ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.
1, ISSUE: 04
PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI
AND WAXMAN TOPOLOGY GENERATION MODEL
R. Ravi
Department of Computer Science and Engineering, Francis Xavier Engineering College, Tamil Nadu, India
E-mail: fxhodcse@gmail.com
Abstract
As internet usage grows exponentially, network security issues
become increasingly important. Network security measures are
needed to protect data during transmission. Various security controls
are used to prevent the access of hackers in networks. They are
firewall, virtual private networks and encryption algorithms. Out of
these, the virtual private network plays a vital role in preventing
hackers from accessing the networks. A Virtual Private Network
(VPN) provides end users with a way to privately access information
on their network over a public network infrastructure such as the
internet. Using a technique called “Tunneling”, data packets are
transmitted across a public routed network, such as the internet that
simulates a point-to-point connection. Virtual private networks
provide customers with a secure and low-cost communication
environment. The basic structure of the virtual circuit is to create a
logical path from the source port to the destination port. This path
may incorporate many hops between routers for the formation of the
circuit. The final, logical path or virtual circuit acts in the same way
as a direct connection between the two ports. Our proposed
Provisioning Restorable Virtual Private Networks Algorithm (PRA) is
used to combine the provisioning and restoration algorithms to
achieve better results than the ones obtained by independent
restoration and provisioning. In order to ensure service quality and
availability in Virtual Private Networks, seamless recovery from
failures is essential. The quality of service of the Virtual Private
Networks is also improved due to the combination of provisioning and
restoration. The bandwidth sharing concept is also applied in link to
improve the quality of service in the Virtual Private Network. The
performance analysis of the proposed algorithm is carried out in
terms of cost, the number of nodes, the number of VPN nodes, delay,
asymmetric ratio and delay with constraints with Disjoint Path
Algorithm and Approximation Restoration Virtual Private Networks
Algorithm. The Provisioning Restorable Virtual Private Networks
Algorithm performs better than the Disjoint Path Algorithm.
Keywords:
Virtual Private Network, Restoration, Provisioning
1. INTRODUCTION
Better quality of service can be achieved by combining the
provisioning and restoration algorithms. In order to meet the
bandwidth requirements specified by customers, the Network
Service Provider needs to reserve in a restorable VPN, a
sufficient amount of bandwidth on data transmission paths
between each endpoint pairs [2]-[5]. The primary bandwidth is
the one needed on the paths under the non-failure case. The
additional bandwidth needed on the alternative paths under the
link failure case is called protected bandwidth [1],[7]-[10]. Here,
we proposed to combine the provisioning and restoration
algorithms to achieve better results than the ones obtained by
independent restoration and provisioning methods.
191
1.1 RESTORABLE VPN
End-to-end restoration provides protection on disjoined paths
from the source to the destination and may rely on fault
signaling to effect recovery switching at the source [6]. Local
restoration for its part effects protection switching at the
upstream node from the point of failure, the point of local repair
and do not require fault signaling. The local restoration is not
efficient in capacity. Path protection, on the other hand, can
optimize capacity allocation on an end-to-end basis [11]-[15].
Our research work investigates the problem of provisioning
restorable virtual private networks algorithm paths that satisfy
delay constraints.
1.2 BANDWIDTH SHARING IN RESTORABLE VPN
In general, bandwidth sharing reduces the total reservation
cost of the virtual private networks. Identifying the optimal
bandwidth reservation needs for supporting the different traffic
is important for both; from the applying customer’s perspective,
it is necessary to minimize the bandwidth reservation cost, and
for a network service provider to optimize the network
utilization [16],[17]. Therefore, the bandwidth reservation, both
in terms of the cost of reservation as well as network utilization,
must be made in such a way that the unused bandwidth during
non-peak periods can be utilized by the best effort data traffic
[18]-[20].
2. PROVISIONING RESTORABLE VIRTUAL
PRIVATE NETWORKS ALGORITHM (PRA)
Input: The topology graph, VPN nodes and ingress/egress
bandwidth.
Output: Provisioned Restorable Virtual Private Networks
Algorithm Tree.
Step 1: Get the number of nodes, VPN nodes and construct the
topology using our program.
Step 2: Fix 10 to 50% of the boundary region nodes as VPN
nodes.
Step 3: Assign the ingress/egress bandwidth for each of the
VPN nodes and assign bandwidth and delay to the
edges in the topology graph.
Step 4: Get the source and destination VPN nodes from the
topology constructed, using our program.
Step 5: Find all the possible shortest paths from the source
VPN node to the destination VPN nodes using the K-
Optimized traffic lane algorithm.
R. RAVI: PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI AND WAXMAN TOPOLOGY GENERATION MODEL
Step 5.1 Find all the shortest paths from the source to
other VPN nodes using the dijkstra algorithm
and add it to the path set.
Step 5.2 For each path in the path set do the following
steps
a. Take the first path from the path set.
b. Take the least cost link in the selected path
and add to the least cost link.
c. Remove the least cost link from the
network and find the new shortest path
using the dijkstra algorithm.
d. If the new path exists then add the path to
the path set.
e. Repeat steps b to d for each least cost link.
f. Reinsert the removed link into the network.
Step 5.3 Repeat the step 5.2 for each path in the path set
until all the paths in the path set have been
processed.
Step 6: Construct a BFS Tree by combining any two paths from
the shortest path list randomly.
Step 7: Repeat step 6 for a different possible combination of the
shortest paths.
Step 8: Construct the topology graph using the different
combination of shortest paths.
Step 9: Get the random value of the delay constraint.
Step 10: Determine the primary path from the source to the
destination.
Step 10.1 Assign the bandwidth and propagation delay
values from the source to all the nodes.
Step 10.2 Find the minimum delay value from the
source to all the adjacent nodes.
Step 10.3 Set the flag value for the visited node in the
network.
Step 10.4 Find the minimum delay of the adjacent
nodes in the network.
Step 11: Step 10 repeated until the destination node is reached.
Step 12: Display the primary quality of service path from the
given source to destination node.
Step 13: Display the minimum propagation delay of the primary
quality of service path that satisfies the delay constraint
with an optimal cost.
Step 14: If any one of the link fails in the primary quality of
service path, use the path restoration for finding the
alternative path.
Step 15: Find a restoration path from the source to the
destination.
Step 15.1 Construct the auxiliary directed graph, by
reversing the links that belong to the primary
quality of service path.
Step 15.2 Assign the delay value as zero for reversing
the link and substitute the bandwidth of each
link of the auxiliary graph.
Step 16: Use the adjustment delay concept for finding the
minimum delay of the restoration path.
192
Step 16.1 Check if any one of the adjacent nodes
belongs to the primary path. Also check if
the delay of the adjacent nodes is less than or
equal to the summation of the delay of the
primary path from the source to the visited
node along with the difference between the
delay constraint and the delay of the primary
path. If both these conditions are satisfied
then find the minimum delay between the
primary path of the source to the visited node
and the delay of the adjacent nodes.
Step 16.2 If both the adjacent nodes belong to the
primary path find if the delay of the route
from the source to the previous node is less
than or equal to the delay of primary path
from the source to the previous node; then,
find the delay of the primary path from the
source to the visited node.
Step 16.3 If the above mentioned two constraints do
not satisfy then add the delay of the route
from the source to the previous node and
delay of the previous node to the visited
node.
Step 17: Step 16 repeated until the destination node is reached.
Step 18: Display the restoration path from the given source to
the destination node.
Step 19: Display the minimum propagation delay of the
restoration path that satisfies the delay constraint at the
optimal cost.
Step 20: Find the delay of the route and cost of the route for the
primary quality of service path and the restoration path,
and whether the total cost for both the paths are
optimum.
3. ALGORITHMS USED FOR COMPARISON
We have proposed a combination of the provisioning and
restoration algorithm, named as the Provisioned Restorable
Virtual Private Networks Algorithm (PRA) for the VPN which is
the main criterion for reducing the total cost. Here, we have used
the Waxman and Barabasi-Albert Model to generate a random
topology of the virtual private network for symmetric as well as
Asymmetric cases. The number of nodes chosen for analysis are
100 to 1000. Also, we have compared our Provisioned
Restorable Algorithm with the disjoint path(DP) algorithm
(Yigal Bejerano et al 2005), and the Adjustment Restoration
Virtual Private Networks Algorithm(ARA) on the basis of the
following parameters: cost, number of nodes, delay and various
delay constraints.
4. SIMULATION RESULTS AND DISCUSSION
(WAXMAN –MODEL)
4.1 NETWORK SIZE VS COST FOR THE
SYMMETRIC VPN USING THE WAXMAN MODEL
Figures 1, 2, 3, 4 and 5 depict the symmetric provisioning
cost of the PRA, ARA and DP algorithms for 500 boundary
ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.1, ISSUE: 04

VPN nodes among the 1000 nodes with various delay constraints Comparison of ARA, DP and PRA Cost
of 3, 4, 8, 9 and 10 seconds respectively. The model used for 5000
DP Cost
topology creation is the Waxman Model. The bandwidth of each
link is assigned randomly by simply specifying the minimum 4000
ARA Cost
and maximum values in the range of 0 to 2500 Mbps. The VPN PRA Cost
nodes are randomly assigned with a symmetric bandwidth so

Cost(Mbps)
3000
that the bandwidth is higher than any randomly assigned
bandwidth for the links. The value plotted in the graph was the
average of the readings taken from fifteen different runs of the 2000
experiment.
1000
Comparison of ARA, DP and PRA Cost
6000
DP Cost 0

5000
ARA Cost
Nodes
Cost(Mbps)

4000
PRA Cost
Fig.3. Effect of the Number of nodes on cost for the
3000
Symmetric case of the Waxman Model with a delay constraint of
2000 8 seconds (500 VPN nodes among 1000 nodes)

1000 Comparison of ARA, DP and PRA Cost

5000 DP Cost
0
ARA Cost
4000
PRA Cost
Nodes
3000
Cost(Mbps)

Fig.1. Effect of the Number of nodes on cost for the

Symmetric case of the Waxman Model with a delay constraint of
2000
3 seconds (500 VPN nodes among 1000 nodes)
1000
Comparison of ARA, DP and PRA Cost
6000 DP Cost 0

5000 ARA Cost

PRA Cost Nodes

Cost (Mbps)

4000
Fig.4. Effect of the Number of nodes on cost for the
3000
Symmetric case of the Waxman Model with a delay constraint of
2000
9 seconds (500 VPN nodes among 1000 nodes)

1000 Comparison of ARA, DP and PRA Cost

0 DP Cost
4000
ARA Cost
PRA Cost
Nodes 3000
Cost(Mbps)

Fig.2. Effect of the Number of nodes on cost for the

Symmetric case of the Waxman Model with a delay constraint of 2000
4 seconds (500 VPN nodes among 1000 nodes)
1000

Nodes
Fig.5. Effect of the Number of nodes on cost for the Symmetric
case of the Waxman Model with a delay constraint of 10 seconds
(500 VPN nodes among 1000 nodes)

193
R. RAVI: PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI AND WAXMAN TOPOLOGY GENERATION MODEL

Figures 1, 2, 3, 4 and 5 show that our PRA algorithm is certainly Comparision of ARA, DP and PRA Cost
optimal for the symmetric case than the ARA and DP. The cost 30
DP Delay
of the PRA is lesser than that of the ARA and DP algorithms.
25 ARA Delay
4.2 NETWORK SIZE VS DELAY FOR THE PRA Delay
SYMMETRIC VPN USING THE WAXMAN MODEL 20

Delay(Seconds)
Figures 6, 7, 8 and 9 depict the symmetric provisioning delay 15
of the PRA, ARA and DP algorithms for 500 boundary VPN
nodes among the 1000 nodes with various delay constraints of 10
11, 9, 8 and 6 seconds respectively. The model used for topology
creation is the Waxman Model. The value plotted in the graph 5
was the average of the readings taken from fifteen different runs
of the experiment. 0

Comparison of ARA, DP and PRA Cost

35 Nodes
DP Delay
ARA Delay Fig.8. Effect of the Number of nodes on delay for the Symmetric
28 case of the Waxman Model with a delay constraint of 8 seconds
PRA Delay
(500 VPN nodes among 1000 nodes)
Delay(Seconds)

21
Comparision of ARA, DP and PRA Cost
14 20
DP Delay

7 ARA Delay
Delay(Seconds) 15 PRA Delay
0
10

Nodes
5
Fig.6. Effect of the Number of nodes on delay for the Symmetric
case of the Waxman Model with a delay constraint of 11 seconds
(500 VPN nodes among 1000 nodes) 0

Comparision of ARA, DP and PRA Cost

30 Nodes
DP Delay

25 ARA Delay Fig.9. Effect of the Number of nodes on delay for the Symmetric
case of the Waxman Model with a delay constraint of 6 seconds
PRA Delay
20 (500 VPN nodes among 1000 nodes)
Delay(Seconds)

15 Figures 6, 7, 8 and 9 show that our PRA algorithm is

certainly optimal for the symmetric case than the ARA and DP.
10 The delay of the PRA is lesser than that of the ARA and DP
algorithms.
5
4.3 ASYMMETRIC RATIO VS COST OF VPN
0
USING THE WAXMAN MODEL
Figures 10, 11, 12, 13 and 14 depict the Asymmetric
Nodes provisioning cost of the PRA, ARA and DP algorithms for 500
Fig.7. Effect of the Number of nodes on delay for the Symmetric boundary VPN nodes among the 1000 nodes with various delay
case of the Waxman Model with a delay constraint of 9 seconds constraints of 3, 4, 8, 9 and 10 seconds respectively. We have
(500 VPN nodes among 1000 nodes) varied the asymmetric ratio ranging from 50 to 250. The model
used for topology creation is the Waxman Model. The
bandwidth of each link is assigned randomly by simply
specifying the minimum and maximum values in the range of 0
to 2000 Mbps. VPN nodes are randomly assigned with
Asymmetric bandwidth so that the bandwidth is higher than any
randomly assigned bandwidth for the links. The value plotted in

194
ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.1, ISSUE: 04

the graph was the average of the readings taken from fifteen Comparision of ARA, DP and PRA Cost
different runs of the experiment. 2500
DP Cost
Comparision of ARA, DP and PRA Cost
3500 ARA Cost
DP Cost 2000
PRA Cost
ARA Cost

Cost (Mbps)
2800 1500
PRA Cost
Cost(Mbps)

2100 1000

1400 500

700 0
50 100 150 200 250
0
50 100 150 200 250 Asymmetric Ratio (%)

Asymmetric Ratio (%) Fig.13. Effect of the Asymmetric Ratio on cost of the Waxman
Model with a delay constraint of 9 seconds (500 VPN nodes
Fig.10. Effect of the Asymmetric Ratio on cost of the Waxman among 1000 nodes)
Model with a delay constraint of 3 seconds (500 VPN nodes
among 1000 nodes) Comparision of ARA, DP and PRA Cost
2000
DP Cost
Comparision of ARA, DP and PRA Cost
4000 ARA Cost
DP Cost 1600
PRA Cost
ARA Cost
Cost (Mbps)
3000 1200
PRA Cost
Cost(Mbps)

2000 800

400
1000

0
0 50 100 150 200 250
50 100 150 200 250
Asymmetric Ratio (%)
Asymmetric Ratio (%)
Fig.14. Effect of the Asymmetric Ratio on cost of the Waxman
Fig.11. Effect of the Asymmetric Ratio on cost of the Waxman Model with a delay constraint of 10 seconds (500 VPN nodes
Model with a delay constraint of 4 seconds (500 VPN nodes among 1000 nodes)
among 1000 nodes)
Figures 10, 12, 13 and 14 show that our PRA algorithm is
Comparision of ARA, DP and PRA Cost certainly optimal for the Asymmetric case than the ARA and
3500 DP Cost DP. The cost of the PRA is lesser than that of the ARA and DP
algorithms.
ARA Cost
2800
PRA Cost 4.4 ASYMMETRIC RATIO VS DELAY OF VPN
USING THE WAXMAN MODEL
Cost (Mbps)

2100
Figures 15, 16, 17 and 18 depict the Asymmetric
1400 provisioning delay of the PRA, ARA and DP algorithms for 500
boundary VPN nodes among the 1000 nodes with various delay
700 constraints of 10, 6, 5 and 4 seconds respectively. The model
used for topology creation is the Waxman Model. The value
0 plotted in the graph was the average of the readings taken from
50 100 150 200 250 fifteen different runs of the experiment.
Asymmetric Ratio (%)
Fig.12. Effect of the Asymmetric Ratio on cost of the Waxman
Model with a delay constraint of 8 seconds (500 VPN nodes
among 1000 nodes)

195
R. RAVI: PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI AND WAXMAN TOPOLOGY GENERATION MODEL

Comparision of ARA, DP and PRA Cost Comparision of ARA, DP and PRA Cost
30 14
DP Delay DP Delay
12 ARA Delay
25 ARA Delay
10 PRA Delay

Delay(Seconds)
PRA Delay
Delay(Seconds)

20
8

15 6

4
10
2
5
0
50 100 150 200 250
0
50 100 150 200 250 Asymmetric Ratio (%)
Asymmetric Ratio (%) Fig.18. Effect of the Asymmetric Ratio on delay of the Waxman
Model with a delay constraint of 4 seconds (500 VPN nodes
Fig.15. Effect of the Asymmetric Ratio on delay of the Waxman among 1000 nodes)
Model with a delay constraint of 10 seconds (500 VPN nodes
among 1000 nodes) Figures 15, 16, 17 & 18 show that PRA algorithm is certainly
optimal for the Asymmetric case than the ARA & DP. The delay
Comparision of ARA, DP and PRA Cost of the PRA is lesser than that of the ARA and DP algorithms.
20
DP Delay
5. SIMULATION RESULTS AND DISCUSSION
ARA Delay
16 (BARABASI-ALBERT MODEL)
Delay (Seconds)

PRA Delay

12 5.1 NETWORK SIZE VS COST FOR THE

SYMMETRIC VPN USING THE BARABASI-
8 ALBERT MODEL
Figures 19, 20, 21, 22 and 23 depict the symmetric
4
provisioning cost of the PRA, ARA and DP algorithms for 500
boundary VPN nodes among the 1000 nodes with various delay
0 constraints of 3, 4, 8, 9 and 10 seconds respectively. The model
50 100 150 200 250 used for topology creation is the Barabasi-Albert Model. The
Asymmetric Ratio (%) bandwidth of each link is assigned randomly by simply
specifying the minimum and maximum values in the range of 0
Fig.16. Effect of the Asymmetric Ratio on delay of the Waxman to 1000 Mbps. VPN nodes are randomly assigned with
Model with a delay constraint of 6 seconds (500 VPN nodes symmetric bandwidth so that the bandwidth is higher than any
among 1000 nodes) randomly assigned bandwidth for the links. The value plotted in
the graph was the average of the readings taken from fifteen
Comparision of ARA, DP and PRA Cost different runs of the experiment.
16
DP Delay Comparision of ARA, DP and PRA Cost
2000 DP Cost
ARA Delay
12 ARA Cost
PRA Delay 1600
Delay(Seconds)

PRA Cost
Cost (Mbps)

1200
8

800

4
400

0
0
50 100 150 200 250

Asymmetic Ratio (%) Nodes

Fig.17. Effect of the Asymmetric Ratio on delay of the Waxman Fig.19. Effect of the Number of nodes on cost for the Symmetric
Model with a delay constraint of 5 seconds (500 VPN nodes case of the Barabasi-Albert Model with a delay constraint of 3
among 1000 nodes) seconds (500 VPN nodes among 1000 nodes)

196
ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.1, ISSUE: 04

Comparision of ARA, DP and PRA Cost Comparision of ARA, DP and PRA Cost
1600 DP Cost 1200
DP Cost
ARA Cost
1000
ARA Cost
1200 PRA Cost

Cost (Mbps)
800 PRA Cost
Cost (Mbps)

800 600

400
400
200

0 0

Nodes Nodes
Fig.20. Effect of the Number of nodes on cost for the Symmetric Fig.23. Effect of the Number of nodes on cost for the Symmetric
case of the Barabasi-Albert Model with a delay onstraint of 4 case of the Barabasi-Albert Model with a delay constraint of 10
seconds (500 VPN nodes among 1000 nodes) seconds (500 VPN nodes among 1000 nodes)

Comparision of ARA, DP and PRA Cost Figures 19, 20, 21, 22 and 23 show that our PRA algorithm is
certainly optimal for the symmetric case than the ARA and DP.
DP Cost The cost of the PRA is lesser than that of the ARA and DP
1200 algorithms.
ARA Cost
5.2 NETWORK SIZE VS DELAY FOR THE
Cost(Mbps)

900 PRA Cost

SYMMETRIC VPN USING THE BARABASI-
600 ALBERT MODEL
Figures 24, 25, 26 and 27 depict the symmetric provisioning
300 delay of the PRA and ARA and DP algorithms for 500 boundary
VPN nodes among the 1000 nodes with various delay constraints
0
of 6, 7, 8 and 9 seconds respectively. The model used for
topology creation is the Barabasi-Albert Model. The value
plotted in the graph was the average of the readings taken from
Nodes fifteen different runs of the experiment.

Fig.21. Effect of the Number of nodes on cost for the Symmetric Comparision of ARA, DP and PRA Cost
case of the Barabasi-Albert Model with a delay constraint of 8 20 DP Delay
seconds (500 VPN nodes among 1000 nodes) ARA Delay
16
PRA Delay
Comparision of ARA, DP and PRA Cost
Delay(Seconds)

1200 12
DP Cost

1000 ARA Cost 8

800 PRA Cost

Cost (Mbps)

4
600
0
400

200 Nodes

0
Fig.24. Effect of the Number of nodes on delay for the
Symmetric case of the Barabasi-Albert Model with a delay
100
200
300
400
500
600
700
800
900
1000

constraint of 6 seconds (500 VPN nodes among 1000 nodes)

Nodes
Fig.22. Effect of the Number of nodes on cost for the Symmetric
case of the Barabasi-Albert Model with a delay constraint of 9
seconds (500 VPN nodes among 1000 nodes)

197
R. RAVI: PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI AND WAXMAN TOPOLOGY GENERATION MODEL

Comparision of ARA, DP and PRA Cost Figures 24, 25, 26 and 27 shows that our RA algorithm is
25 DP Delay
certainly optimal for the symmetric case than the ARA and DP.
The delay of the PRA is lesser than that of the ARA and DP
ARA Delay
algorithms.
20
PRA Delay
Delay (Seconds)

5.3 ASYMMETRIC RATIO VS COST OF VPN

15
USING THE BARABASI-ALBERT MODEL
10 Figures 28, 29, 30, 31 and 32 depict the Asymmetric
provisioning cost of the PRA and ARA and DP algorithms for
5 500 boundary VPN nodes among the 1000 nodes with various
delay constraints of 3, 4, 8, 9 and 10 seconds respectively. We
0
have varied the asymmetric ratio ranging from 50 to 250. The
model used for topology creation is the Barabasi-Albert Model.
The bandwidth of each link is assigned randomly by simply
Nodes specifying the minimum and maximum values in the range of 0
to 1000 Mbps. VPN nodes are randomly assigned with
Fig.25. Effect of the Number of nodes on delay for the Asymmetric bandwidth so that the bandwidth is higher than any
Symmetric case of the Barabasi-Albert Model with a delay randomly assigned bandwidth for the links. The value plotted in
constraint of 7 seconds (500 VPN nodes among 1000 nodes) the graph was the average of the readings taken from fifteen
different runs of the experiment.
Comparision of ARA, DP and PRA Cost
Comparision of ARA, DP and PRA Cost
25 DP Delay 2000
DP Cost
ARA Delay
20 ARA Cost
PRA Delay 1600
PRA Cost
Delay(Seconds)

Cost (Mbps)

15
1200

10
800

5
400

0
0
50 100 150 200 250
Nodes Asymmetric Ratio (%)
Fig.26. Effect of the Number of nodes on delay for the Fig.28. Effect of the Asymmetric Ratio on cost of the Barabasi-
Symmetric case of the Barabasi-Albert Model with a delay Albert Model with a delay constraint of 3 seconds (500 VPN
constraint of 8 seconds (500 VPN nodes among 1000 nodes) nodes among 1000 nodes)

Comparision of ARA, DP and PRA Cost Comparision of ARA, DP and PRA Cost
30 1800
DP Delay DP Cost
25 1500 ARA Cost
ARA Delay
Delay(Seconds)

PRA Cost
20 PRA Delay
Cost (Mbps)

1200

15 900

10
600

5
300

0
0
50 100 150 200 250
Nodes Asymmetric Ratio (%)

Fig.27. Effect of the Number of nodes on delay for the Fig.29. Effect of the Asymmetric Ratio on cost of the Barabasi-
Symmetric case of the Barabasi-Albert Model with a delay Albert Model with a delay constraint of 4 seconds (500 VPN
constraint of 9 seconds (500 VPN nodes among 1000 nodes) nodes among 1000 nodes)

198
ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.1, ISSUE: 04

Comparision of ARA, DP and PRA Cost Figures 28, 29, 30, 31 and 32 show that our PRA algorithm is
1600 certainly optimal for the Asymmetric case than the ARA and
DP Cost DP. The cost of the PRA is lesser than that of the ARA and DP
ARA Cost algorithms.
1200 PRA Cost
5.4 ASYMMETRIC RATIO VS DELAY OF VPN
Cost (Mbps)

USING THE BARABASI-ALBERT MODEL

800
Figures 33, 34, 35 and 36 depict the Asymmetric
provisioning delay of the PRA and ARA and DP algorithms for
400 500 boundary VPN nodes among the 1000 nodes with various
delay constraints of 4, 5, 7 and 8 seconds respectively. The
model used for topology creation is the Barabasi-Albert Model.
0 The value plotted in the graph was the average of the readings
50 100 150 200 250 taken from fifteen different runs of the experiment.
Asymmetric Ratio (%)
Fig.30. Effect of the Asymmetric Ratio on cost of the Barabasi-
Albert Model with a delay constraint of 8 seconds (500 VPN Comparision of ARA, DP and PRA Cost
nodes among 1000 nodes) 14
DP Delay
12 ARA Delay
Comparision of ARA, DP and PRA Cost
10 PRA Delay

Delay (Seconds)
DP Cost
1200 ARA Cost 8

PRA Cost 6
Cost (Mbps)

900
4

600 2

0
300 50 100 150 200 250
Asymmetric Ratio (%)
0 Fig.33. Effect of the Asymmetric Ratio on delay of the Barabasi-
50 100 150 200 250 Albert Model with a delay constraint of 4 seconds (500 VPN
Asymmetric Ratio (%) nodes among 1000 nodes)
Fig.31. Effect of the Asymmetric Ratio on cost of the Barabasi-
Albert Model with a delay constraint of 9 seconds (500 VPN Comparision of ARA, DP and PRA Cost
nodes among 1000 nodes) 16
DP Delay
ARA Delay
Comparision of ARA, DP and PRA Cost
12 PRA Delay
Delay(Seconds)

1200
DP Cost

1000 ARA Cost

8
PRA Cost
Cost (Mbps)

800

4
600

400
0
50 100 150 200 250
200
Asymmetric Ratio (%)
0 Fig.34. Effect of the Asymmetric Ratio on delay of the Barabasi-
50 100 150 200 250
Albert Model with a delay constraint of 5 seconds (500 VPN
Asymmetric Ratio (%) nodes among 1000 nodes)
Fig.32. Effect of the Asymmetric Ratio on cost of the Barabasi-
Albert Model with a delay constraint of 10 seconds (500 VPN
nodes among 1000 nodes)

199
R. RAVI: PROVISIONING RESTORABLE VIRTUAL PRIVATE NETWORKS USING BARABASI AND WAXMAN TOPOLOGY GENERATION MODEL

Comparision of ARA, DP and PRA Cost reserves less bandwidth when compared to the DP and ARA.
25 The Provisioning Restorable Virtual Private Networks
DP Delay Algorithm (PRA) requires minimum delay when compared to
20 ARA Delay the Disjoint Path Algorithm and the Adjustment Restoration
Virtual Private Networks Algorithm, even though the delay
Delay(Seconds)

PRA Delay
threshold is specified in the networks.
15

ACKNOWLEDGEMENT
10
It is, indeed, a great privilege to express my deep sense of
5 gratitude to my respected supervisor Dr. K. Rajagopal M.E.,
Ph.D, Principal, Francis Xavier Engineering College,for his
invaluable guidance, stimulating encouragement, sparkling
0
50 100 150 200 250
discussions, and the thorough explanations which he gave me for
the successful completion of this work.
Asymmetric Ratio (%)
I would like to place on record, my indebtedness to
Fig.35. Effect of the Asymmetric Ratio on delay of the Barabasi- Chairman Dr. S.Cletus Babu, Chairman,Francis Xavier
Albert Model with a delay constraint of 7 seconds (500 VPN Engineering College, for his inspiration and constant motivation.
nodes among 1000 nodes) Finally, I wish to express my gratitude to my family and my
friends, for providing unending support in every respect
Comparision of ARA, DP and PRA Cost possible, to ease my path on this journey.
25
DP Delay
ARA Delay REFERENCES
20
PRA Delay
[1] Chittaranjan Hota, Sanjay Kumar Jha and Raghurama G,
Delay(Seconds)

15
10
5 [3] Chris Metz, 2004, “The Latest in VPNs: Part II”, IEEE
0 [4] Chris Metz, 2006, “Multi protocol Label Switching and IP,
50 100 150 200 250
Asymmetric Ratio (%)
Fig.36. Effect of the Asymmetric Ratio on delay of the Barabasi-
Albert Model with a delay constraint of 8 seconds (500 VPN
nodes among 1000 nodes)
Figures 33, 34, 35 and 36 show that our PRA algorithm is
certainly optimal for the Asymmetric case than the ARA and
DP. The delay of the PRA is lesser than that of the ARA and DP
algorithms.
6. SUMMARY OF CONTRIBUTIONS
In the above simulation study, we have used both the
Waxman and Barabasi - Albert Model for topology generation.
The numbers of nodes are varied from 0 to 1000 and the
numbers of VPN nodes are fixed as 500. The delay constraint is
varied from 3 to 10 seconds. Both symmetric and Asymmetric
bandwidths were studied. The Asymmetric Ratio is also varied
from 0 to 250.
Our Provisioning Restorable Virtual Private Networks
Algorithm (PRA) shows better performance than the Disjoint
Path Algorithm (DP) and Adjustment Restoration Virtual Private
Networks Algorithms (ARA) by optimizing the total bandwidth
reserved on the edges of the Restorable VPN tree. The PRA
2005, “Restoration of Virtual Private Networks with QoS
Guarantees in the Pipe Model’, 6th International Workshop
on Distributed Computing, Vol. 3326, pp. 127-141.
[2] Chris Metz, 2003, “The Latest in Virtual Private Networks:
PartI”, IEEE Internet Computing, Vol. 7, No. 1, pp. 87-91.
Internet Computing, Vol. 8, No. 3, pp. 60-65.
Part 2: Multicast Virtual Private Networks”, IEEE Internet
Computing, Vol. 10, No. 1, pp. 76-81.
[5] Christian Labonte and Srinivas Sampalli, 2000, “Group
Management Strategies for Secure Multicasting on Active
Virtual Private Networks”, Proceedings of 25th Annual
IEEE Conference on Local Computer Networks,
pp.213-222.
[6] Christian Labonte and Srinivas Sampalli, 2004, “Secure
Architecture for Multicasting on Active Virtual Private
Networks”, IEEE Military Communications Conference,
Vol. 1, pp. 301-307.
[7] Christian Muller, Emmanuel Dotaro and Dimitri
Papadimitriou, 2006, “A Practical Approach to VPN
Resource Management using a Dynamic Hose Model”,
IEEE 2nd Conference on Next Generation Internet Design
and Engineering, pp. 147-153.
[8] Cicek Cavdar and Aysegul Gencata, 2007, “CATZ: Time-
Zone-Aware Bandwidth Allocation in Layer 1 VPNs”,
IEEE Communications Magazine, Vol. 45, No.4, pp.60-66.
[9] Cicek Cavdar, Aysegul Gencata Yayimli and Biswanath
Mukherjee, 2008, “Multi-Layer Resilient Design for Layer-
1 VPNs”, Optical Fiber communication/National Fiber
Optic Engineers Conference, pp. 1-3.

200
ISSN: 2229– 6948 (ONLINE) ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, DECEMBER 2010, VOL.1, ISSUE: 04
[10] Craig Shue, Youngsang Shin, Minaxi Gupta and Jong Youl
Choi 2005, “Analysis of IPSec Overheads for VPN
Servers”, 1st IEEE ICNP Workshop on Secure Network
Protocols, pp. 25-30.
[11] Cristian Lambiri, Dan Ionescu and Voicu Groza, 2004, “A
New Method for the Estimation and Measurement of
Traffic Packet Loss for Virtual Private Networks Services”,
Proceedings of the 21st IEEE Instrumentation and
Measurement Technology Conference, Vol. 1, pp. 401-406.
[12] Csaba Antal, Janos Harmatos, Alpar Juttner and Gabor
Toth, 2005, “A Cluster-based Resource Provisioning
[15] David Lewis, Bjerring L.H. and Thorarensen I.H., 1996,
“An Inter-Domain Virtual Private Network Management
Service”, IEEE Network Operations and Management
Symposium, Vol. 1, pp. 115-123.
[16] Debasis Mitra and Ilze Ziedins, 1997, “Hierarchical Virtual
Partitioning: Algorithms for Virtual Private Networking’,
IEEE Global Telecommunications Conference, Vol. 3,
pp.1784-1791.
[17] Debasis Mitra, Morrison J.A. and Ramakrishnan K.G.,
1999, “Virtual Private Networks: Joint Resource Allocation
and Routing Design”, Eighteenth Annual Joint Conference
of the IEEE Computer and Communications Societies,
Vol.2, pp.480-490.
201
Model in Virtual Private Networks”, IEEE Proceedings of
5th International Workshop on Design of Reliable
Communication Networks, pp.367-374.
[13] Daniel A., 2004, “IP Virtual Private Networks- A Service
Provider Perspective”, IEE Proceedings Communications,
Vol.151, No. 1, pp. 62-70.
[14] Darrell Kindred and Daniel Sterne, 2001, “Dynamic VPN
Communities: Implementation and Experience”,
Proceedings of DARPA Information Survivability
conference and Exposition II, Vol. 1, pp. 254-263.
[18] Dongli Zhang and Dan Ionescu, 2004, “A New Practical
Packet Loss Estimator for MPLS VPN Services”, 10th
Asia-Pacific Conference on Communications and the 5th
International Symposium on Multi-Dimensional Mobile
Communications Proceedings, Vol. 2, pp. 587-591.
[19] Dongli Zhang and Dan Ionescu, 2004, “Feedback Control
of Packet Loss for VPN Services’, IEEE Proceedings of
12th International Conference on Networks, Vol. 2,
pp.583-587.
[20] Dongli Zhang and Dan Ionescu, 2004, “On Packet Loss
Estimation for Virtual Private Networks Services”, IEEE
Proceedings of 13th International Conference on Computer
Communications and Networks, pp.175-180.