Académique Documents
Professionnel Documents
Culture Documents
1 INTRODUCTION
Transparent optical networks are the recent technological
advancement for supporting huge traffic in the future
communication networks. To transfer data in wavelength
routed optical network, a lightpath must be established
between source-destination pair by a Routing and Wave-
length Assignment (RWA) technique . A lightpath is an
all optical communication channel between two nodes,
which may consist of one or more fiber links. The com-
plexity arises for selection of lightpath with appropriate
wavelength between pair of nodes in the network. Wave-
length continuity is a common problem in wavelength
routed networks, where every link in the lightpath must
share a free common wavelength. A proper RWA scheme
must be implemented in order to avoid blocking in future
connection requests. As the optical signal from the source
node propagates along a lightpath in an all optical trans-
parent wavelength routed network towards the destina-
tion node, the quality of signal degrades as there is no
OEO (Optical-Electronic-Optical) conversion and there-
fore, the Bit Error Rate (BER) of the signal increases or
OSNR decreases. OSNR below the threshold level is un-
acceptable by the user. Again, it is not acceptable if estab-
lishment of a lightpath decreases the OSNR of other exist-
ing lightpath. RWA technique added with physical layer
impairment constraints is much more practical for
lightpath establishment and is called quality of transmis-
sion aware routing and wavelength assignment tech-
2015 JOT
www.journaloftelecommunications.co.uk
4
W 1
lp
ij
( i , j )E
lp
ji
= 1, if i = S , lp LP
lp
ij
ij
( i , j )E
( i , j )E
lp
ji
ji
(1a)
= 1, if i = D, lp LP
ij
ij
lp
w=0 x
lp ( x , y ) ( x , y )
ijw
W 1
lp ( y , x ) ( y , x )
I ijw
l
= I ijlp , y = i
w=0 x
lp ( x , y ) ( x , y )
ijw
W 1
lp ( y , x ) ( y , x )
I ijw
l
= 0, y i , y j
w=0 x
2(d)
( j ,i )E
lp
1, if i D , lp LP
lp
= 0, if i = D , lp LP
i j
( i , j )E
(1b)
The constraint (1a) guarantees that the solutions obtained are valid paths from S to D. The first part of the
constraint (1a) ensures that for each pair of source and
destination node only one lightpath exists from source
node S. The second part of the constraint (1a) ensures
that for each pair of source and destination nodes only
one lightpath reaches the destination node D. Third part
of the constraint (1a) ensures that a path that reaches an
intermediate node also exits that node through another
link. The constraint (1b) ensures that the lightpath is
without any loops [3-5],[ 19-21].
lp
I ijlp = I ijw
(i , j )
2(a)
w=0
lp ( x , y )
lp
I ijw
I ijw
(i, j ), ( x, y ), w
2(b)
lp ( x , y )
ijw
W 1
lp ( y , x ) ( y , x )
I ijw
l
= I ijlp , y = j
= 0, if i S , i D, lp LP
i j
( i , j )E
W 1
w=0 x
w=0 x
( j ,i )E
lp
lp ( x , y ) ( x , y )
ijw
w=0 x
W 1
( j ,i )E
1( x, y ), w
i, j
2(c)
3. Proposed Algorithm
Figure 4 below depict the block diagram of the proposed
QoT-aware-FDRWA respectively. The following steps
sequentially describe the algorithm.
Step 1: A call request arrival is assumed to follow a Pois-
son distribution. Source-destination pair follow a uniform
random distribution and holding time follows (exponen-
tial/ pareto) distribution [2-5], [19].Numbers of possible
paths using dijkstras shortest path algorithm are gener-
ated [22]. A path in this work is defined as the collection
of physically connected links from a source to a destina-
tion.
Step 2: Each of these paths is checked with the availabil-
ity of free wavelengths or channels (FCA).Different wave-
Name
Type
Range
(per
path)
Membership
function
1
2
OSNR (dB)
Free Chan-
nel
Availability
Input
Input
13-30
1-16
Triangular
Triangular
3
4
Hop Count
Cost
Input
Output
1-7
0-1
Triangular
Triangular
OSNR
(dB)
FCA
HOP
COUNT
COST
VG
VG
VG
VG
VG
VG
VG
VG
VG
VG
10
VG
11
VG
VG
VG
12
VG
VG
13
VG
VG
VG
14
VG
VG
VG
VG
15
VG
VG
16
VG
VG
17
VG
VG
18
VG
19
20
VG
VG
21
VG
22
VG
VG
23
VG
VG
VG
24
VG
25
VG
26
VG
27
Values
Wavelength (nm)
1544.5 to1556.5
Number of Wavelength
Sixteen
0.8
40 Gbps
ASE factor (
90 GHz
- 4 dB
Demultiplexer loss (
- 4 dB
Switch loss (
- 8 dB
- 0.2 dB/km
18 dB
14 dB
Loss of tap (
- 1 dB
- 30 dB
Fig. 3(c) Signals in the time domain before passing through the
PMD emulator
1.5
Multiplexer loss (
Fiber loss (
1
30 GHz
Optical bandwidth (
-30 dB
0.13 ps/
Fig. 3(d) Signals in the time domain after passing through the
PMD emulator (shows delay)
(k))
Fiber Pulse broadening (
0.1
OSNR threshold
13.09 dB (for BER=
)
Fig. 3(b) Representation of subsystem using optisystem
Fig. 3(f) Signals in the frequency domain after passing through the
PMD emulator (shows spreading)
TRAFFIC
IN-
CALL
FORMATION
REQUEST
NETWORK LAYER
Generation of Sourcedestination
pair
FF
RR
ASSIGNMENT TECHNIQUE
N=0
M=0
Each
or
set
of
input
defuzzified value is
selected for establishment of LP
ly passed to FLC as
input parameters.
Request Blocked
PHYSICAL LAYER
PMD CONSTRAINT
M paths out of N paths satisfying the
OSNR CONSTRAINT
Good
Very good
Excellent
Membership degree
Membership degree
1
0.8
0.6
0.4
0.6
0.4
0.2
0.2
0
12
Good
Very good
Excellent
0.8
0
0
14
16
18
20
22
OSNR (dB)
24
26
28
30
10
12
14
16
Excellent
Very good
Good
0.6
Membership degree
Membership degree
1
0.8
0.4
0.2
0
1
Hop Count
0.8
0.6
Excellent
Very good
Good
Poor
0.4
0.2
0
0
0.2
0.4
Cost
0.6
0.8
In this work, a MATLAB Version of 7.5 with Intel (R)
Core-Duo CPU (3.3 GHz) is used for simulation. In this
paper, a 23 node topology is used as shown in figure.7,
which has 40 bi-directional links [23].
Network or traffic load in the proposed model indicates
the duration of time a network remains busy and is de-
fined as mean arrival rate x mean holding time of connec-
tion requests. It is measured in erlang. Mean blocking
probability is defined as the number of requests blocked
divided by total number of requests processed. A connec-
tion request is said to be blocked when a wavelength con-
tinuous route is not found. Mean execution time is esti-
mated as total simulation time divided by total number of
requests processed.The experiment is carried out for
fuzzy controlled mechanisms (with FLC), two types of
holding times (exponential, pareto) and for various wave-
length assignment techniques (FF, RR, R) to estimate dif-
ferent network related QoT parameters. Following steps
are carried out for the proposed QoT-aware FDRWA. The
algorithm is simulated for eight different network loads
i.e. 20 erlangs to 160 erlangs and for 100000 call requests.
For all the network loads the mean holding time is kept
constant (0.5) and mean arrival rate is varied to meet the
desired network load conditions. Different network relat-
ed QoT performance metrics are evaluated for each of the
network loads. Each of these network parameters is plot-
ted versus network load with exponential and pareto
holding times and for various wavelength assignment
techniques. Figures below show the details of the simula-
tion results.
Fig.6 below shows the graph of the average blocking
probability vs. traffic load for exponential and Pareto
holding times respectively for different wavelength as-
signment techniques such as FF, RR, and R.
0.25
0.2
0.15
0.1
0.05
0
20
40
60
80
100
120
140
160
Traffic Load in Erlang
Fig.6 Average Blocking Probability versus Traffic load
(Erlangs)
It is observed from the graphs that the average blocking
probability increases with increase in network loads for
the proposed QoT-aware FDRWA with different assign-
ment techniques. It is observed from the graph that the
average blocking probability corresponding to exponen-
tial holding time at various network load (from 50 erlangs
to 160 erlangs) is lower in comparison to that of Pareto
holding time and the mean blocking probability at differ-
ent network traffic loads (especially greater than 50 er-
QoT-Aware-FDRWA-Expo-FF
QoT-Aware-FDRWA-Expo-R
QoT-Aware-FDRWA-Expo-RR
QoT-Aware-FDRWA-Pareto-FF
QoT-Aware-FDRWA-Pareto-R
QoT-Aware-FDRWA-Pareto-RR
QoT-Aware-FDRWA-Expo-FF
QoT-Aware-FDRWA-Expo-R
QoT-Aware-FDRWA-Expo-RR
QoT-Aware-FDRWA-Pareto-FF
QoT-Aware-FDRWA-Pareto-R
QoT-Aware-FDRWA-Pareto-RR
0.07
0.06
QoT-Aware-FDRWA-Expo-FF
QoT-Aware-FDRWA-Expo-R
QoT-Aware-FDRWA-Expo-RR
QoT-Aware-FDRWA-Pareto-FF
QoT-Aware-FDRWA-Pareto-R
QoT-Aware-FDRWA-Pareto-RR
0.05
0.04
0.03
0.02
0.01
0
20
40
60
80
100
Traffic Load in Erlang
120
140
160
Fig. 8 Average Blocking Probability due to PLI versus
Traffic load (Erlangs)
Acknowledgement
The authors would like to extend their sincere appreciation to the all India council of technical education
(AICTE) for the funding of this research through the research project number. 20/AICTE/RIFD/RPS (POLICYII) 2/2012-13.
Fig.9 3D- plot of input variables (FCA & Hop-Count) vs.
output variable (COST) fixing third input variable
OSNR at 21dB
Above figures shows the surface viewer invoked using
GUI (Graphical User Interface) tool in FIS (Fuzzy Infer-
ence Systems) , Which exhibit the output surface of FIS
stored in a file for any one or two input. It does not alter
the fuzzy system or its FIS structure in any way. Surface
viewer is a read only editor and reflects fuzzy rules. In
this work, there are three inputs (OSNR, FCA, and Hop-
Count) and one output (COST) is passed to FIS. Each of
the above figures shows the surface view of the output
(COST) vs. any other two inputs, keeping third input
fixed at certain value.The fig.9 shows the surface of the
output COST (in Z-Axis) vs. input FCA (in X-Axis) and
Hop-Count (in Y-Axis), fixing the third input, OSNR at
twenty-one. As observed in the plot, there is steep varia-
tion in xz-plane in comparison to yz-plane, as according
to the fuzzy rules higher preference is given to FCA than
Hop-Count. At lower value of Hop-Count (ranging from
3 to1) and at higher value of FCA (ranging from 10 to 16)
the surface started converging towards origin. Finally at
FCA at 14 to16 and Hop-Count at 1, the surface touches
the origin.
6. Conclusion
10
Reference