Part III
Wide-Area (Wavelength-Routed)
Optical Networks –
1. Virtual Topology Design
2. Wavelength Conversion
3. Control and Management
BM-UC Davis
Optical Networks
122
Lightpaths and Wavelength Routing
 Lightpath
 Virtual topology
 Wavelength-continuity
constraint
 Wavelength conversion
 Packet routing
BM-UC Davis
Optical Networks
123
Illustrative example
NY
WA
MI
NJ
PA
UT
CA1
CO
NE
IL
MD
CA2
GA
TX
BM-UC Davis
Optical Networks
124
Solution 1a: Infocom’94 and ToN-Oct96
 More than one laser filter pair at any node can tune to the
same wavelength
BM-UC Davis
Optical Networks
125
Solution 1b: Infocom’94 and ToN-Oct96
 All laser filter pairs at any node must be tuned to different
wavelengths
BM-UC Davis
Optical Networks
126
Virtual Topology
BM-UC Davis
Optical Networks
127
Wavelength Routing Switch (WRS)–Details of the UT Node
BM-UC Davis
Optical Networks
128
Optimization Problem Formulation

On virtual topology connection matrix Vij
V
ij
 Ti

On virtual topology traffic variables
ijsd  0
i

j
Vij  R j
sd
sj
j
On physical route variables
pij

sd
sj

sd
ik
i
ij
pmn
 Vij
p
m
p
ij
kn
p
p
ij
in
 Vij
sd
ij
 Vij
m
if k  s, d
 Vij  C
s ,d

On coloring of lightpaths cijk
c
ij
k
n
ij
mj
  sdkj
j

if k  i, j
n
 sd
j
mn
ij
pmn
 Pmn
ij
mk
 sd
j
i

sdij
p
 Vij
k
ij
mn
 ckij  1 m, n, k
ij

Objective: Optimality criterion
New optimality criterion
(a) Delay minimization:


 (c) Minimize average hop distance
1
ij

Minim ize    ijsd  pmn
 d mn 
sd 

C  sd ij  
ij
 mn
 sd

1
ijsd

(b) Maximizing offered load (equivalent to minimizing maximum flow inMinimize
a link):
s,d sd i, j s,d




min max ijsd   i, j
 sd


BM-UC Davis
Optical Networks
129
Solution Approach to Virtual Topology WDM WAN Design
1. Choice of optimal virtual topology

Simulated annealing; optimization based on maximizing
throughput, minimizing delay, maximizing single-hop traffic, etc.
2. Routing of lightpaths over the physical topology

Alternate-path routing, multicommodity flow formulation,
randomized routing
3. Wavelength assignment: Coloring of lightpaths to avoid
wavelength clashes

Graph-coloring algorithms, layered graph models
4. (Optimal) routing of packets over the virtual topology

Shortest-path routing, flow-deviation algorithm, etc.
5. Iterate

Check for convergence and go back to Step 1, if necessary.
BM-UC Davis
Optical Networks
130
Details of Virtual Topology Design
 Simulated Annealing
 Start with random virtual topology
 Perform node exchange operations on two random nodes
 Route packet traffic (optimally) using flow deviation
 Calculate maximum traffic scaleup for current configuration
 If maximum scaleup is higher then previous maximum,
then accept current configuration;
else accept current configuration with certain decreasing
probability
 Repeat until problem solution stabilizes (frozen).
 Flow Deviation
 Perform shortest-path routing of the traffic
 Select path with large traffic congestion
 Route a fraction of this traffic to less-congested links
 Repeat above two steps iteratively, until solution is
acceptable
BM-UC Davis
Optical Networks
131
NSFNET Traffic Matrix (11:45 PM to midnight, ET, Jan. 12, 1992)
BM-UC Davis
Optical Networks
132
The WDM Advantage
BM-UC Davis
Transceivers
/node
Scaleup
4
106
5
135
6
163
Optical Networks
133
Delay Components in a WDM Solution
BM-UC Davis
Optical Networks
134
Scaling of Bandwidth – The WDM Advantage
 No WDM (Physical Topology)
p 
Lp  C
Hp
C = link speed (Mbps)
Hp = avg. hop distance (physical)
Mbps
N = number of nodes
 WDM (with P transmitters/receivers per node)
v 
Lv  C NP  C

Mbps
Hv
Hv
 WDM Advantage
 v Lv H p NP H p




 p Lp H v
Lp H v
v
P

 p Hv
Increasing P decreasing Hv
BM-UC Davis
Optical Networks
135
Problems/Limitations of Solution 1
 Nonlinear objective functions.
 Nonlinear constraints – on wavelength continuity.
 Resorted to heuristics
 Optimal virtual topology design (Simulated Annealing)
 Optimal packet routing on V.T. (Flow Deviation
Algorithm)
 No routing and wavelength assignment
(Shortest-path lightpath routing; no constraints on
wavelengths).
BM-UC Davis
Optical Networks
136
Highlights/Contributions of Solution 2
 Complete Virtual Topology Design
 Linear formulation  Optimal solution
 Objective: Minimize average hop distance
 Assume: Wavelength conversion
(Sparse conversion provides almost full conversion
benefits).
 Resource Budgeting Tradeoffs
 Important/Expensive Resources: Transceivers and
wavelengths
 Don’t under-utilize either of them!
 Hardware cost model.
 Optimal Reconfiguration Algorithm
 Minimize reconfiguration time.
BM-UC Davis
Optical Networks
137
Optional Constraints / Simplifying Assumptions
 Need scalability.
 Physical topology is a subset of the virtual
topology.
 Bounded lightpath length
Prevent long convoluted lightpaths from occuring.
 Prune the search space
Consider K shortest paths (bounded K).
BM-UC Davis
Optical Networks
138
Two Solutions from the LP
(a) Two-wavelength
solution
(b) Five-wavelength
solution
BM-UC Davis
Optical Networks
139
Hop Distance, Transceiver + Wavelength Utilization
BM-UC Davis
Optical Networks
140
Average Hop Distance
BM-UC Davis
Optical Networks
141
Transceiver Utilization
BM-UC Davis
Optical Networks
142
Wavelength Utilization
BM-UC Davis
Optical Networks
143
Heuristic Solutions
BM-UC Davis
Optical Networks
144
WDM Network Cost Model
N



 N

 Ti  N  R j  

C  Ct    Ti   Ri   Cm   2M           WC x    m  log m  / 2
i
i 1 W 
j 1  W  
 i

 m 1


BM-UC Davis
Optical Networks
145
Reconfiguration Algorithm
 Generate linear formulations F(1) and F(2) corresponding to traffic
matrices sd1 and sd2.
 Derive solutions and S(1) and S(2), corresponding to F(1) and F(2)
 Modify F(2) to F’(2) by adding the new constraint:
1
sd  sd
 
sd
ij
 OPT2
i , j s ,d
 New objective function for F’(2) :
ij
ij
Minimize:  pmn
(2)  pmn
(1)
or
ij
mn
Minimize:  Vij (2)  Vij (1)
ij
 Although mod is nonlinear, above reconfiguration formulation is
linear since the variables p’s and V’s are binary.
BM-UC Davis
Optical Networks
146
Reconfiguration Statistics
BM-UC Davis
Optical Networks
147
Summary of Virtual Topology Design Principles
 Use WDM to scale up an existing fiber-based WAN
(Network’s information carrying capacity increased
manifold)
 Employ packet-switched virtual topology
… imbedded on a physical topology
… as if we have a virtual Internet
(which is reconfigurable under user control)
… need optimum graph-imbedding algorithms
 Reuse electronic switch of existing WAN
… as part of the WRS in the scaled-up WAN
BM-UC Davis
Optical Networks
148