, methods for network reliability
evaluation is basic and important.Different methods for determining reliability of networks
require different means of reliability analyses.To set up a scientific evaluation system,one must
firstly have an accurate understanding of the special characteristics of networks as well as the
needs of the consumer.
Networks are complex huge systems. Relationships between different net components are
complex.Furthermore, the designers and users of the network directly influence the network
function.So,research result of the reliability of ordinary products is not applicable to networks.
Another special characteristic of networks is that they simultaneously serve multiple customers.
As the size of networks and technology level of networks continuously increases,there is a
corresponding continuous change in the needs of the consumers. Traditional research on
reliability of aircrafts, tanks, automobiles and so on are inadequate for reliability research of
network systems. In conclusion, the complex nature of networks, both in network communication
and network consumer’s needs, requires the research and development of new methods for
evaluation of network reliability.
Analysis of Network Reliability Evaluation System
1. Existing Network Reliability Evaluation Methods
The first question in research on network reliability is to find a rational evaluation of
network reliability. From 1960 to the present many evaluation methods are set forth.
(1) 1960's
probability: 1.Network states enumeration method[4 5]; 2.

— The algorithms for S-T connection
[6,7]
Network cut-set method .
— Based on component survival probability, Baron[12] simulated random damages that resulted
in failure of nodes and links using Monte Carlo theory. The percentage of the average
number of nodes of the biggest survived connectional sub-network among the original
network nodes is the link-success probability of the communication network.
— Frank[13] studied the case of symmetric networks. He gave a recursive result through Poisson
distribution to simulate the random damages that the nodes and links suffered. The
link-success probability of the communication network is, as he defined, the percentage of
the number of nodes among the original network nodes , where the former nodes are those
that can connect to any arbitrarily selected node in the survived networks.
— Gilbelt[14] and Kelmans[15] assumed that the nodes in network are absolutely reliable, and all
links have a common survival probability, so that the link-success probability of the
communication network is defined as the probability of the overall communication networks
forming a complete link-success network, and a analytic expression was concluded.
— While O.Frank further developed the achievements of Gilbelt and Kelmans to the cases
when the nodes and links have certain diversified survival probabilities.
(2) 1970's
— Two S-T connection probability algorithms were improved and network paths method was
concluded. [8 9]

— Frank and others[1] raised the concept of network cohesion (CH) and network connectivity
(CN) of link-success networks. In fact, cohesion is related to the minimal links cut-set of
network, whereas connectivity is related to the minimal nodes cut-set of the network.
— Wilkov[2] held that the minimal cut-set of network also depicts the network vulnerability.
— Boesch and Thomas[3] raised the generalized network cohesion and connectivity, i.e., the
minimum links NL(m ) or nodes NN (m ) required to be removed if a subnet of m
nodes is to be separated from a network, that is
CH=min[ NL(m ) ], CN=min[ NN (m ) ]
(While m = 1 , the CH , CN here is identical to the fore-mentioned CH , CN .)
(3) End of 1970's and early 1980's up to date
— O.Frank [16] further developed the achievements of Gilbelt and Kelmans to the cases when
the nodes and links have certain diversified survival probabilities.
— Pan[17] studied the link-success probability of the network with the consideration of
dependency of network components .
— Barberis and others[18 ] presented the weighted S-T connection probability, i.e., averaging the
overall network's S-T connection probability value after it is being weighted by network S-T
information flow. Barberis also raised a new index which is the probability of the network's
throughput exceeding a given threshold value L. Due to the presence nof +l
network
components failure, a network of n nodes and L links can form N = 2 subnets,
each of which corresponds to a failure pattern of network components. Let's assume that the
probability of the pattern's presence is P (S i ) , a indicative function can be defined as:
0, if the throughput of the subnet' s S i ≥ L ;
F (S i ) = 
 1, otherwise;
Then the network performability can be expressed as : A(L ) = ∑ F (S i )P (S i )

A more general expression could be: X = ∑ P (S )X (S )


i∈Ω
r i i

— Where X (S i ) is the network's operation performance index under the condition of S i ,


Pr (S i ) is the presence probability of the network state S i . Ω denotes state space of the
network.
— Park[19] adopted an index based on network transmission delay: probability that the delay of
transmission from a central node to any other nodes isn’t exceeding the threshold. This index
is applicable [20]
to networks with a central node.
— Bonarentura raised two network reliability indexes for circuits switch and packet switch
network respectively.
— K.Li and Silvester[22] proposed ORDER algorithm to evaluate the approximation of
performability. [23]
— Yang and Kubat made improvements to the algorithm of K. Li and Silvester.
2. Further development of network reliability evaluation method
From 1980 to present network performability has become the main issue. The
performability index actually is the expectation of network operation performance.
Perfprmability is related to network’s topology structure, route options, congestion control,
distribution of network tasks and so on. When comparing two networks,firstly consider
reliability and performance,if their reliability and performance are the same, then compare
their performability. In certain conditions some indexes of performability could be the same.
For example, the two networks shown in fig 1 and fig 2 (the numbers near the links outside
brackets denote links’ reliability and the numbers inside brackets denote links’ capability) .
2 2

0.8(4) 0.8(6)
0.8(6) 0.7(2) 0.8(9)

1 0.9(2)
5 3 1 3
0.7(2) 0.4762(3)
5
0.9143(1)
0.7(4) 0.7(5) 0.7(2) 0.7(4)

4 4

Fig1 Topology structure of network 1 Fig2 Topology structure of network 2


The reliability of both networks is 0.8594.Operation performance (the maximum flow) in
both cases is 10. The performability indexes are also the same when work to 13333.3 hours
calculated with the simulation method (fig6). The conventional definition of performability can
not show the difference, a new method for network reliability evaluation is needed .
If the distribution of the two networks’ maximum flow is analyzed at 13333.3 hours as in fig
3 and 4, the probability(approximately 0.3)that the maximum flow of network 1 is more than
6 is larger than that of network two (less than 0.27). In other words, there is a larger possibility
for network 1 to maintain a high level performance at the 13333.3 hour.
probability
probability

Maximum Maximum
flow flow
Fig3 The maximum flow distribution Fig4 The maximum flow distribution
of network 1 of network 2

From this simple case, it is clear that if increase consideration of network operation
performance threshold to the conventional network performability, the evaluation of network
reliability will be better. Network performance reliability is defined as the probability that
network operation performance is greater than a threshold under a given condition at a given time.
The expression is as follow:
PF(t)=P(x≥x0 t)
In this expression, PF(t) denotes the network performance reliability at t time , x denotes the
network performance index , x0 denotes the performance threshold.
3. Network reliability evaluation system
rel. evaluation system
of communication
network
evaluation system

network network rel. network function


topology rel. rel.

network vuln. network surv. network perf. network


performance rel.

all pair all pair all pair all terminal pair-terminal


terminal terminal terminal terminal terminal terminal performance performance
vuln. vuln. surv. surv. perf. perf. rel. rel.

general analysed s-t weighted s-t parameter of


CH CH unknow connection unknow performance
evaluation

and brought connectio


parameter

forward by n probability rel.


general this paper probabili parameter
CN CN ty based on
operation
performance
rel.:reliability vuln.:vulnerability surv.:survivability perf.:performability
Fig 5 The reliability evaluation system of communication network
From the research history of network reliability, it can be seen that evaluation methods of
network reliability could be divided into three categories, that is methods based on network
topology structure,based on network topology structure and random failures, and based on
the integration of network performance and network reliability.
The first category primarily deals with deterministic network. In this category the chart
theory is applied to analyze network reliability and brought forward network vulnerability. The
second category focuses on the network whose nodes and links survive(or fail) with certain
probability. In this category, the chart theory and probability theory are applied to analyze the
network connection probability and brought forward the network survivability. The third
category is a comprehensive research phase based on the network operation performance and
reliability .Performability is the symbol of the beginning of this phase .
Based on the above analysis, this paper proposes an evaluation system for network
reliability (fig 5). But the evaluation index is not in a closed form yet.
This system in fig5 evaluates network from the aspects of topological reliability, traditional
network reliability and network functional reliability. In other words, it evaluates network in
graph theory, combination of graph theory and probability theory and a synthesizing of graph
theory, probability theory and communication theory. In addition to the traditional network
performability, network performance reliability is developed to evaluate network functional
reliability, which perfects the whole evaluation system. In each specific evaluation branches,
there are two types of all-terminals and pair-terminal. For pair-terminal, specific evaluation
parameters are given, of which the performability of pair-terminal is determined by other
probability indexes based on network performance. For all-terminals, the vulnerability evaluation
parameter is an existent research result. The survivability evaluation parameter is raised and
discussed in this paper according to the relevant knowledge of the graph theory. However, the
evaluation parameters about the performability and performance reliability are still to be further
discussed.Therefore, the network reliability has been analyzed and evaluated in various aspects,
on which a relatively complete evaluation system forms.
Algorithm for calculating network performability and performance reliability
Currently there is no analytic way to calculate network performability and performance
reliability .This paper uses Monte Carlo method to calculate it, the Algorithm is as follows:
input data:network's total nodes number(n),
flux matrix(C),start node code,terminal node
code,operation performance threshold,failure
rate matrix.The total No.of the not 0
elements in C is the double No. of links (h)

produce every link and node's random failure


time ti(i=1,2,…,h+n)

order ti(i=1,2,……,h+n) to give the links


and nodes failure sequence

caculate the maximum flow at ti(i=1,2,……,


h+n)
select m simulate time point within a
defined time

compare these m points with ti(i=1,2,……,


h+n) in order to get these points' Max flow

N does the simulate times


reach the given value
Y
caculate the times in every point which the
Max flow is greater than the threshold ,the
percentage of the times in the simulate
times is the performance reliability
parameter

draw the simulate chart

Fig 6 The network performability and performance reliability simulation flow chart
Application example
The predecessor to Internet, SRI UTAH ILL MTI LINC CASE
Arpanet, is selected to elucidate this 0.868(90) 0.91(97) 0.946(102)
evaluation system. Arpanet’s 0.916(99) 0.944(110)

topology structure[21] is shown in


0.932
(100) 0.888
0.93 0.899
fig7. (the numbers near the links (104) (107) (98)

outside brackets denote links’ UCSB STAN SDC


0.892
(110) CARX
reliability and the numbers inside 0.869
(103)
brackets denote links’ capability). 0.92
(105)
0.908 0.973
UCSB node is selected as the 0.971
(108) (94)

input node while CARX is selected (102)

as the output node. Then S-T 0.94(106) 0.979(109) 0.986(109) 0.86(110)

cohesion is 2(the smallest destruct UCLA RAND BBN HARV BURROUGH

links to make node 1 and node 10


disconnected ) , connectivity also is 2 ( the Fig 7 Arpanet’s topology structure
smallest destruct nodes to make node 1 and node
10 disconnected).
Calculation shows that Arpanet has a total of 26 paths. The reliability data of Arpanet’s
nodes are shown in table 1:
Table 1 Reliability value of Arpanet’s nodes

UCSB SRI UTAH ILL MTI LINC CASE STAN SDC CARX UCLA RAND BBN HARV BURROUGH

0.9997 0.9997 0.9998 0.9996 0.9996 0.9994 0.9999 0.9999 0.9998 0.9998 0.9998 0.9995 0.9999 0.9993 0.9997

Without accounting for node failure, the S-T connection probability for Arpanet between
UCSB and CARX is 0.95279. Accounting for node failure, the S-T connection probability
becomes 0.84547.
Next, we analyze Arpanet’s performability and performance reliability. Select the Max flow
between node 1 and node 10 as the network operation performance index. To calculate the Max
flow for Arpanet, a capacity matrix is fixed in table 2.(The element at row i, column j of the
matrix represents the capacity of the link from node i to node j, while the elements on the
diagonal represent capacity of the nodes).
The maximum flow calculated for a connection on Arpanet between UCSB and CARX is
129.Using a simulation method the Arpanet Max flow’ average at different times are designated
as the conventional estimation index of performability. Set the threshold of the Max flow as 20,
Then figure out the probability of Arpanet’s Max flow higher than the threshold at different
times as the index of performance
Table 2 Arpanet’s capability matrix
reliability. In order to make the
simulation result more clear, we 0 100 0 0 0 0 0 0 0 0 102 0 0 0 0
100 0 90 0 0 0 0 104 0 0 103 0 0 0 0
draw the performability curve and 0 90 0 97 0 0 0 0 107 0 0 0 0 0 0
performance reliability curve in fig8, 0 0 97 0 120 0 0 0 0 0 0 0 0 0 0
0 0 0 120 0 99 0 0 0 0 0 0 110 0 0
fig9. 0 0 0 0 99 0 110 0 0 0 0 0 0 0 0
Arpanet has been evaluated 0 0 0 0 0 110 0 0 0 98 0 0 0 0 0
0 104 0 0 0 0 0 0 0 0 0 105 0 0 0
through four aspects, namely, 0 0 107 0 0 0 0 0 0 0 0 108 0 0 0
vulnerability, survivability, 0 0 0 0 0 0 98 0 0 0 0 0 0 0 94
performability and performance 102 0
103 0
0 0
0
0
0
0
0
0
0 0 0 0 0 106 0
0 105 108 0 106 0 109 0
0 0
0
reliability. So that we could have a 0 0 0 0 110 0 0 0 0 0 0 109 0 109 0
profound conception toward the 0 0 0 0 0 0 0 0 0 0 0 0 109 0 110
0 0 0 0 0 0 0 0 0 94 0 0 0 110 0
reliability of the Arpanet under the
conditions of intentional/random damages, as well as its operation performance variation and its
ability to suffice communication requirements under the occasion of unavailability of network
components.

Fig 8 Arpanet’s performability Fig 9 Arpanet’s performance reliability


Conclusion
This paper has summarized the research history of network reliability and analyzed many
different methods for evaluating network reliability .A new evaluation method that combines
network performability and network performance reliability is proposed. Furthermore, a network
reliability evaluation system is formed with addition of conventional network invulnerability and
survivability. Next, the predecessor of Internet, Arpanet, is analyzed with this network evaluation
system .The results shows that this network reliability evaluation system could yield a more
complete evaluation result in the evaluation of S-T network reliability.
However, a network is a system that comprises many nodes and links. The most concern of
consumers is the overall reliability, operation performance, availability, and performance
reliability of the whole network. What kind of estimation index is most suitable for overall
network performability? How is it to be calculated? How is overall network performance
reliability to be defined? These are all questions that will be answered upon further research.

Reference
[1]H.Frank and I.T.Frisch, Analysis and Design of Computer Network, IEEEE Trans. on Communication
Technology,Vol.COM-18,No.5,1970.
[2]R.S.Wilkov,Design of Computer Nets.Based on a New Reliability Measure,Proc.Symp.on Computer
Communs.Nets .and Teletraffic,New York ,1972.
[3]F.T.Boesch and R.E.Thomas,On Graphs of Invulnerable Communication Nets,IEEE
Trans.On Circuit Theory vol.CT-17,No.2,1970.
[4]O.Wing and P.Demetriou,Analysis of Probabilistic Networks , IEEE Trans .on Communications
Technology,Vol.COM-12,No.3, 1964.
[5]L.Fratta and U.G.Montanari , A Recursive Method Based on Case Analysis for Computing
Network Terminal Reliability,IEEE Trans.on Communications, Vol.COM-26,No.8, 1978.
[6]J.K.Cavers,Cutset Manipulations for Commun.Net.Reliability Estimation IEEE Trans.on
Communications,Vol.COM-23,No.6,1975.
[7]P.A.Jensen.An Algorithm to Determine the Reliabity of a Complex System IEEE Trans, on
Reliability,Vol.R-18,No.4, 1969.
[8]L.Fratta and U.G.Montanari , A Boolean Algebra Method for Computing the Terminal Reliability
in a Communication Network,IEEE Trans.on Circuit Theory Vol.CT-20 , No.3,1972.
[9]K.K.Aggarwal,A Fast Algorithm for Reliability Evaluation,IEEE Trans.on Reliability,Vol.R-24,
No.1 , 1975.
[10]E.Hansler , A Fast Recursive Algorithm to Calculate the Reliability of a Communication Network,
IEEE Trans.on Communications , Vol.COM-20,No.3, 1972.
[11]P.M.Lin,A New Algorithm for Symbolic System Analysis,IEEE Trans.on Reliability,Vol.R-25,
No.1,1976.
[12]P.Baran,On Distributed Communication Networks,IEEE Trans. on Communications Systems,
Vol.CS-12,No.1,1964.
[13]H.Frank,Vulnerability of Communication Networks,IEEE Trans.on Communication Technology,
Vol.COM-15,No.6,1967.
[14]E.N.Gilbert,Random Graphs Ann.Of Math.Statistics,Vol.30,N o.4,1959.
[15]A.K.Kelmans , Connectivity of Probabilistic Networks,Automatic Remote Control ,No 3,1967.
[16]O.Frank,on Reliability in Stochastic Graphs,Network, Vol.12,No.2,1982.
[17]S.N.Pan,Dependent Failure Reliability Models for Tactical Communs.Networks.Proc.
IEEE ICC’83,Boston,MA,Vol.1,Session-C5.5,June 1983.
[18]G.Barberis and U.Mazzei,Traffic-Based Criteria for Reliability and Availability Analysis of Computer
Networks,Proc.IEEE ICC’77,Chicago,IL,Vol.1,Session-22.6,June 1977.
[19]Y.J.Park and S.Tanaka,Reliability Evaluation of a Network with Delay,IEEE Trans.on Reliability,
Vol R-31,No 4,1979.
[20]V.Bonaventura, Service Availability of Communication Networks ,Proc.IEEE NTC’80,Houston,TX ,
Vol.1,Session -15.2,Nov.1980.
[21] Dr .Sujata,Order-P:An Algorithm to Order Network Partitionings,IEEE Trans.on Reliability, Vol 43,No
2, June 1994.
[22]V.O.K.Li & J.A.Silvester ,Performance Analysis of Network with Unreliable Components,IEEE
Trans.Comm.,Vol.COM -32,No.10,1984.
[23]C.L.Yang & P.Kubat , Efficient Computation Of Most Probable States for Communication Networks with
Multimode Components ,IEEE Trans.Comm.,Vol.COM -37,No.5,1989.
[24]Liu Youheng ,The new progress of the research on communication network’s reliability, proceeding of the
Chinese communication academy,1995,P172-P176.
[25]Xiong Wei ming, The research on communication network’s reliability, Ph. Dissertation, 1990.
[26] Xiong Qingxu, The analysis and design of communication network’s reliability, Ph. Dissertation,1993.

View publication stats

You might also like