Академический Документы
Профессиональный Документы
Культура Документы
INTRODUCTION
1
related problem is converter allocation determining the number of converters each node in a
network can have in order to optimize network performance for a given number of
converters. This problem and a solution based on converter utilization information at various
nodes are discussed.
2
2.NEED FOR WAVELENGTH CONVERTERS
In wavelength route networks, light paths are used to carry messages. A message is
transmitted between the end nodes of a light path without undergoing any electronic-optical
conversion at intermediate nodes. In other words, the message is carried in optical form from
the source node to the destination node. This results in shorter message propagation time.
Also, the need for large buffers at intermediate nodes is eliminated. On the other hand, it has
the drawback of higher connection blocking probability. A lightpath must be wavelength-
continuous, If a route is free but no common wavelength is available on it, then it cannot be
used by a light path. This results in blocking a connection which would otherwise have been
accepted if there were no such wavelength continuity constraints. The question which
naturally arises is: Is there any means by which we can have all the advantages- fully optical
message transmission, no need for intermediate buffering with no penalty in network
performance? Wavelength converters are a viable solution that offers all the above
advantages.
3
Example:-
Let us consider a network with four nodes and two wavelengths per fiber as shown in figure.
The figure shows two copies of network, one for each wavelength. The vertical edge from a
node on one wavelength to the same node on a different wavelength signifies that it is a
converting node. Assuming that connections are required between nodes<0, 2>, <1, 3> and
<2, 1>.Let p1, p2 and p3 be the connections that corresponds to these node pairs. The routes
that can be used by p1, p2 and p3 are 0→1→2, 1→2→3 and 2→3→0→1, respectively. It can
be observed that all three connections cannot be established in a network with no converting
nodes. If p1 uses wavelength w0 then p2 cannot use w0, as they share a common link 1→2.
Consequently, p2 will use wavelength w1. Now, p3 cannot use w0, as it is not free on link
0→1. Also, it cannot use w1, as it is not free on link 2→3. Therefore, p3 cannot be established
because the route is not wavelength continuous, even though it is free.
The connection p3 can be established along with the other two connections, p1 and p2, if node
0 has a converter which can convert from w0 to w1. It can use w0 on links 2→3 and 3→0, and
w1 on link 0→1, as shown in figure. It can be noted that a message on this lightpath
undergoes wavelength conversion from w0 to w1 at node 0. While the connections p1 and p2
are realized by lightpaths, p3 is realized by a semi-lightpath.
4
A wavelength converter is said to have full degree of wavelength conversion if it is capable
of shifting any wavelength. Otherwise, it is said to have partial degree of wavelength
conversion. A converter is said to have partial degree of wavelength conversion. A converter
is said to have conversion degree D if it can shift any wavelength to one of D wavelengths .A
converter which is capable of shifting a wavelength I to a range of wavelengths from i-k to
i+k is called a limited range converter. Wavelength converters are expensive.
There are two kinds of wavelength conversion: optical electronic conversion (O-E) and all
optical wavelength conversion (A-O).This classification is based on how wavelength
conversion is realized.
In A-O conversion, the conversion is realized in a fully optical way by means of optical
wavelength converter devices. They include absorption saturation of a semiconductor laser
diode, four-wave mixing, and gain saturation of an optical amplifier. The A-O method is
faster, and it increase network speed to a greater extent than the optical-electronic method.
5
3. ENABLING TECHNOLOGIES
Significant advances in optical and electronic device technologies have made wavelength
conversion feasible. Several different techniques have been demonstrated to perform
wavelength conversion. Many novel switch designs have been proposed for utilizing these
wave- length converters in a wavelength-convertible network.
A. Wavelength-Converter Design
The function of a wavelength converter is to convert data on an input wavelength onto a
possibly different output wavelength among the wavelengths in the system. In the figure,
.
and throughout this section, denotes the input signal wavelength; , the output
(converted) wavelength; , the pump wavelength; , the input frequency; , the
converted frequency; , the pump frequency; and , the continuous wave generated as the
signal.
6
transparency achievable is digital transparency, where digital signals of any bit rates up to a
certain limit are accommodated.
[The extinction ratio is defined as the ratio of the optical power transmitted for a bit 0 to the
power transmitted for a bit 1].
2) All-Optical Wavelength Conversion: In this method, the optical signal is allowed to
remain in the optical domain throughout the conversion process. Note that, in this context,
all-optical, refers to the fact that there is no O/E conversion involved. Such all-optical
methods can be further divided into the following categories-
7
4. WAVELENGTH-CONVERIBLE SWITCH ARCHITECTURES
8
An aggregate message received on an incoming fiber is first demultiplexed into different
messages, each on a different wavelength. Each message corresponds to a unique connection.
These messages are sent to the appropriate wavelength switch. Each wavelength is switched
to one of the output ports connected to the desired output link by the optical switch. The
output message of the optical switch may have its wavelength shifted by the wavelength
converter. Finally, several messages, each on a different wavelength, which are outbound to a
particular output link, are wavelength-multiplexed into an aggregated message coupled to the
output link. Since for every link and wavelength there is a converter, no connection request
will be blocked due to the nonavailability of a wavelength converter at a node. For example,
if at a node, wavelength wi is available on its input link and wj is available on its output link,
there will always be a free converter at the node that can be used by a connection. This results
in increased acceptance rate of connection requests. However, this architecture requires many
wavelength converters, thus increasing the cost of the switch. In practice, at any instant of
time, the number of connections that require wavelength converters at a node is much less
than DW. Therefor this switch may not use the converters effectively. A cost-effective
solution is to use only a few converters and allow their sharing. The share-per-node WC
switch and share-per-link WC switch differ in the way the converters are shared.
The share-pre-node wavelength-convertible switch uses two optical switches and a
wavelength converter bank (WCB) as shown in figure. A wavelength converter bank is a
collection of a few wavelength converters. Assume that there are k converters in the WCB.
The first optical switch is a DW× (DW+k) nonblocking switch and the second switch is a
k×Dk nonblocking switch. The aggregate message arriving at an input link is demultiplexed
into different wavelengths. These wavelengths are fed to the first optical switch. If
wavelength does not require conversion, it is fed directly to the multiplexer associated with
the desired output link. Otherwise, it is fed to the converter bank. The second optical switch
directs the converted wavelength to the multiplexer associated with the desired output link.
All the messages which are outbound to an output link are multiplexed into an aggregate
message by the multiplexer, before it is sent. Here, the wavelength converters are shared
efficiently. This architecture is cost-effective in the sense that the performance-cost ration is
high. The performance is good, although it is lower than that yielded by the other
architectures. The optical switching complexity is higher when compared to other
architectures.
9
Fig:-The share per node WC switch architecture
10
Fig:- The share per link WC switch architecture
All the message which are outbound to an output link are multiplexed into an aggregate
message by the multiplexer, before it sent. Here, the sharing efficiency of wavelength
converters is medium, as a converter bank is dedicated to a particular output link only. The
cost is low, as it uses fewer numbers of converters. The optical switching complexity is
higher than that of the dedicated-switch and is lower than for the share-per-node switch. In
Table the three different convertible switch architectures are compared.
11
5. ROUTING IN CONVERTIBLE NETWORKS
12
6.PERFORMANCE EVALUATION OF CONVERTIBLE NETWORK
Recently, several attempts have been made to evaluate the performance of WDM networks
with and without wavelength conversion through analytical models. In this section we give an
overview of the models developed. In the next section we present a detailed discussion of a
model developed in. The analytical models available in the literature differ in their
assumptions about the correlation of load on successive link, use of a wavelength on a link in
relation to the use of other wavelengths, and the route and wavelength selectin methods. The
above criteria are discussed below-
Link load correlation- The correlation between load or wavelength use on successive link is
termed as link load correlation. If the correlation is not taken into consideration, then the
model is termed a link interdependence model. The link independence model is not accurate
for WDM networks and in particular for sparsely connected is high. Consequently, the link
independence model overestimates the blocking probability.
Wavelength use on a link- A model which makes an assumption that a wavelength is used
on a link with a fixed probability independent of the other wavelengths is known as a
wavelength independence model. Like the link independence model, this model is inaccurate
and overestimates the blocking probability.
Route selection method- The route selection method assumed by a model greatly affects the
result of the blocking probability computation. The possible methods are fixed routing, fixed
alternate routing, and least-costed-path routing.
Wavelength selection method- Like the route selection method, the wavelength selection
method assumed by a model also greatly affects the result of the blocking probability
computation. The possible wavelength selection methods are random and first-fit.
13
7. NETWORK DESIGN, CONTROL, AND MANAGEMENT ISSUES
2) Sharing of converters: Even among the switches capable of wavelength conversion, it may
not be cost-effective to equip all the output ports of a switch with this capability. Designs of
switch architectures have been proposed (see Section II-B) which allow sharing of converters
among the various signals at a switch. It has been shown in that the performance of such a
network saturates when the number of converters at a switch increases beyond a certain
threshold. An interesting problem is to quantify the dependence of this threshold on the
routing algorithm used and the blocking probability desired.
14
shorter wave- length. Moreover, since the signal quality usually worsens after multiple
such conversions, the effect of a cascade of these converters can be substantial. The
implications of such a device on the design of the network need to be studied further. Apart
from efficient wavelength-convertible switch architectures and their optimal placement,
several other design techniques offer promise. Networks equipped with multiple fibers on
each link have been considered for potential improvement in wavelength-convertible
networks and suggested as a possible alternative to conversion. Another important problem
is the design of a fault-tolerant wavelength-convertible network. Such a network could
reserve capacity on the links to handle disruptions due to link failure caused by a cut in
the fiber. Quantitative comparisons need to be developed for the suitability of a
wavelength-convertible network in such scenarios.
15
the fixed path and which, if any, lightpaths to block selectively. However, design of efficient
routing algorithms which incorporate the limitations still remains an open problem.
2) Static RWA: In contrast to the dynamic routing problem described above, the static RWA
problem assumes that all the lightpaths that are to be set up in the network are known
initially. The objective is to maximize the total throughput in the network, i.e., the total
number of lightpaths which can be established simultaneously in the network. An upper
bound on the carried traffic per available wavelength has been obtained (for a network with
and without wavelength conversion) by relaxing the corresponding integer linear program
(ILP) .Several heuristic-based approaches have been proposed for solving the static RWA
problem in a network without wavelength conversion . A special case of the static RWA
problem, in which all the lightpath requests can be accommodated, is discussed for networks
with limited wavelength conversion. Again, efficient algorithms which incorporate the
limitations for a wavelength-convertible network are still unavailable.
C. Network Management
Wavelength conversion may be used to promote interloper- ability across subnetworks which
are managed independently. Thus, it supports the distribution of network control and
management functionalities among smaller subnetworks by allowing flexible wavelength
assignments within each subnet- work. Network operators 1, 2, and 3 manage their own
subnetworks and may use wavelength conversion for communication across subnetworks. In
,the authors propose to install wavelength converters at the borders between non overlapping
network partitions in the Cost 239 European optical network.
16
8.NETWORK WITH SPARSE WAVELENGTH CONVERSION
In this section we explain the analytical model developed to study blocking performance in
the networks with spares wavelength conversion, where the converters with full degree of
wavelength conversion are placed only at a few nodes. We first present an analytical model
called the Markovian link load correlation (MLLC) model to compute the call blocking
probability in a network with no wavelength conversion capability. Then we extend this
model to compute the call blocking probability in a network with a certain number of
converters.
17
9.A MODEL FOR NETWORK WITH NO CONVERSION
We now explain the MLLC model to compute the mean call blocking probability in a
network with no wavelength converters. The model is attractive and useful because it is
accurate, taking into consideration the link load correlation with moderate computational
complexity. Also, it does not make the wavelength independence assumption. The fixed
routing and random wavelength selection methods are used. The assumptions made by this
model are summarized below.
The model starts by considering a two-hop path deriving the conditional free wavelength
distribution on the path. We use “hop” and “link” interchangeably in this section. The states
of a two-hop path can be modelled using a three-dimensional Markov chain as follows.
Consider Figs Let Cl be the number of calls that enter the path at node 0 and leave at node 1.
Let Cc be the number of calls that enter the path at node 0 and continue on to the second link.
And let Ce be the number of calls that enter the at node 1. Therefore, the number of calls that
use the first link Cl+Cc and the number of cals that use the second link is Cc+Ce. Since the
number of calls on a link cannot exceed the total number of available wavelengths W, we
have Cl+Cc ≤ W and Cc+Ce ≤ W.
18
10.USEFULNESS OF CONVERTERS
The performance improves rapidly as the conversion density increases from zero, but the rate
of improvement slows down with increasing conversion density, The conversion density
beyond which the performance improves only marginally depends on the number of
wavelengths and the network load. This suggests that placing a converter at every node is
rarely necessary, if at all. In general, wavelength converters are more effective when the
number of wavelengths is larger and when the load is lower.
19
11.CONVERTER PLACEMENT PROBLEM
The previous section discussed how the performance of a convertible network in terms of the
connection blocking probability can be predicted for a given number of converting nodes. In
this section we study another important problem known as the converter placement problem.
This problem can be stated as follows.
Given K number of converters, and a network with N≥K number of nodes, how do we choose
K nodes such that placing the converters at these nodes yields the best performance
achievable using K converters?
The converter placement problem for an arbitrary network is NP-complete. The topology of a
network and the traffic pattern greatly affect the optimal solution to the placement problem.
Some of the factors which affect the optimal converter placement are discussed below-
1. A node with heavy transit traffic could be a possible candidate for converter
placement. However if does very little missing (or switching) of traffic, it may not be
a good choice for an optimal solution.
2. A node which mixes a significant amount of traffic may not be a good candidate for
an optimal solution, if the transit traffic at this node is les.
3. The distance between two converting nodes is a significant factor in obtaining an
optimal solution. The blocking probability increases with increasing distance between
two converting nodes.
20
12.CONVERTER ALLOCATION PROBLEM
The converter placement problem we discussed earlier was concerned with selection a gew
nodes which can use wavelength converters, so as to optimize some performance matric such
as connection blocking probability. As a result, some nodes in a network have conversion
capability while others do not. Usually, it is assumed that enough converters are available at a
converting node so that no connection request is blocked due to the nonavailability of
converters at a converting node. This assumption is valid when dedicated WC switch
architecture is used. This architecture uses a number of converters equal to the number of
outgoing fiber times the number of wavelengths per fiber. For other architectures, such as
share-per-node and share-per-link, only a limited number of wavelength converters is
assumed to be available a converting node. Here, determining the number of wavelength
converters to be used at every node in the network is mandatory to optimize the network
performance for a given number of converters. This leads to the converter allocation problem,
which is formally defined below.
Given a network with a fixed number with a fixed number of wavelengths per fiber, traffic
demanded pattern between node pairs, and C number of wavelength converters, the converter
allocation problem is to distribute the converters over the network nodes so as to optimize a
certain network performance metric such as connection blocking probability.
The converter allocation problem is an important and valid problem in WDM networks, as
some of the nodes are required to handle heavier volumes of traffic. This is because the
network topology is usually irregular and the traffic demand pattern between ode pairs is
often no uniform. Therefore, allocating more converters to nodes that handle a heavy volume
of traffic will probably result in improved network performance. Solutions to this problem
will vary depending upon the kind of architecture and the kind of wavelength converters
used. The architecture can be either share-per-node or share-per-link. The wavelength
converters can have either full or partial degree of wavelength conversion. In other words,
full-degree wavelength converters (FWCs) or partial-degree wavelength converters (PWCs)
can be employed.
21
SUMMARY
In this chapter several important problems in wavelength convertible networks were studied.
The benefit of employing wavelength converters was explained with an illustration. Three
different wavelength-convertible switch architectures, dedicated, share-per-node, and share-
per-link, were explained, and a comparison study was made based on criteria such as
performance and cost.
The routing problem was then addressed. A procedure to model a convertible network as a
graph was described. A shortest-path-finding algorithm on this graph was presented and its
computational complexity was studied. A brief description of some analytical models to
predict the performance of convertible networks was given and the performance results were
discussed. A detailed discussion of networks with sparse wavelength conversion and an
analytical model for such networks was presented. The use fullness of converters and how
they are affected by network topological connectivity were discussed.
The converter placement problem was studied. Some optimal solutions and heuristic solution
for placement of converters on a path were presented. Finally, the converter allocation
problem was studied. The importance of the converter allocation problem and a solution
based on converter utilization information at he nodes obtained by simulation experiments
were explained.
22
REFRERENCES
BOOKS:
C.Siva Ram Murthy ,Mohan Gurusamy, “WDM OPTICAL NETWORKS, Concept design and
algorithms” ,Department of Computer Science and Information Engineering Indian Institute of
Technology, Madras, India, 2011; page no. 107-144.
WEBSITES:
htp://digitalcommons.unl.edu/csearticles/68
htp://digitalcommons.unl.edu/csearticles
https://www.researchgate.net/publication
ieeexplore.ieee.org/document
https://www.collectionscanada.gc.ca
23