See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254020041

Lightpath-level active rerouting algorithms in all-optical WDM networks

with alternate routing and traffic grooming

Article · February 2012

DOI: 10.1109/ICOIN.2012.6164346

CITATIONS READS

18 33

2 authors, including:

Sheng-Wei Wang
Fo Guang University
16 PUBLICATIONS   86 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Sheng-Wei Wang on 27 October 2015.

The user has requested enhancement of the downloaded file.

 Lightpath-Level Active Rerouting Algorithms in
All-Optical WDM Networks with Alternate Routing
and Traffic Grooming
Sheng-Wei Wang Chin-Yen Wen
Department of Applied Informatics Department of Applied Informatics
Fo Guang University Fo Guang University
Yilan 26247, TAIWAN Yilan 26247, TAIWAN
email: [email protected] email: [email protected]

Abstract— This paper proposes two lightpath-level active assigned two or more wavelengths on different links, one or
rerouting algorithms, namely, the least resources rerouting algo- more nodes along its routing path must have the capability
rithm and the load balanced rerouting algorithm, in all-optical of converting the signal from one wavelength to another
WDM networks with alternate routing and traffic grooming. The
proposed algorithms consist of three major components to be wavelength. This paper considers all-optical WDM networks
determined: i) when the rerouting algorithm initiates, ii) which without wavelength conversion. The proposed approach and
lightpath is rerouted, and iii) which routing path and wavelength alternate routing algorithms can also be applied to all-optical
for the rerouted lightpath is newly allocated. The proposed active WDM networks with wavelength conversion.
rerouting algorithms initiate the rerouting procedure when a In an all-optical WDM network without wavelength con-
connection leaves and a lightpath is released. The lightpath to be
rerouted is selected according to the objectives of two different version, a connection must be assigned the same wavelength
algorithms. The routing path and wavelength allocated to the on all links along its routing path. Otherwise, the connection
rerouted lightpath is that the departure connection released. request is blocked. This is known as the wavelength continuity
Simulation results show that the proposed load-balanced active constraint [2]. A wavelength which is available on all links
rerouting algorithm yields much lower connection blocking along a routing path is referred to as a common available
probability than the least resources rerouting algorithms. We
also observe that the number of rerouted lightpaths is very small. wavelength along the routing path. A connection that is
This result implies that the overhead for rerouting a lightpath is assigned the same wavelength on all links along its routing
small and the performance of the networks can be significantly path is called a lightpath [2].
improved. The amount of bandwidth provided in a lightpath can be
over several gigabits per second, e.g., OC-48, OC-192, and
I. I NTRODUCTION
OC-768. However, the bandwidth requirement of a connection
A number of multimedia applications such as video on is always much smaller than the bandwidth provided. The large
demand, video conference, distance education and etc. require discrepancy between the provided bandwidth and bandwidth
a vast amount of bandwidth. The demand for networks with requirement results in the waste of bandwidth and high con-
high bandwidth is increasing. An optical fiber can provide nection blocking probability if a lightpath can be used by a
a large amount of bandwidth (nearly 50 Tb/s [16]) by us- single connection only . To solve the problem, traffic grooming
ing wavelength division multiplexing (WDM) technology [1], technology [5] is proposed. Traffic grooming technology is
[16]. Wavelength division multiplexing (WDM) technology is able to multiplex multiple low rate connections into a high
able to divide the vast bandwidth of a fiber into a number speed lightpath such that the bandwidth of a lightpath can be
of high-speed channels each of which is located at different used more efficiently [5].
wavelength. In this paper, we consider alternate routing algorithms.
Optical WDM networks are promising transport networks When a network uses alternate routing algorithm, a lightpath
for providing vast amount of bandwidth. If some or all of the may be established along a longer routing path due to wave-
switches in an optical WDM network are electronic switches, length continuity constraint or lack of wavelength resources
the bandwidth of the optical network is limited by the speed of on the shortest routing path. Using longer routing paths for a
the optical-electrical-optical (O-E-O) conversion. All-optical connection may lead to more bandwidth usages such that the
WDM networks in which the signals remain in the optical connection blocking probability for future connection request
domain are desirable for providing large amount of bandwidth. will be high. Rerouting the connection or lightpath to a better
In all-optical WDM networks, a connection is assigned routing path is a solution for this problem [16]. A number
a wavelength on each of the links along the routing path of researches for rerouting algorithms in WDM networks
for the connection. In case that the connection has to be without traffic grooming have been studied in [16]. In these
researches, a lightpath is able to carry a single connection probability will be smaller. We also hope that the overhead of
while multiple connections are multiplexed in a lightpath in a rerouting algorithms is also minimized.
WDM network with traffic grooming. Therefore, rerouting a Simulation results show that the proposed load-balanced
lightpath in a WDM network with traffic grooming requires rerouting algorithm yields much lower connection blocking
more consideration since all the connections in the lightpath probability than the least resources rerouting algorithms. The
will be rerouted simultaneously. blocking probability produced by the normal connection es-
Rerouting in a WDM network with traffic grooming can be tablishment procedure with the least-resources rerouting al-
classified into two categories with respect to the timestamp of gorithm will be slightly higher than that without rerouting
initiating the rerouting procedure. The first is passive rerouting algorithm. We also observe that the number of rerouted
which initiates the procedure of rerouting only when the lightpaths is very small. This result implies that the overhead
normal connection setup procedure fails. On the other hand, for rerouting a lightpath is small and the performance of the
the active rerouting actively reroutes the lightpaths or connec- networks can be significantly improved.
tions periodically or initiates the rerouting procedure when a The rest of this paper is organized as follows. The normal
connection leaves the network. Since the active rerouting is connection establishment procedure used in this paper is in-
able to adopt the traffic patterns not only when the normal troduced in Section II. The proposed least resources rerouting
connection setup procedure fails, this paper considers the algorithm and load balanced rerouting algorithm are described
active rerouting. in Section III. In Section IV, the simulations are conducted and
With respect to the level of rerouting, rerouting algorithms the results are discussed. Finally, some concluding remarks are
can be classified into connection level rerouting and lightpath given.
level rerouting. In connection-level rerouting algorithm, a
connection can be rerouted to a lightpath in which the available II. C ONNECTION E STABLISHMENT P ROCEDURE
bandwidth is sufficient for the connection. In lightpath-level In this section, the normal connection establishment pro-
rerouting algorithm, a lightpath is rerouted and all connec- cedure used in this paper is introduced. In all-optical WDM
tions carried by the lightpath are also rerouted to the newly networks, several lightpath establishment policies have been
established lightpath. Since the connection-level rerouting is proposed and compared [13]. The least physical hop first
much more complicated, we consider the active lightpath-level (LPHF) method is employed as the connection establishment
rerouting in this paper. The active connection-level rerouting procedure in this paper since the produced connection blocking
algorithms are left for future researches. probability is the smallest one among the various methods
A number of lightpath rerouting algorithms in all-optical in [13].
WDM networks without traffic grooming capability has been The least physical hop first procedure in an all-optical WDM
proposed in [6], [7], [10], [12]. In this paper, we consider the networks with traffic grooming and alternate routing consists
lightpath level rerouting in the networks with traffic grooming of two phases. The first phase is to find the routing path in
which is more complicated than those in the networks without logical topology. Each link in the logical topology is assigned
traffic grooming. The lightpath-level rerouting algorithm has a weight. For example, the weight is defined as the number of
been studied in [14]. In [14], the authors proposed two rerout- fibers the lightpath traversed in least physical hop first method.
ing heuristics, namely, CWA-1L and CLA-1C, for rerouting on If the available bandwidth of the lightpath is smaller than the
lightpath level and connection level respectively. The proposed bandwidth requirement of the connection request, the weight
rerouting algorithms are passive rerouting while the rerouting of the link should be set to infinity and the link cannot be
algorithms proposed in this paper are active rerouting. In used for connection establishment. The routing path in logical
this paper, we proposed two lightpath level active rerouting topology is found by the Dijkstra’s shortest path algorithm [3].
algorithms to reduce the connection blocking probability. If the routing path can be found, the connection is established
In an active lightpath-level rerouting algorithm, three major on the existing lightpaths. Otherwise, the connection tries to
problems should be solved. The first problem is that when find a routing path with common available wavelength in the
the rerouting procedure initiates. The second problem is that physical topology. If the routing path in physical topology
which existing lightpath is selected to be rerouted when the can be found, the procedure to establish a new lightpath starts
rerouting procedure starts. The final problem is that which to setup the connection. Otherwise, the connection request is
routing path and wavelength is determined for the rerouted blocked.
lightpath. To solve the three problems, we proposed two If no routing path on the logical topology is found, the hop-
rerouting algorithms called least resources rerouting algo- count based fixed-alternate routing algorithm [8], [16] is used
rithm and load balanced rerouting algorithm to minimize the to establish a new light path on physical topology. In hop
connection blocking probability. The main idea of the least- count fixed-alternate routing algorithm, the multiple routing
resources rerouting algorithm is that if the resources can be paths between the specified source-destination pair are tried
saved after rerouting, the connection blocking probability can one by one in ascending order according to the hop counts of
be reduced. The main idea of the load-balanced rerouting each routing path to establish a lightpath for the connection
algorithm is that if the traffic pattern after rerouting is more request until a lightpath is successfully established. The first-
balanced than that before rerouting, the connection blocking fit wavelength assignment algorithm [16] which selects the
available wavelength with the smallest index is used to assign routing path is first calculated. Let C = C1 , C2 , · · · , Cn(s,d)
a wavelength to a lightpath. Otherwise, the connection request denote the set of number of common available wavelength
is blocked. along the routing path between the source-destination pair of
the leaving connection where n(s, d) is the number of routing
III. T HE PROPOSED REROUTING ALGORITHMS paths between the source-destination pair (s, d). The range ω
In this paper , we proposed two lightpath-level active rerout- is the calculated as follows:
ing algorithms, namely, the least resources rerouting algorithm
ω = Cmax − Cmin (1)
and the load balanced rerouting algorithm , to reduce the
connection blocking probability. In the proposed algorithms, where Cmax and Cmin are the maximum and minimum
three major problems will be solved separately: 1) When will numbers of common available wavelengths in C. Let k be
the rerouting procedure initiate? 2) Which existing lightpath denoted as the index of routing path that the released lightpath.
will be rerouted? 3) Which routing path and wavelength will For each routing path i 6= k, calculate the range ωi under the
be allocated for the rerouted lightpath? The two proposed situation that if a lightpath on the routing path i is rerouted to
algorithms differ in the methods to solve the second problem routing path k. After calculating all ωi , ∀i 6= k, a lightpath
where the first and the third problems are solved by the same on routing path i with smallest ωi , ∀i 6= k and ωi < ω
method in the two proposed algorithms. is searched for rerouting. Otherwise, the rerouting procedure
stops.
A. Least resources rerouting
After the routing path i is found, the next step is to find a
The main idea of the least resources rerouting algorithm lightpath on routing path i for rerouting. Since the wavelength
is that if the resources consumed by the existing lightpath assignment used in the connection establishment procedure is
can be reduced by rerouting algorithm, it is expected that the first-fit algorithm [16] which selects the available wavelength
connection blocking probability can be reduced. According to with the smallest index, the lightpath is searched from those
this idea, the rerouting procedure initiates when a connection use the wavelength with largest index. The wavelength usage
leaves and this connection is the last one in the lightpath. That will be more compacted if the wavelength selection is from
is, the wavelength occupied by the lightpath will be released that with large index to small one. If a lightpath is found, the
when the connection completes the departure procedure. When lightpath is rerouted to routing path k and the wavelength
a lightpath is released, there is at least one common available used by the leaving lightpath is allocated for the rerouted
wavelength along the routing path. If there is a lightpath lightpath. If no lightpath can be found, the rerouting procedure
between the same source-destination pair and consumed more terminates.
wavelength resources than the released lightpath, the lightpath
can be rerouted such that the consumed wavelength resources IV. S IMULATIONS
can be reduced. The lightpath is searched in the descending Simulations are performed to study the performance of
order according to the number of physical hops the lightpath the proposed rerouting algorithms. We first compare various
traversed. If the lightpath can be found, the lightpath is then traffic grooming policies and choose the best policy for
rerouted to the routing path and wavelength which the released rerouting algorithms. The performances of proposed rerouting
lightpath used. algorithms and the normal connection establishment procedure
without rerouting procedure are compared in order to study
B. Load balanced rerouting
how much performance can be gained by the proposed rerout-
The main idea of the load balanced rerouting algorithm is ing algorithms. We also compare the number of connections
that is the traffic load on the networks can be more balanced, established on the logical topology and physical topology to
the connection blocking probability can be further reduced. study why the proposed load balanced lightpath level rerouting
When a lightpath is newly established, the lightpath may be algorithm performs best among the three algorithms. Finally,
established on a routing path with high traffic load. If the the number of rerouted lightpaths is studied.
lightpath can be rerouted to a routing path with low traffic
load, the traffics on the networks can be more balanced such A. Simulation Model
that the connection blocking probability can be reduced [8], Flat random graphs generated using the GT-ITM [15] tool
[16]. are used to represent the networks. The network size consid-
In the load balanced rerouting algorithm, the traffic load ered in our simulations is 50-nodes networks. The nodes in
is defined by the number of common available wavelength a graph are classified into two types, namely, core nodes and
along the routing path. A routing path with small number of edge nodes. Edge nodes are those nodes that can be the end
common available wavelength is said to be with high traffic points of light paths while core nodes are transit nodes. In
load. The routing path is said to be with low traffic load when the internet, the percentage of edge routers among all routers
the number of common available wavelength is large. is roughly 63.2% [11]. Therefore, we randomly select 63.2%
When a connection leaves and the lightpath used by the of the nodes in a graph as edge nodes. Any two of the edge
connection is also released, the rerouting procedure initiates. nodes can be a source-destination pair. For each destination
The number of common available wavelength along each pair, the link-disjoint paths are pre-calculated.
 0 −1
10 10
No Traffic Grooming
Single Lightpath
Connection blocking probabilities

Connection blocking probabilities

Least Virtual Hop First
−1
Least Physical Hop First
10

−2
10

−2
10 Least resources rerouting
No rerouting
Load balanced rerouting

−3 −3
10 10
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
Traffic intensity per source−destination pair Traffic intensity per source−destination pair
Fig. 1. Comparison of connection blocking probability no rerouting. Fig. 2. Comparison of connection blocking probability with and without
rerouting algorithms

5
In our simulations, the maximum number of link-disjoint x 10
5.5
paths between each source-destination pair is 3. All source-

Number of established connections

destination pairs have the same connection request arrival rate. On logical topology
The rate at which connection requests arrives at each source-
destination pair ranges from 0.6 to 1.4. The mean holding 5
Least resources rerouting
time is exponentially distributed with unit mean. The number
No rerouting
of wavelengths, W , on each link is 24. The wavelengths on Load balanced rerouting
each link is indexed from 1 to 24. The amount of bandwidth
provided in a wavelength is 768 units where the bandwidth 4.5
requirement in each connection request is uniformly generated
On physical topoloty
between 1 and 768. The first-fit algorithm [16] is used to assign
a wavelength to a lightpath.
4
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
B. Simulation Results Traffic intensity per source−destinatio pair
Fig. 3. Comparison of the numbers of established connections
For each random graph and each routing algorithm consid-
ered, 106 connection requests are generated. The connection
requests are routed using one of the routing algorithms under
consideration. The connection blocking probability is then cal-
Next, we compared the performances of the proposed
culated. Each data point of our simulation results is the average
rerouting algorithms with that without rerouting algorithm.
connection blocking probability over 20 random graphs.
Fig. 2 shows the connection blocking probabilities when the
We first compared the connection blocking probabilities
least physical hop first traffic grooming policy is used and
produced by different grooming policies in logical topology.
various rerouting algorithms are employed. From the figure,
Four policies are compared in the simulation. The single
we can make the following observations.
lightpath grooming policy establishes the connection on logi-
• The load balanced rerouting algorithm performs better
cal topology only when a single lightpath between the same
source-destination pair with sufficient amount of available than the least resources rerouting algorithm and that
bandwidth exists. The least logical hop first and the least without rerouting algorithm.
• The connection blocking probability produced when no
physical hop first setup the connection on logical topology
when there is a routing path between the source-destination rerouting algorithm is employed is slightly smaller than
pair. The difference between the two methods is the method to that with least resources rerouting algorithm. That is, the
assign weight on the links in the logical topology. The weights connection blocking probability will increase when the
of the links on logical topology are assigned to 1 in the least least resources rerouting algorithm is employed.
logical hop first method and the number of physical hops the We are interested in the reason why the proposed algorithms
lightpath traversed in the least physical hop first method [13]. produce such a different performances. Fig. 3 shows the
Fig. 1 shows the connection blocking probabilities of the four number of connections established on the existing lightpaths
grooming policies. From the figure, we can observe that the (on the logical topology) and the number of connections
least physical hop first method is able to yield the smallest established on the new lightpath (on the physical topology).
connection blocking probability and the least physical hop first From the figure, the following observations can be made:
method is used to evaluate the proposed rerouting algorithms. • The number of connections which are successfully es-
 4
x 10 nate routing and traffic grooming. The two proposed rerouting
2
algorithms initiate the rerouting procedure and select the
lightpath for rerouting differently. Simulation results show that
Number of rerouted lightpaths

1.5 the proposed load-balanced rerouting algorithm yields much

lower connection blocking probability than the least resources
active rerouting algorithm. Furthermore, the least resources
1 rerouting algorithm performs worse than that no rerouting
algorithm is used. The main reason is that the least resources
rerouting algorithm produces unbalanced traffic patterns. We
0.5 also observe that the number of rerouted lightpath compared
Load balanced rerouting with the total number of connection requests is very small.
Least resources rerouting This result implies that the overhead for rerouting a lightpath is
0
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 small and the performance of the networks can be significantly
Traffic intensity per source−destination pair
improved.
Fig. 4. Comparison of the percentages of departure connections which
rerouted the lightpath ACKNOWLEDGMENTS
This research was supported by the National Science Coun-
cil, Taiwan, under grant NSC99-2218-E-431-001-MY3 .
tablished by traffic grooming on logical topology when
the three methods are employed are very similar. This is R EFERENCES
because the logical topology does not change when the [1] C. A. Brackett, , “Dense Wavelength Division Multiplexing Networks:
rerouting algorithm migrates a lightpath from a routing Principles and Applications,” in IEEE Journal on Selected Areas in
path to another. Communications, vol. 8, no. 6, pp.948-964, August 1990.
[2] I. Chlamtac, A. Ganz, and G. Karmi, “Lightpath Communications: An
• The number of connections which are successfully estab- Approach to High Bandwidth Optical WAN’s,” in IEEE Transactions
lished by a new lightpath on physical topology are dif- on Communications, vol. 40, no. 7, pp. 1171-1182, July 1992.
ferent. The load balanced rerouting algorithm establishes [3] E.W. Dijkstra, “A Note on Two Problems in Connexion with Graphs,”
in Numerische Mathematik 1, pp. 269-271, 1959.
most connections than the other two methods do. [4] A. K. Dutta, N. K. Dutta, and M. Fujiwara, WDM Technologies: Optical
According to the two observations, we found that the Networks, Elsevier Academic Press, Burlington (2004).
[5] R. Dutta, Rudra, A.E. Kamal, and G. N. Rouskas, Traffic Grooming
connection blocking probability is mainly affected by the for Optical Networks: Foundations, Techniques and Frontiers, Springer,
probability that a connection can be established in the phys- 2008.
ical topology. Since the lightpaths between the same source- [6] M. Kouba‘a and M. Gagnaire, “Lightpath Rerouting Strategies in
WDM All-Optical Networks Under Scheduled and Random Traffic,” in
destination pair may always be rerouted to the routing path IEEE/OSA Journal of Optical Communications and Networking, vol. 2,
with less hop counts when the least resources rerouting algo- no. 10, pp.859-871, December 2010.
rithm is used, the wavelength resources which are occupied [7] K. Lee and V.O.K. Li, “A Wavelength Rerouting Algorithm in Wide-
Area All-Optical Networks,” in IEEE/OSA Journal of Lightwave
the lightpath with least hop counts will not be released for Technology, vol. 14, no. 6, June 1996.
a long time. Therefore, the traffic loads under the least re- [8] H.C. Lin, S.W. Wang, C.P. Tsai, and M.L. Hung, “Traffic Intensity
sources rerouting algorithm are less balanced than that without Based Alternate Routing for All-Optical WDM Networks,” in IEEE/OSA
Journal of Lightwave Technology, vol. 26, pp. 3604-3616 , November
rerouting algorithm and the load balanced rerouting algorithm. 2008.
This is why the probability that a new lightpath cannot be [9] D. Mitra and J. B. Seery, “Comparative Evaluations of Randomized
established in physical topology when the least resources and Dynamic Routing Strategies for Circuit-switched Networks,” IEEE
Transactions on Communications, vol. 39, no. 1, pp.102-116, January
rerouting algorithm is employed is higher than the other two 1991.
methods. [10] G. Mohan and C. Siva Ram Murthy, “A Time Optimal Wavelength
Since the rerouting algorithms may be criticized that the Rerouting Algorithm for Dynamic Traffic in WDM Networks,” in
IEEE/OSA Journal of Lightwave Technology, vol. 17, no. 3, March 1999.
complexity for rerouting algorithms is very high, we are [11] N. Spring, R. Mahajan, and D. Wetherall, “Measuring ISP Topologies
interested in the number departure connections which incur with Rocketfuel,” in Proceedings of ACM SIGCOMM, Pittsburgh, PA,
the rerouting procedure and a lightpath will be rerouted August 2002.
[12] Amit Wason and R.S. Kaler, “Rerouting technique with dynamic traffic
successfully for the proposed rerouting algorithms. Fig. 4 in WDM optical networks ,” in Optical Fiber Technology, vol. 16, no.
shows the average number of lightpaths which is rerouted 1, pp. 50-54, January 2010.
to another routing paths when 106 connection requests arrive [13] W. Yao and B. Ramamurthy, “Dynamic Traffic Grooming using Fixed-
Alternate Routing in WDM Mesh Optical Networks,” in the Workshop
the networks. We observe that the numbers are very small on Traffic Grooming, 2004.
compared with the total arriving connection requests. There- [14] W. Yao and B. Ramamurthy, “Rerouting Schemes for Dynamic Traffic
fore, the proposed rerouting algorithms do not incur much Grooming in Optical WDM Networks,” in Computer Networks, vol. 52,
pp. 1891-1904, March 2008.
computational overhead and high time complexity. [15] E. W. Zegura, “GT-ITM: Georgia Tech Internetwork Topology Mod-
V. C ONCLUSIONS els (software),” http://www.cc.gatech.edu/fac/Ellen. Zegura/gt-itm/gt-
itm.tar.gz, 1996.
In this paper, we have proposed two lightpath-level active [16] J. Zheng and H. T. Mouftah, Optical WDM Networks: Concepts and
rerouting algorithms in all-optical WDM networks with alter- Design Priciples, A John Wiley & Sons, 2004.

View publication stats