QoT-aware Routing in Impairment-constrained OpticalNetworks
Jun He∗, Maıte Brandt-Pearce∗, Yvan Pointurier?, Suresh Subramaniam+{jh2eu,mb-p}@virginia.edu, [email protected], [email protected]
∗Charles L. Brown Department of Electrical and Computer Engineering
University of Virginia, USA?Department of Electrical and Computer Engineering
McGill University, Montreal, Quebec, Canada+Department of Electrical and Computer Engineering
George Washington University, USA
Funded by NSF
IEEE GLOBECOM 2007November 28, 2007
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
1/13
Background
I Wavelength routed all optical networks
¥ Circuit switched end-to-end optical channel or lightpath
¥ No optical-electrical-optical regenerators
¥ Advantages• Huge bandwidth,
• Transparency,
• Cost
• Wavelength reuse, etc.
¥ Applications• Backbone network, now being deployed in metropolitan area networks
• Large file transfer and E-science
¥ Problems• Routing and Wavelength Assignment
• Physical impairments in the physical layer
• Challenges to find a good route simultaneously considering both constraints
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
2/13
Research Motivation
¥ Previous works;
¥ The impact of physical impairments on RWA;
� � � � �� � � � �� � � �� � � � ���� � �� �� �� � ��
� � � �� � � � � �� �
� �� �� �� � � �� �� �� �� �� �� �� � � � � � �� � � � �
� � � �� � � �
� �� �� �� �
� �� �� �� �
� � �� � �� � �� � �� � �� � � � �� �� �� �� �� � !
""##switching
fabric
crosstalk
$ $ $ $ $$ $ $ $ $% % % %% % % % &'WRS
WRS
crosstak
adjacent−port( ( ( (( ( ( () ) )) ) )
LP3
λ4λ2 λ3λ1
CH1 CH2 (crosstalk)CH2(λ2)
CH2 CH1 (crosstalk)
CH1
CH2
CH1 CH2 CH3 CH4
CH1 (λ1)
CH2
CH1
crosstalk fromLP3
CH1 (λ1)
LP2
LP1
¥ Two QoT-aware routing schemes.
• Route admission control with impairment constraints.
• Routing incorporates impairments into the cost of routes.
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
3/13
BER Estimation Model
DC AG
Hop1
1Link
0 1 i n
DestinationA lightpath
i−1
Linki
Other spansOne spanSource
1
2
Si
Si
λ1, λ2, . . . , λm
12
. . .. . .
. . .. . .
¦ Dominant physical degradations
◦ Amplified spontaneous emission (ASE) noise from EDFAs;
◦ Linear crosstalk from OXCs and Demux;
◦ Nonlinear effect from four-wave mixing (FWM) and cross-phase
modulation (XPM);
◦ Shot noise and thermal noise.
¦ Power of impairments propagates to the next hop until arriving at receiver;
¦ Other linear/nonlinear impairments can be integrated in the model.
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
4/13
QoT-Aware Routing Algorithms
No
Block the request
threshold BER <
the requestAccept
A request arrives
Find the available path with theshortest distance, fewest hops,
the chosen wavelength
minimized measured Q factors, or lowest variance of noises for
BER estimation of the lightpathand other involved lightpaths
Choose a new wavelength based on wavelengthassignment algorithm, such as FF or RP
No free wavelength
Yes
¦ SD: find a route with shortest fiber length (decreasing link impairments);
¦ FH: find a route passing fewest intermediate nodes (decreasing node impairments);
¦ QM: the link weight is estimated by the average link degradation to the existing;
¦ LV: the link weight is estimated by the impairments variance on candidate wavelength.
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
5/13
Cost Functions in QoT-Aware Routing
• Routing with QoT constraints: the link weight doesn’t represent QoT
¦ The weight of hopi in Shortest distance routing (SD) is assigned asD(i) =
Length of linki;
¦ The weight of hopi in fewest hop routing (FH) is assigned asD(i) = 1;
• Routing based on QoT: the link weight considers QoT
¦ The weight of hopi in Q-maximizing routing (QM) is assigned as
D(i) =
PTil=1 10 log[Q(s)(i, l)/Q(d)(i, l)]
Ti; (1)
¦ The weight of hopi (if consideringλj) in routing based on least variance (LV) is
assigned as
W (i, j) = σ2XT (i, j) + σ2
NL(i, j) + σ2ASE(i, j) + αi; (2)
αi ⇐= max{αi + ∆α, αmax} if detecting wavelength blocking;
αi ⇐= min{αi −∆α, αmin} if detecting QoT blocking.
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
6/13
Simulation Settings
NY
CA1CO
MD
4
21
2
2
4
1
11
4
1
1 2
2
2
12
1
2
1
TX
1
MI
NJ
GA
CA2
UT
WA
NE
IL
PA
Topology of a downsized version of the NSF network with 14
nodes and 21 bidirectional links. The numbers on the links
represent number of spans along the link. Each span is70
km long.
Parameters Value
Number of Wavelength 32
Wavelength spacing 25 GHz
Data rate per channel 10 Gps
Fiber loss (Lf ) 0.2 dB/km
ASE factor(nsp) 1.5
Laser source power 0 dBm
ρ 0.95 A/W
BER threshold 10−12
Dispersion compensator 100% post
Chromatic dispersion 17 ps/nm/km
Nonlinearity constant 2.2 (W.m)−1
αmax 3× 10−10 A2
αmin 1× 10−10 A2
∆α 1× 10−11 A2
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
7/13
Blocking Probability vs Network Load if WA uses FF
120 140 160 180 200 220 24010
−5
10−4
10−3
10−2
10−1
100
Load, erlangs
Blo
ckin
g pr
obab
ility
FFwFRFFwSDFFwFHFFwQMFFwLV
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
8/13
Blocking Probability vs Network Load if WA uses RP
120 140 160 180 200 220 24010
−5
10−4
10−3
10−2
10−1
100
Load, erlangs
Blo
ckin
g pr
obab
ility
RPwFRRPwSDRPwFHRPwQMRPwLV
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
9/13
Blocking Probability vs Node Crosstalk if WA uses FF
(−30, −15) (−35, −20) (−40, −20) (−45, −25) (−50, −30)10
−4
10−3
10−2
10−1
100
(Xsw
dB, Xadj
dB)
Blo
ckin
g pr
obab
ility
FFwFRFFwSDFFwFHFFwQMFFwLV
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
10/13
Blocking Probability vs Node Crosstalk if WA uses RP
(−30, −15) (−35, −20) (−40, −20) (−45, −25) (−50, −30)10
−4
10−3
10−2
10−1
100
(Xsw
dB, Xadj
dB)
Blo
ckin
g pr
obab
ility
RPwFRRPwSDRPwFHRPwQMRPwLV
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
11/13
Conclusions and Future Work
I Current work
¥ We studied the impact of incorporating QoT constraints on theperformance of adaptive routing algorithms for wavelength-routednetworks.
¥ A new adaptive routing algorithm is proposed and compared with otherrouting algorithms.• New routing algorithm can effectively minimize the blocking probability.
• New routing algorithm only uses the state of the links between a node and its
neighbors.
. Future work◦ More network topologies and traffic patterns are applied to test the new routing
algorithm.
◦ QoT-aware WA algorithms can be applied to our adaptive routing algorithm to
further improve the performance.
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
12/13
Questions?
IEEE GLOBECOM 2007ONSS04-2
Jun He∗, Maıte Brandt-Pearce∗ , Yvan Pointurier?, Suresh Subramaniam+
13/13