Optical Networks

Chapter 10:WDM Network Design

Prof. Yogendra K. Prajapati

Department of Electronics and Communication Engineering

Motilal Nehru National Institute of Technology Allahabad
Prayagraj-INDIA
 Introduction
• In this chapter, we will design a wavelength routing network as well as
the higher-layer SONET or IP network.
• The optical layer realizes the light paths over the physical fiber using
elements such as optical line terminals (OLTs), optical add/drop
multiplexers (OADMs), and optical cross connects (OXCs). We called a
network using such light paths a wavelength-routing network.
• We will discuss the use of optical add/drop in the network, how do
they affect the cost and efficiency of network in the next slide.
A light path refers to the route or channel through which optical signals travel in a fiber optic network
Use of optical Add/drop in network
 No optical add/drop
corresponds to (b)

In the first method, we set up 10 wavelengths on each of the links A–B

and B–C connecting the routers at the ends of these links. We observe
that the traffic flowing on link A–B is 50 Gb/s (traffic from A–B)+ 50
Gb/s (traffic from A–C that must use link A–B) = 100 Gb/s. Similarly ,
the traffic flowing on link B–C is also 100 Gb/s. Thus the 10 wavelengths
on each of the links A–B and B–C are sufficient to carry this traffic. In
this case ,we use 10 router ports at node A, 20 router ports at node B,
and 10 router ports at node C, for a total of 40 router ports. At the
optical layer, nodes A and C have OLTs, whereas node B has a pair of
OLTs that terminate all the wavelengths passing through node B.
in second design, node B is OADM instead of OLT
 With Optical Add/drop
In the second design, we set up only five lightpaths each on the routes
A–B, B–C, and A–C. The five lightpaths on the route A–C pass through
the node B within the optical layer, without being converted to an
electrical signal. This design requires only 10 router ports at each of
the three nodes, A, B, and C, for a total of 30 router ports, compared to
40 router ports in the design without optical add/drop. However, this
design requires node B to have an OADM node that is capable of
adding and dropping 10 of the 20 lightpaths that terminate at the
router at node B, while passing the other 10 lightpaths through.
Unit 10.1:Cost Trade-Offs: A Detailed Ring
Network Example
• While designing an optical network there are different network
topologies for designing the network to meet the traffic demand of t
units.
• There are mainly three network topologies on the basis
of arrangement of router, optical links: Mesh,Hub and fully connected

• Engineers try to lower the cost of design and meeting the traffic
demand by optimising the network.
 Different
Network
Topologies
 • Router Ports: Minimising the router ports means to
minimise the number of light paths that must be set up
to handle the network.
• Wavelengths: Minimising the number of wavelengths
would reduce the additional equipment cost of the
network.
Components • Hops: Designing of transmission system becomes difficult
as the number of hops taken by light in the system
in an Optical increases.
Network However there is a tradeoff between these different
parameters.
For example, we will see that the PWDM ring uses a large
number of router ports, but the smallest possible number
of wavelengths. In the hub and all-optical design examples
that follow, we will use fewer router ports at the cost of
requiring more wavelengths.
Analysis of Network
topology: PWDM ring
Passive

• Number of
lightpaths/wavelengths: W
• Number of nodes: N
• Traffic Load: L
• Assuming shortest path
routing algorithm and single
hop light paths
• L=(N+1+1/(N-1))t/8
• W=ceil(L) t--> no of transmitters per wavelength
• Q=2W
Q = no of router ports; mutiplied with 2 because 2 on each side of router
Analysis of
Network topology:
Hub Network

• Number of lightpaths/wavelengths: W
• Number of nodes: N
• Traffic Load: L
• Number of hops: H
• Number of router ports: Q
• Assuming shortest path routing
algorithm and single hop light paths
• Q=2*ceil(T)
• W=(N/2)*ceil(t)
• H=N-1 t--> no of transmitters per wavelenth
T--> no of transmitters per node
Analysis of
Network topology: All optical design Network

• Number
of lightpaths/wavelengths: W
• Number of nodes: N
• Traffic Load: L
• Number of router ports: Q
• Assuming shortest path
routing algorithm and single
hop light paths
• Q=(N-1)*ceil(t/(N-1))
• W=ceil(t/N-1)*(N^2/8+N/4)
Comparision of the designs on the basis number of minimum
number of IP router ports vs traffic
 Comparision of the designs on the basis number of
number of wavelengths vs traffic
*large number of router ports* but *smaller number of wavelengths* for PWDM ring design
LTD (Lightpath Topology Design): Focuses on creating an optimal network layout for connecting nodes with minimal cost and high performance.

RWA (Routing and Wavelength Assignment): Involves determining the best paths for data transmission and assigning wavelengths to avoid conflicts and
ensure efficient use of resources.

10.2 LTD and RWA Problems

• The wavelength-routing network design problem can be split into LTD and RWA problems as a
heuristic.
• Different lightpath topologies have different costs and trade-offs, which can be evaluated by
solving the RWA problem for each of them.
• The next sections will discuss the LTD and RWA problems in more detail.
• The lightpath topology design problem is simplified by assuming no constraints from the fiber
topology or the optical layer.
• All lightpaths are bidirectional and each IP router has at most Δ ports to connect to other routers.
• The number of lightpaths is limited by nΔ and the cost of lightpaths is assumed to be uniform
within a region.
• The routing of packets over the lightpath topology is also part of the design problem.
 10.2.1 Lightpath Topology Design (LTD)

• The problem is modeled by using binary variables bij to indicate if there is a lightpath from node i to node j.
• The traffic between nodes is assumed to be IP packets that can be split over different paths.
• The fraction of traffic between s-d pair (s, d) that goes over link (i, j) is asd ij and the total traffic over link (i, j) is λij .
• The congestion is defined as the maximum traffic over any link, λmax .
• The problem involves finding the values of bij and asd ij that specify the lightpath topology and the routing algorithm.
• The congestion is important because it determines the queuing delay and the throughput of the network.
• The network can be modeled as a collection of M/M/1 queues with service rate μ and arrival rate λij for each link (i, j).
• The queuing delay on link (i, j) is dij = 1/(μ - λij) and the throughput is the minimum offered load that makes λmax = μ.
• The performance objective is to minimize the congestion λmax.

s,d --> source/destination

• The problem is formulated as a mathematical program with variables
λsd ij , λij , λmax, bij .
• The flow conservation constraint ensures that the net flow out of a
node equals the arrival rate for the source node, the negative arrival
rate for the destination node, and zero for other nodes.
• The definition of λij and the congestion constraint ensure that λmax is
the maximum traffic over any link.
• The lightpath existence constraint ensures that no packets are routed
over non-existing links.
• The degree constraints limit the number of links per node to Δ and
the bidirectional lightpath constraint ensures that all lightpaths are
bidirectional.
• The problem is a mixed integer linear program (MILP) which is NP-hard and difficult to solve exactly.
• Commercial packages can solve LPs, ILPs, and MILPs but they are often too slow for large problems.
linear program
• Heuristics can be used to find approximate solutions using LP-relaxation and rounding techniques.
• A feasible solution satisfies all the constraints and an optimal solution minimizes or maximizes the objective
function. The value of the program is the objective function value at the optimal solution.
• The LTD-MILP can be relaxed to an LTD-LP by allowing the bij variables to take fractional values between 0 and 1.
• The LTD-LP has the same or lower value than the LTD-MILP and its optimal solution gives a lower bound on the
value of the LTD-MILP.
• The LTD-MILP can be reduced to a routing-LP by fixing the values of the bij variables to 0 or 1 and satisfying the
degree constraints.
• The routing-LP has the same or higher value than the LTD-MILP and its optimal solution gives an upper bound on
the value of the LTD-MILP.
• A heuristic approach is to round the bij values in the LTD-LP to the closest integer without violating the degree
constraints. This is done by a rounding algorithm.
bij<=Delta
10.2.2 Routing and Wavelength Assignment
• The RWA problem is to find routes and wavelengths for a set of lightpath requests using the minimum number of wavelengths.
• The RWA problem can be modeled as an ILP but it is too hard to solve for large networks.
• The RWA problem can be split into an LR problem and a WA problem.
• The LR problem is to find routes for the lightpaths that minimize the maximum number of lightpaths on a link or some other
network cost.
• The WA problem is to assign wavelengths to the lightpaths that minimize the maximum wavelength on a link.
• The LR problem can be solved by routing the lightpaths one by one using shortest path algorithms with link weights that reflect the
network objective.
• The WA problem must ensure that no wavelength conflicts occur on a link or through a switch without wavelength conversion.
• A simple WA algorithm is first fit, which assigns the lowest available wavelength to each lightpath. This minimizes the number of
wavelengths used.
• Disjoint paths are needed to survive single link, node, or SRLG failures.
• One method to find disjoint paths is to compute the shortest path first and then another shortest path
that avoids the first path. This may not always work.
• Another method is to solve the minimum disjoint paths problem that finds disjoint paths with minimum
total weight. This is more complicated but more general.
• The protection bandwidth depends on the protection mechanism. For 1+1 and 1:1 protection, it is
dedicated.
 10.2.3 Wavelength Conversion
• Wavelength conversion can be full, fixed, or limited.
• Full conversion can change any wavelength to any other wavelength.
• Fixed conversion has a predetermined mapping between input and
output wavelengths.
• Limited conversion can change a wavelength to a subset of other
wavelengths. Limited conversion allows only a subset of wavelength changes.
Need of wavelength conversion:

Without wavelength conversion, a lightpath must use the same wavelength across the entire path from
source to destination. If the required wavelength is unavailable on any link, the lightpath cannot be
established, leading to wavelength blocking. Conversion allows lightpaths to switch wavelengths at
certain points, reducing the likelihood of blocking.
Optical Cross-Connects (OXCs)
O/E/O-->Optical-Electrical-Optica
• Fixed and limited conversion models can save on switch cost but still
need O/E/Os for each signal.
• These models are useful when all-optical wavelength converters are
available or when networks have multiple fibers and no conversion.
• Networks with multiple fibers and no conversion can be modeled as
single fiber networks with limited-degree conversion at the nodes.
• The suffixes NC, FC, C, and LC are used to denote different types of
wavelength conversion.
• The WA problem depends on the type of wavelength conversion
available in the network.
• Full conversion allows any wavelength change and makes the WA
problem trivial.
• Limited conversion allows only a subset of wavelength changes and
makes the WA problem more complex.
• Fixed conversion requires a predetermined wavelength change at
each node and makes the WA problem even more complex.
• The load of a request is the maximum number of lightpaths on a link
and it is a lower bound on the number of wavelengths needed. The
actual number of wavelengths could be much higher without
conversion.
Need of wavelength conversion
10.3- Dimensioning Wavelength-Routing
Networks
• Wavelength dimensioning is basically providing the set of wavelength
in every WDM link.
Dimensioning Wavelength-Routing Networks

• The solution of the RWA problem determines the specific set of

wavelengths that must be provided on each link to realize the
required lightpath topology, and thus solves the dimensioning
problem.
• This is the offline RWA problem since we are given all the lightpaths at
once.
• Once a network is operational, the RWA problem has to be solved for
one lightpath at a time, when the lightpath is required to be set up.
This is the online RWA problem.
Dimensioning Wavelength-Routing Networks

30 ports OXC
Dimensioning Wavelength-Routing Networks
Dimensioning Wavelength-Routing Networks
• Example 10.6 -To illustrate the flexibility obtained by using OXCs in
the network, consider the three-node linear network example again.
By replacing the static OADM in Figure 10.1(c) by a reconfigurable
OADM, or OXC, with 30 ports, we obtain the node design shown in
Figure 10.14. This design can handle any combination of traffic that
does not require termination of more than 100 Gb/s of traffic at each
node, in contrast to the design of Figure 10.1(c), which was designed
for a specific traffic matrix: 50 Gb/s of traffic between each pair of
nodes.
Dimensioning Wavelength-Routing Networks
• Solving the dimensioning problem determines not only the number of
wavelengths that need to be supported on each link, but also the
sizes of the OLTs and the OXCs.

• The size of the OXC also depends on the maximum number of

lightpaths to be terminated at each node, which corresponds to the
number of router interface cards provided at that node.