DYNAMIC CONGESTION CONTROL IN WDM OPTICAL NETWORK

Rupak Bhattacharyya et al. (Eds) : ACER 2013,
pp. 233–242, 2013. © CS & IT-CSCP 2013 DOI : 10.5121/csit.2013.3221
DYNAMIC CONGESTION CONTROL IN WDM
OPTICAL NETWORK
Sangita Samajpati1
, Abhisek Daw2
and Swapan Kumar Mondal3
1
Regent Education & Research Foundation, Barrackpore, WB, India.
eva.samajpati@gmail.com
2
Tata Consultancy Services
abhisek.daw@tcs.com
3
Kalyani Govt.Engg.College, Kalyani, WB, India.
paltaswapan@yahoo.co.in
ABSTRACT
This paper is based on Wavelength Division Multiplexing (WDM) optical networking. In this
optical networking, prior to data transfer, lightpath establishment between source and
destination nodes is usually carried out through a wavelength reservation protocol. This
wavelength is reserved corresponding to a route between the source and destination and the
route is chosen following any standard routing protocol based on shortest path. The backward
reservation protocol is implemented initially. A fixed connected and weighted network is
considered. The inputs of this implementation are the fixed network itself and its corresponding
shortest path matrix. After this initial level of implementation, the average node usage over a
time period is calculated and various thresholds for node usage are considered. Above
threshold value, request arriving at that path selects its next shortest path. This concept is
implemented on various wavelengths. The output represents the performance issues of dynamic
congestion control.
KEYWORDS
Optical network, Wavelength Division Multiplexing, Backward reservation protocol,
Congestion, Threshold.
1. INTRODUCTION
In recent years extensive research has taken place in the area of wavelength division multiplexing
(WDM) networks. Wavelength-routed networks allow building high bandwidth networks. Optical
networks employing wavelength division multiplexing (WDM) offer the promise of meeting the
high bandwidth requirements of emerging communication applications. WDM technology has
been recognized as one of the key components of the future networks. The commercialization of
WDM technology is progressing rapidly. This work focuses on the dynamic control of fully
optical wavelength routed WDM network, in other words how the arriving connections, i.e.
lightpath requests, should be configured into the network in order to maximize the expected
revenues [1].
A continuous path, having the same wavelength reserved in all the hops of the route, is called a
lightpath. Since lightpaths are the basic building block of this network architecture, their effective
establishment is crucial. It is thus important to provide routes to the lightpath requests and to
assign wavelengths on each of the links along this route among the possible choices so as to

234 Computer Science & Information Technology (CS & IT)
optimize a certain performance metric [2].The wavelengths assigned must be such that no two
lightpaths that share a physical link use the same wavelength on that link. In distributed WDM
optical networks having no wavelength conversion facility, usually a dedicated lightpath is first
established between the source node (source) and destination node (destination), before the actual
data transfer starts.
It is noted that with increase in Arrival Rate, blocking probability also increases. When more
number of requests arrives in a specific time, more is the contention for free wavelength and more
chance of getting blocked. Again with increase in the number of wavelength the blocking
probability decreases as more wavelengths allow more requests to successfully select and reserve
the selected wavelength. Thereby, successful data transmissions occur. But assigning more
wavelength in a network increases the cost. Few protocols have been used for reducing the
blocking probability. Among the various protocols, backward reservation protocol helps in
reducing the chance of getting blocked to a significant extent. But with growing demand for
wavelength due to increase in number of requests, backward reservation protocol also shows the
increase in blocking probability. A brief description of existing backward reservation protocol
with its problem has also been studied.
It has been observed in backward reservation protocol that when traffic fight for the selection of
wavelength through their shortest path resulting in congestion when some nodes belong to the
shortest path of a large number of requests, the wavelength available in the next shortest path
remain unused [3].With a good solution of this problem, more customers can be accommodated
by the given system, and fewer customers need to be rejected during periods of congestion.
2. OBJECTIVE
Implement a protocol to dynamically reduce the congestion problem in wavelength assignment in
WDM network and to maximize the resource utilization as well as minimize the cost. Also, the
optimization goal is to minimize some infinite time horizon cost functions such as blocking
probability.
3. BACKWARD RESERVATION PROTOCOL (BRP)
Lightpath establishment in a distributed environment requires a control protocol that will be
responsible for a lightpath establishment and teardown facilities. These protocols can be
categorized based on their working mechanisms and whether they require the use of global state
information or not.
Parallel reservation scheme requires global state information in particular topology and
wavelength usage information on each of the links in the network to work. A separate control
message is sent from source node to each of the nodes in the path from source to destination for
wavelength reservation. After a successful reservation of wavelengths by each node along the
path it will send separate acknowledgement messages to the source node.
In BRP scheme control message (PROB) originated from the source node travels from source to
destination on a hop by hop basis collecting wavelength usage on each node in the route. When
this control message reaches at the destination node it selects a free wavelength from a set of free
wavelengths along the entire path and then initiates a reservation message (RES) in the reverse
direction from destination to source. This message reserves that particular wavelength on each
node on the reverse path from destination to source [4]. However, it suffers from “outdated
information”. Fig. 1 shows the principle of BRP.

Computer Science & Information Technology (CS & IT) 235
Fig. 1: Backward Reservation Protocol
4. CONGESTION CONTROL
For every node pair, there may exist more than one feasible route and wavelength. An optimal
algorithm should select a route and wavelength pair such that more requests can be
accommodated into the network. Though this optimality criterion cannot be addressed directly, it
is addressed in couple of indirect ways. Minimum amount of resources are used to establish the
current request, so as to maximize the amount of free resources available in the network for future
requests [5]. However, this may result in increasing the congestion in the network. In an alternate
way of addressing this problem, algorithms try to balance the load in the network by selecting the
next shortest path.
A network is said to be congested when the users of the network collectively demand more
resources than the network has to offer. Congestion in a network leads to delayed packet arrivals,
packet droppings and even denial of service for certain hosts [6].
4.1. Congestion and the need to control it
Ideally we want unlimited bandwidth for every user connected to the network. Reality however is
different. We have limitations to the available bandwidth. Limitations that are not only because of
the transport protocol but also of the medium for transferring informations (copper wires,
electrical transmitters etc). Since the resources are limited and users and their demands are
exponentially increasing we need to control the way network resources are used. We want to send
as much traffic as possible through a link to maximize our throughput. On the other hand if we
send at excessive rates then we will experience congestion. The network throughput will
dramatically drop and we may even have a collapse. To avoid such situation congestion control is
needed [7]. The need for new technology and network independent, congestion control algorithms
is more demanding than ever. To satisfy these needs we must consider a new totally different
approach.
4.2. Working Principle of Dynamic Congestion Control
The schemes considered under dynamic traffic can be divided into two cases reconfigurable and
non-reconfigurable. If it is possible to reconfigure the whole network when blocking would
occur, the blocking probability can be considerably reduced. Such an operation, however,
interrupts all (or at least many) active lightpaths and requires a lot of coordination between all the
nodes.

In large networks the reconfiguration seems infeasible. In any case, the reconfiguration algorithm
should try to minimize the number of reconfigured lightpaths in order to minimize the amount of
interruptions in the service. The other case occurs when active lightpaths may not be
reconfigured. In this case it is important to decide which route and wavelength are assigned to an
arriving connection request in order to balance the load and minimize the future congestion in the
network.
In general, one is interested in the optimal policy which maximizes or minimizes the expectation
(infinite time horizon) of a given objective function. The objective is defined in terms of
maximizing or minimizing some revenue or cost function. The cost may represent e.g. the loss of
revenue due to blocked calls, where different revenue may be associated to each type of call.
5. THRESHOLD IN NETWORK
Based on average node usage, a value can be considered so that high usage of a node can be
prevented. This value can be referred as threshold.
Congestion occurs when the traffic and resource mismatch at some points of the network. It is
therefore usually local and changes rapidly. Changing routing table, which involves information
exchange and route computation across the entire network, is often too costly and also too slow.
When congestion does occur, the routing algorithm has to wait until next updating period to
respond. At that time, the congestion may have already dissipated [8].And the reported high
distances caused by the congestion, which are no longer valid, nevertheless has a misleading
effect on the traffic that shares the same link when the next route updating is due.
As the traffic approaches the capacity of a path, it is often impossible for a single-path routing
algorithm to derive satisfactory routes. It is necessary to provide some degree of load sharing by
using multi-path routing.
Although most multi-path routing algorithms are too complex to be used in an adaptive fashion,
some simple form of multi-path routing can be easily adopted and can still improve the
performance. To make full use of the multi-equal-path routing with traffic-sensitive metrics, it
may be necessary to transform the values of link distances into a set of discrete values. When the
traffic is heavy and changing rapidly, it is difficult for the routing algorithm to determine the
entire routes based on the past link distance information. It has to allow continuous changes in
routing decisions. When the traffic accumulated in some areas of the network exceeds a
threshold, an alternate route can be provided to ease the congestion [5].
Fig. 2: Selection of new route due to congestion
In network, the available shortest path selection algorithms were proposed and evaluated. The
shortest path selects the shortest hop path among the all paths [4]. It is important to note that
though every link on a path computed has free wavelengths, a connection may still get blocked

due to unavailability of a single continuous free wavelength. The available shortest path routing
algorithm considers wavelength availability while computing a route. When a request arrives in a
particular node, it compares the node’s wavelength usage with a given threshold value & if the
usage is above of threshold then it chooses the next shortest path by the shortest path algorithm.
In this work not only considered wavelength availability but also link congestion information to
dynamically compute a route. Fig. 2 shows that a request has to change its path as one of the
nodes’ usages is greater than the threshold value in which “Nu” represents the node’s usages and
“T” represents threshold value.
6. SIMULATION OF RESULTS
In order to evaluate the proposed protocol with different thresholds and compare it with backward
reservation protocol (BRP) varying we had to implement the proper test environment. In this
section we see how that was accomplished and then comment on the results that were gathered
from our simulations.
6.1. Implementation
In order to simulate the proposed protocol the following steps are considered.
Step – 1: Generation of shortest path: Implementation of the input to Floyd’s algorithm with
random fit approach and obtaining the output of Floyd’s algorithm.
• Input: Numbers of the nodes in the graph and percentage of connectedness in the graph.
• Output: Randomly generated connected, weighted network, all pairs shortest distance
matrix and path matrix.
Step – 2: Generation of requests: Generation of Requests with a mean arrival rate, randomly
generated source and destination pairs. Queuing of the requests is done.
Step – 3: Probing: A source sends a PROB packet to its destination to gather information about
wavelength usage without locking any wavelength. The PROB packet carries a bit vector to
represent the set of wavelength available to establish the connection.
Step – 4: Decision based on Average node’s usage: After arriving at each node in the
designated path, the request checks the average usage of the node.
Step – 5: Compare with threshold: If the average usage is above threshold value, the next
shortest path for that request is calculated by Floyd Algorithm and execution again begins from
(3), else it proceeds through that path.
Step – 6: Reservation: The destination picks up one of the wavelengths randomly from the pool
of available wavelengths and reserves the selected wavelength by traversing in the reverse order.
Step – 7: Unavailability of wavelength: In case of unavailability of wavelength during
reservation, the request is blocked.
Step – 8: Data Transmission: After successful reservation, data transmission takes place
completely.
Step – 9: Simulation: Simulation of the relation of blocking probability of connection request
with mean arrival rate of request, wavelength and threshold of node’s usage in a network.

7. SIMULATION RESULTS AND DISCUSSION
Simulation of the relation of blocking probability of connection request for varying mean arrival
rate of request with different thresholds for different wavelengths, varying thresholds (TH) with
various mean arrival rate and for number of wavelengths per channel using random fit approach
in a network. Arrival of connection request follows poison arrival process. In this simulation the
input is the shortest distance matrix and shortest path matrix (obtained following Floyd’s
algorithm) corresponding to the fixed network.
7.1 Analysis Based on Arrival Rate (AR):
In Fig. 3 – 5, the blocking probability of connection request for varying mean arrival rate of
request with different thresholds is shown. The x-axis represents the arrival rate of a connection
request and the y-axis represents the blocking probability of the request.
Fig. 3: Blocking Probability vs. Arrival Rate with wavelength = 100
The Fig. 3 shows the performance of the proposed protocol with respect to backward reservation
protocol .The number of wavelength channel is set to 100. Threshold 40, 50 does not show better
result than brp as when threshold is too low most of the requests are forced to change their paths
and they in turn increase the nodes' usage in the changed path. The graph shows two natures. One
is upto arrival rate 500 and another above arrival rate 500.So it is better to study them separately.
Fig. 3a: Blocking Probability vs Arrival Rate upto AR = 500

Fig. 3b: Blocking Probability vs Arrival Rate above AR = 500
Thresholds 80, 90 shows better result here as most of the requests are allowed to go through the
original path. From Fig. 3a it can be seen that till arrival rate 500, threshold 90 shows better
result. While from Fig. 3b, it is seen that above AR 500 threshold 80 is better as in case of 90, the
chance of blocking increases during reservation. In threshold 80, however as comparatively less
number of requests is allowed to probe through original path, less chance of getting blocked
during reservation.
Fig. 4: Blocking Probability vs. Arrival Rate with Wavelength = 200
Fig. 5: Blocking Probability vs. Arrival Rate with wavelength = 300

Fig. 4 & Fig. 5 show the blocking probability when wavelength is 200 and 300 respectively with
varying different thresholds. Low threshold 40 does not show better result than BRP as when
threshold is too low most of the requests are forced to change their path and they in turn increase
the nodes' usage in the changed path. Higher thresholds 80, 90 shows better result here as most of
the requests are allowed to go through the original path. Till arrival rate 400, threshold 90 shows
better result .above AR 500 threshold 70-80 are better as in case of 90, the chance of blocking
increases during reservation. In threshold 70-80, however as comparatively less number of
requests is allowed to probe through original path, less chance of getting blocked during
reservation.
7.2 Analysis Based on Wavelength:
In case of analysis based on wavelength we have studied the performance of wavelengths with
respect to brp varying different thresholds with a fixed arrival rate 300 i.e. high arrival rate.
Fig. 6: Blocking Probability vs. Wavelength with mean Arrival Rate = 300
We can see in Fig. 6 in where threshold 80 clearly shows better result than other thresholds as
pointed in other curves. With increase in wavelength , brp decreases as in normal case .
7.3 Analysis Based on Threshold:
The graphical analysis based on thresholds shows the blocking probability of a
connection request with respect to thresholds varying different wavelengths.
Fig. 7: Blocking Probability vs. Arrival Rate with Wavelength = 100

Fig. 8: Blocking Probability vs. Thresholds with Wavelength = 200
From Fig. 7 and Fig. 8 it can be conferred that when the threshold increases the blocking
probability decreases. But after a certain point the Blocking probability increases. Threshold 80
shows better result than other thresholds.
The requests follow the path obtained by using shortest path routing protocol. Some nodes may
be used frequently as they belong to shortest paths of many requests. As a result the link usages
of those nodes are very high. So, the pool of free wavelengths for that link gets depleted and
resulting in blocking of few requests. But those requests can use other sub-optimal paths and
successful data transmission can occur.
The threshold value needs to be determined accurately. As, some inaccurate selection of threshold
like in case of low threshold values, might result in chaotic and abnormal situation as requests,
being forced to select next shortest path, nodes belonging to next shortest path might be in some
request’s optimal path and hence increases that node’s usage. So, it has been found from analysis
of all the results that using a high threshold value in the range of 70-90 one can achieve a better
result
Again a situation might arise when the inter-connections are very low. So, many sub-optimal
shortest paths may not exist as a result requests get blocked. Selection of a threshold value is very
important.
8. CONCLUSIONS
In this work, however, we have considered only the wavelength routed networks. Nonetheless,
the promising future of WDM networks makes them an interesting topic for the research. The
objective was to search for a way to reduce the node congestion in a heavy busty traffic in the
area of optical networking. Because of the large amount of activity on this subject over the past
dozen years, capturing the important and most significant developments in a concise format is a
challenging task; thus, apologize if we inadvertently left out some contributions which could be
considered to be more important. We have introduced a backward reservation protocol and
implemented a solution to control the congestion in WDM network by lowering the blocking
probability. This is accomplished by allowing the nodes that are considered to be the main
decision points for selecting the shortest path and also try another shortest path if the first choice
does not succeed.
It has been observed that for comparatively low arrival rates threshold 80 or 90 gives better result.
But as the arrival rate is increased it has been observed that threshold 70 or 80 provide better
result compare to 80 or 90.

More study can be done on the effect of other parameters like use of control packets, setup time,
time period etc. Piggybacking of intermittent control packets with link usage information for
updating the network to reduce the effect of outdated information of wavelength usage in the
network can also be implemented. Results can be made more statistically stable. Concept of
threshold can be tested in other protocols also. The running time of the algorithm can also be
reduced.
REFERENCES
[1] Zang, H., Jue, J., Sahasrabuddhe, L., Ramamurthy, R., & Mukherjee, B., (2001) “Dynamic lightpath
establishment in wavelength–routed WDM networks”, IEEE Communication Magazine, Vol. 39, No.
9, pp. 100–108.
[2] Ozdaglar, Asuman E., & Bertsekas, Dimitri P., (2003) “Routing and Wavelength Assignment in
Optical Networks”, IEEE/ACM Transactions on Networking (TON), Vol. 11, No. 2, pp. 259–272.
[3] Todimala, A., & Ramamurthy, B., (2003) “Congestion-based Algorithms for Online Routing in
Optical WDM Mesh Networks”, Proceedings of Second IASTED International Conference on
Communications, Internet, and Information Technology.
[4] Zhang, H. & Mukherjee, B., (2003) “Lectures on Routing and wavelength assignments for
Wavelength-routed WDM Networks”, Internetworking Technologies Handbook, Vol. 2, pp. 1 – 20.
[5] Thodime, G.P.V., Vokkarane, V. M., & Jue, J. P., (2003) “Dynamic congestion-based load balanced
routing in optical burst-switched networks”, In GLOBECOM 2003 – IEEE Global
Telecommunications Conference, pp. 2628–2632.
[6] Jonnadula,Venkata R., (2005) “Performance Analysis of Congestion Control Schemes in OBS Mesh
Networks”.
[7] Rossides, L., (1999) “Congestion Control in Integrated Services Networks”, Masters Thesis, Dept of
Computer Science, University of Cyprus, pp. 1–64.
[8] Wang, Z., & Crowcroft, J., (1992) “Analysis of Shortest-Path Routing Algorithms in a Dynamic
Network Environment”, ACM SIGCOMM Computer Communication Review, Vol. 22, Issue 2, pp.
63–71.
AUTHORS
Sangita Samajpati received his B.Tech. and M.Tech. degrees in Computer Science
Engineering from the Kalyani Govt. Engineering College, Kalyani, West Bengal, India.
She is currently an Assistant Professor at Regent Education & Research Foundation. Her
research interests include Optical Network, Fuzzy Logic and soft computing Technique.
Abhisek Daw received his B.Tech. and M.Tech. degrees in Computer Science
Engineering from Institute of Technology and Marine Engineering, West Bengal, India
and Kalyani Govt. Engineering College, Kalyani, West Bengal, India respectively. He is
currently employed as an Assistant System Engineer in Tata Consultancy Services. His
research interests include Optical Network, Fuzzy Logic and soft computing Technique.
Dr. Swapan Kr. Mondal received his Ph.D. degree from Jadavpur University, West
Bengal, India. He did B.E. and M.Tech. degrees in Electrical Engineering and Computer
Science Engineering from Jalpaiguri Government Engineering College, West Bengal,
India and Rajabazar Science College, University of Calcutta, West Bengal, India
respectively. Also he has M.B.A. in portfolio from IGNOU. Currently he is a Professor at
Kalyani Govt. Engineering College, Kalyani, West Bengal, India and he has more than 23
years of teaching experience. He has some research publications in different reputed
Journals and Conferences. His research interests include Computer Networking, Optical Networking and
Computer Architecture.

DYNAMIC CONGESTION CONTROL IN WDM OPTICAL NETWORK

More Related Content

What's hot (19)

Similar to DYNAMIC CONGESTION CONTROL IN WDM OPTICAL NETWORK (20)

More from cscpconf (20)

Recently uploaded (20)

DYNAMIC CONGESTION CONTROL IN WDM OPTICAL NETWORK