Performance Analysis of Continuous Wavelength Optical Burst Switching Networks

Abstract

Considering the economic and technical aspects of wavelength converters, fullwavelength conversion capability will not be available throughout opticalnetworks in the foreseeable future. In this paper we have used balanced staticwavelength assignment (BSWA) algorithm. It is a technique to improve theperformance of an all-optical network in respect of blocking probability & otherparameters to provide cost effective optical communication system. We carriedout the performance analysis of wavelength-continuous optical burst switchingnetwork-using BSWA for the network topologies namely NSFNET andINTERNET2NET with the initial assumption that there is no wavelength converteror optical buffer in the networks. For simulation we have used a powerful toolMatplanWDM. The performance analysis of burst loss probability v/s traffic loadusing BSWA algorithm shows that its performance is better than other staticwavelength assignment algorithm. We have also explored the performances ofNSFNET and INTERNET2NET networks on the basis of channel utilization,message delay, and the ratio of traffic transferred in single hop to total trafficoffered against percentage of reused wavelength in the network. It has beenobserved that channel utilization becomes the maximum for 40% of reusedwavelengths per fiber and beyond that cost of network unnecessary increases.By the performance analysis of message propagation delay for both networkTopology, it is observed that message propagation delay, is minimum at 40% ofreused wavelength per fiber and it becomes constant beyond that. It is thereforerecommended that reused wavelengths per fiber should not exceed 40 % of thetotal available wavelength so as to yield cost effective solution with optimumperformance.

Full text

Page 1

ADITYA GOEL, VITTHAL J GOND, R K SETHI and A K SINGH

International Journal of Engineering, Volume (3): Issue (6)
609
Performance Analysis of Continuous Wavelength Optical Burst
Switching Networks


Aditya Goel
Department of Electronics & Communication
MANIT (Deemed University)
Bhopal-462051, India

Vitthal J. Gond vitthal_g@rediffmail.com
Department of Electronics & Communication
MANIT (Deemed University)
Bhopal-462051, India

R K Sethi
Department of Electronics & Communication
MANIT (Deemed University)
Bhopal-462051, India

And
A K SINGH
Department of Electronics & Communication
MANIT (Deemed University)
Bhopal-462051, India
adityagoel2@rediffmail.com
sethirk@hotmail.com
akhil26783@gmail.com

Abstract
Considering the economic and technical aspects of wavelength converters, full
wavelength conversion capability will not be available throughout optical
networks in the foreseeable future. In this paper we have used balanced static
wavelength assignment (BSWA) algorithm. It is a technique to improve the
performance of an all-optical network in respect of blocking probability & other
parameters to provide cost effective optical communication system. We carried
out the performance analysis of wavelength-continuous optical burst switching
network-using BSWA for the network topologies namely NSFNET and
INTERNET2NET with the initial assumption that there is no wavelength converter
or optical buffer in the networks. For simulation we have used a powerful tool
MatplanWDM. The performance analysis of burst loss probability v/s traffic load
using BSWA algorithm shows that its performance is better than other static
wavelength assignment algorithm. We have also explored the performances of
NSFNET and INTERNET2NET networks on the basis of channel utilization,
message delay, and the ratio of traffic transferred in single hop to total traffic
offered against percentage of reused wavelength in the network. It has been
observed that channel utilization becomes the maximum for 40% of reused
wavelengths per fiber and beyond that cost of network unnecessary increases.
By the performance analysis of message propagation delay for both network
Topology, it is observed that message propagation delay, is minimum at 40% of
reused wavelength per fiber and it becomes constant beyond that. It is therefore
recommended that reused wavelengths per fiber should not exceed 40 % of the
total available wavelength so as to yield cost effective solution with optimum
performance.

Page 2

ADITYA GOEL, VITTHAL J GOND, R K SETHI and A K SINGH


Keywords: BSWA: Balanced Static Wavelength Assignment, OBS: Optical Burst Switching, OADM:
Optical Add/Drop Multiplexers, OXC: Optical Cross-Connect, TBS: Terabit Burst Switching, PWA: Priority
Wavelength Assignment.
International Journal of Engineering, Volume (3): Issue (6)
610

1. INTRODUCTION
Due to the introduction of various wide band applications, such as video conferencing, broadband
wireless communication [1] etc the demand of transmission capacity is growing exponentially.
This has motivated the researcher to device & employ different techniques [2,3] to enhance the
existing transmission capacity of optical terrestrial and wireless networks. In order to
accommodate this wide capacity requirement, efficient and cost-effective wavelength division
multiplexing (WDM) can be used between network nodes [4]. Today up to several Tbit/s traffic
can be carried by the optical link over long distances. e.g. in an ultra-dense WDM transmission
total capacity of 50 Tbit/s have already been experimentally demonstrated [5]. However, with the
introduction of very high capacity WDM optical transmission techniques, the discrepancy between
optical transmission capacity and electronic switching capability increases. Moreover, due to cost
and complexity aspects, it is advantageous to keep data in the optical domain and to avoid bit-
level signal processing. Burst switched WDM optical networks are coming up as suitable network
architectures for future Optical Internet backbones [6]. Optical Burst Switching (OBS) has
emerged as a new paradigm for an optical Internet [7]. During the past years, there are large
efforts to realize OBS. Although the concept of burst switching had already been known since the
1980s, it has never been a big success in electrical networks. The main reason is that its
complexity and realization requirements are comparable to that of more flexible electronic packet
switching techniques. Optical circuit switching (OCS) does not support fine switching granularity
on the packet-level. Only poor bandwidth utilization is obtained due to the absence of statistical
multiplexing. On the other hand, Optical Packet Switching (OPS) are still too complex to be
realized in the near future. Therefore, a hybrid approach like burst switching seems promising for
the WDM-based optical networks. Optical burst switching (OBS) was proposed in the late 1990
[7,8] and it tries to exploit the advantages of both OCS and OPS approaches and avoid a part of
their limitations. OBS is considered as one of the most promising approaches to implement IP
over WDM [9]. At the moment, due to the fast development of this new approach and the large
number of new proposals, OBS is still at its definition phase and no common definition of OBS
can be found. However, among several different definitions two main definitions can be found.
One is more oriented towards fast wavelength switching. Synonyms like optical flow switching
(OFS) [10] and wavelength routed OBS (WR-OBS) [11,12,13] can be found in the literature. Here,
an acknowledged two-way reservation is generally used. A burst will be only transmitted after
having received a positive acknowledgement of its reservation request. Another most widely used
definition for OBS is more packets oriented and was proposed by Qiao and Turner [14,7,8].
To realize an OBS network, a new optical layer is mandatory. Furthermore, an OBS network will
be only possible if fast and efficient switching nodes can be built. To constitute the OBS network
nodes several key components are needed e.g. optical space switches, tunable filters/lasers,
wavelength converters etc. Most of the studies on OBS networks assume that full wavelength
conversion is available throughout the network. However, such capability is not at present a
realistic assumption. Although use of wavelength converters give better performance with respect
to blocking probability but it is quite expensive. So, the absence of (full) wavelength conversion
capabilities calls for good and efficient wavelength assignment policies. The Wavelength
Assignment Algorithm is a method to improve the performance in respect of blocking probability
and to provide cost effective communication system. To solve the above problems there have
been developed number of wavelength assignment techniques e.g. Random, FF-TE, BSWA etc.
Traditional first-fit or random wavelength assignment algorithm in OCS network is not appropriate
for wavelength-continuous OBS networks, since the burst is transmitted without knowledge of the
wavelength occupation status of the following links.
Out of the number of wavelength assignment techniques we have selected Balanced Static
wavelength assignment (BSWA) algorithm. The BSWA algorithm is new and improved
wavelength assignment policy [15] In BSWA wavelength order lists are carefully designed during

Page 3

ADITYA GOEL, VITTHAL J GOND, R K SETHI and A K SINGH

the network-planning phase, and no extra information exchange or signaling is needed. The
BSWA improves the performance of blocking probability vs. traffic load than other existing static
wavelength assignment policies. The BSWA also reduces the cost of communication system
because it reduces the number of wavelength converters, which is too costly. The BSWA has
improved the performance in respect of blocking probability as compared to other static
approaches. Numerical results show that BSWA reduces the burst loss probability in comparison
with existing static approaches, and almost achieves the same performance as PWA (Priority
Wavelength Assignment) with the advantage that no extra dynamic information is needed. In this
work performance analysis for NSFNET and INTERNET2NET networks have been carried out for
various crucial parameters like burst loss probability, message propagation delay and channel
utilization.

1.1 Optical Burst Switching (OBS)
The main characteristics of OBS can be summarized as follows:
• Client layer data (IP-Packet) is aggregated and assembled into variable length optical
bursts in edge nodes. OBS assumes more extensive burst aggregation, to realize bursts
with payloads typically carrying tens of Kbytes.
• There is a separation between control information (header) and user information (data) in
space and in time. Control header packets are sent on a separate wavelength (out-of-
band) and processed electronically in all OBS core nodes to set up the switch matrix
before the data bursts arrive.
• Data bursts are asynchronously switched in core nodes and stay in the optical domain
until they reach their destination edge node. Only wavelength conversion and/or some
degree of regeneration are applied to the signal.
• Resources are allocated by using one-pass reservation, i.e., burst transmission is not
delayed until an acknowledgment of successful end- to-end path setup is received but is
initiated after the burst was assembled and the control packet was sent out.
• Burst switching does not require buffering inside the core network.
International Journal of Engineering, Volume (3): Issue (6)
611
2. BSWA ALGORITHM & IMPLEMENTATION
To reduce the blocking probability, wavelength assignment techniques play an important role in
OBS Network, which are not having full wavelength conversion capability. To address the burning
problem of network congestion along with low cost, various wavelength assignment algorithms
have been developed like First-Fit, Random, First-Fit-TE, etc. In this paper we have carried out
the performance analysis of newly developed wavelength assignment algorithm, Balanced Static
Wavelength Assignment algorithm. BSWA is capable of reducing the burning problem of network
congestion to a substantial amount in comparison to other traditional wavelength assignment
algorithm.

2.1 BSWA Algorithm
BSWA algorithm is static approach, which does not require the exchange of any information
among network routers. Thus, the assignment rules are determined during the network-planning
phase, depending on certain properties of the network like topology, routing paths, and traffic. In
this work we used traffic information for wavelength assignment. In this BSWA tries to fully
balance the traffic that shares common links between different source-destination pairs among
wavelength on those links.

An OBS network is represented by a graph G (V, E), where V = {V1, V2,...,VN} is the set of nodes,
and E={E1, E2,…,EM} is the set of unidirectional fiber links. The link E(Vs, Vd) connects Vs to Vd
and contains W different wavelengths labeled arbitrarily, say, , λ1, λ2 ,……, λw . By P{Vs ,Vd}, we
denote the routing path from Vs to Vd, and ρ(Vs ,Vd) is the given traffic load from Vs to Vd. We define
virtual_wl_cost(Vi, Vj, λk) to be the virtual cost of wavelength λk on link E(Vi, Vj), and path_cost(Vs, Vd,
λk) to be the cost of the the path on λk along which the source Vs sends bursts to the destination Vd. Thus ,
path_cost(Vs, Vd, λk) is calculated as

Page 4

ADITYA GOEL, VITTHAL J GOND, R K SETHI and A K SINGH

International Journal of Engineering, Volume (3): Issue (6)
612
path_cost(Vs, Vd, λk) =
()
()
()
ijsd
ijk
E V , VP V ,V
virtual_wl_cost V, V,
λ
∈∑
(1)
Balanced static wavelength assignment (BSWA) algorithm:
Input: G(V,E), P(Vs ,Vd), ρ(Vs ,Vd) for all s,d є {1,2,……..,n}
Output: WA(Vs, Vd, t) for all s,d є {1,2,……..,n}, t є {1,2,……..,w}
Initialization: virtual_wl_cost (Vi, Vj, λk) = 0 for all i, j є {1,2, ... ,N},
t є {1,2, ... ,W}
FOR t← from 1 to W
FOR s← from 1 to N
FOR d← from 1 to N
Step 1: Compute the path_cost for each wavelength that has not been assigned before for this
s-d pair.
Step 2: Choose the λk that has the least path_cost, and make it the wavelength assigned in this
loop.That is, WA(Vs, Vd, t)= λk.
Step 3: Update virtual_wl_cost(Vi, Vj, λk) for all E(Vi, Vj) є P{Vs, Vd} by virtual_wl_cost(Vi, Vj, λk) =
virtual_wl_cost(Vi, Vj, λk) + ρ(Vs, Vd) (2)
The process of BSWA is stated as follows. First the cost of each wavelength on all links,
virtual_wl_cost(Vi, Vj, λk) is initialized to 0. Then we begin to assign wavelengths for s-d pairs one by one.
The value of path_cost(Vs, Vd, λk), k є {1, 2, ….,w}, is calculated for each (Vs, Vd) pair, and λk that has the
least path_cost is chosen as the wavelength assigned for the (Vs, Vd) pair.Every time we assign a
wavelength (e.g. λk) for an s-d pair, we increase virtual_wl_cost(Vi, Vj, λk) on all the passing links of P{Vs
,Vd} by ρ(Vs ,Vd) so that the wavelength is less likely to be chosen by other s-d pairs if it is already heavily
loaded. We continue to do this for all s-d pairs, so that each pair is assigned a wavelength. The
process described is called searching loop W times, we can get a list of the wavelength
assignment order for each s-d pair.

2.2 Implementation of BSWA Algorithm
To reduce the data congestion with low cost, we have used balanced static wavelength
assignment algorithm to assign the wavelength for each source-destination pair (s-d pair). The
programming of the BSWA algorithm has been coded in Matlab. We have used a very powerful
tool named as Matplan-wdm to implement the coded algorithm. This works on the platform of
Matlab7.0 or above version. The optimization in this tool will be done with the help of Tomlab,
which is also a powerful tool for optimization. The algorithm has been implemented in two
network topology namely NSFNET & INTERNET2NET as shown in Figure 1 & Figure 2
respectively. The NSFNET consist of 14 nodes and 21 bi-directional fiber links, where as
INTERNET2NET consist of 9 nodes and 13 bi-directional fiber links. The traffic loads have been
taken in erlangs per wavelength. The NSFNET have 40 maximum wavelengths per fiber where
as INTERNET2NET have 32 maximum wavelengths per fiber. It is assumed that there are not
available wavelengths converters and optical buffers in both networks. We have used the
Dijkastra’s shortest-path algorithm to calculate the route for each s-d pair [15]. In this algorithm
each s-d pair will select shortest path in network to transfer the data. This algorithm has been
also coded in Matlab. We have analyzed the burst loss probability by varying the traffic load in
erlangs per wavelength. Similarly we have done performance analysis of channel utilization,
traffic transferred in single hop, message propagation delay by varying the number of
wavelengths per fiber. The channel utilization has been calculated as how many wavelengths
channels are being used to transfer the data out of total wavelengths channels available in the
network. The traffic transferred in single hop has been calculated as ratio of traffic transferred in
single hop to total traffic offered in percentage. The message propagation delay has been
calculated as the difference between the time that message being transmitted at transmitter and
that is being received at receiver. It is assumed that burst dropped at transmitter have negligible
propagation delay. The number of wavelengths per fiber varied as percentage of maximum
wavelengths used per fiber in the particular network. For example in NSFNET maximum
wavelengths per fiber is 40 and in INTERNET2NET that is 32.

Page 5

ADITYA GOEL, VITTHAL J GOND, R K SETHI and A K SINGH

International Journal of Engineering, Volume (3): Issue (6)
613
1
2
3
4
5
6
7
89
10
1112
13
14

FIGURE 1: NSFNET (14 Nodes, 21 Bi-directional Link)











FIGURE 2:INTERNET2NET (9 Nodes, 13 Bi-directional Link)

2.3 Validation of BSWA
We have validated our simulated results generated through our code with the previously
published results [15] as reported by Chunlei Zhu, Wenda Ni at all in their paper titled “New and
Improved Approaches for Wavelength Assignment in Wavelength-Continuous Optical Burst
Switching Network,” By The tabular comparison of the two shows that the simulated result is in
reasonably good agreement with the reported result [15]. The comparison of result is for NSFNET
with 14 nodes and 21 bidirectional fiber links. The load is taken in erlangs per wavelength.

Table: 2.1 Result Validations
LOAD
(Erlang)
Result (1)
0.1
0.0015
Published Simulated
0.002
0.2
0.009
0.026
0.009
0.03
0.3
0.4
0.05 0.06
0.5
0.08
0.12
0.08
0.12
0.6
0.7
0.15
0.2
0.15
0.19
0.8
0.9
0.25
0.30
0.26
0.30
1.0


1
3
67
8
4
2
5
9