-
1
A
SEMINAR REPORT
ON
ANALYSIS OF QUEUING DELAY IN OPTICAL SPACE NETWORK ON LEO
SATELLITE
CONSTELLATIONS
Under the guidance of:
Asso.prof(Dr.) NILAMANI BHOI
Submitted by:
Name: SAUMMIT KANOONGO
Roll no.:13040111
Branch : M.Tech , 2 nd
sem (CSE)
DEPARTMENT OF ELECTRONICS AND TELECOMMUNICATION ENGINEERING
(VEER SURENDRA SAI UNIVERSITY OF TECHNOLOGY, ODISHA)
-
2
ACKNOWLEDGEMENT
It is difficult to acknowledge the precious debt of knowledge
and learning. But we
can only repay it through our gratitude. First and foremost I
wish to express my
profound gratitude to the almighty. It is my privilege to
express my sincere thanks
to Prof. Sanjay Agrawal, H.O.D. EL & TC who has always been
a constant
source of inspiration. All the faculty members have helped me
during the
preparation of report by spending their precious time.
I convey my sincere thanks to Dr. Nilamani Bhoi who has given
his most valuable
time and effort in guiding me to complete this seminar in due
time and in this
shape.
Last but not the least I would like to thank my parents, friends
for their co-
operation and continuous support during the course of the
assignment.
saummit kanoongo
Regd no: 13040111
2nd
Semester, M.Tech ( CSE)
-
3
CERTIFICATE
This is to certify that saummit kanoongo Of 2 nd
semester M.Tech in Electronics
and Telecommunication Engineering dept, Communication System
Engineering
specialization bearing Regd No.13040111 has given his seminar
talk and prepared
seminar report on Analysis of queuing delay in optical space
network on LEO satellite constellations Under our guidance and
advice.
Dr. Sanjay Agrawal DrNilamani Bhoi
-
4
A B S T R A C T ON
Analysis of queuing delay in optical space network on LEO
satellite constellations
Low earth orbit (LEO) satellite constellations using laser
inter-
satellite links (ISLs) are recognized as a promising
technology
to provide global broadband network services. In this paper,
the
queuing delay model of an optical space network built on LEO
satellite constellations is established. It is assumed that
the
optical space network employs wavelength division
multiplexing ISLs with wavelength routing technology to
communication satellites and makes routing decisions. With
consideration of the network task characterizations such as
distribution of task arrival time and task holding duration,
simulation experiment results are analyzed and the expression
of
optical space network queuing delay is given. Both
theoretical
analysis and simulation results show that features of
queuing
delay vary with distribution characterizations of the
network
tasks. It is hoped that the study can be helpful to evaluate
the
design of constellation networking.
-
5
CONTENTS
1. Introduction..6
2. What Is New In This Paper.7
3. Model And Analysis Of The Network
.......................................12
4. Theoretical Analysis Of Network Traffic Parameters..12
5. Numerical Results And Analysis ....13
6. Future Work ..17
7. Features Technology....18
8. Conclusion......19
9. References.20
-
6
Introduction :-
Optical space network based on wavelength division multiplexing
(WDM)
architecture and wavelength routing technology has emerged as
the befitting
backbone for the dramatic increase in bandwidth demand of
emerging applications
[1]. Several optical space network have been proposed such as
Celestri
[2],Teledesic [3], and NeLS [4]. These satellite systems built
on non-
geosynchronous orbits, use a single wavelength or WDM laser
inter-satellite links
(ISLs) for broadband inter-satellite communications.
In contrast to single wavelength ISLs, WDM ISLs with onboard
wavelength
routing equipment perform better by offering more routing
selection and by
reducing processor delays in high band-width communications
[5].The key factor
of optical space network transmission feasibility is the
transmission delay, also
called network delay, which is the duration of the signal
passage from the source
satellite transmission to the destination satellite. In optical
space networks based on
WDM technology, network delay is mostly decided by the queuing
delay. Queuing
delay indicates the duration from one signal transmission
network task arrived at
the source satellite to the time when this signal was ready to
be transmitted.
During queuing delay, the satellite system is expected to assign
the light path for
communication and the wavelength for the corresponding network
task, and
check the light path and wavelength till they are available. Due
to the periodic
motion of the satellites, each network task has a waiting
limitation until it can be
executed. Thus, the increase of the queuing delay will be a
destructive defect to
the optical satellite network. For this reason, it is necessary
to research the
generative mechanism and characterization of the queuing delay
in the optical
space network in order to reduce it to an acceptable range.
Previous study has
mainly focused on the network structure in LEO satellite
communication system,
in order to optimize the orbit parameters to enhance the
performance of
communication service [6].
In these feasibility studies on the LEO satellite networks,
Walker constellations
with optical ISL are employed to form a global satellite network
and the
-
7
constellation parameters such as orbits altitudes, number of
orbital planes,
number of satellites in the orbital plane and the orbital
inclination were studied
for optimization [7]. However, to the best of our knowledge, no
previous work
concerned the performance of queuing delay in random network
tasks on the LEO
satellite networks. In this paper, on the assumption that WDM
architectures with
wavelength routing were available for the ISLs, and in view of
random network
tasks, queuing delay in the optical space network is analyzed
for the Walker
constellation with optical ISLs.
The paper is organized as follows. Section 2 describes the
system model and
theoretical analysis of the queuing delay in the optical space
network. Section 3 is
devoted to the simulation and numerical analysis with
discussion. Section 4 states
the conclusions of this work.
Model and analysis of the network :-
Logical connections of LEO space network :-
In optical space network employing low earth orbit (LEO)
satellite constellations,
the geographic topology of the network changes periodically.
This dynamic
network architecture is considered as a great challenge to the
light path routing and
wavelength assignment of the WDM based optical network.
Therefore, in order to
improve the effectiveness of routing and wavelength assignment,
the topology of
the network has to be simplified.
Through the use of continuous ISL connections, a typical LEO
optical space
network in the Walker constellation can be seen as a modified
Manhattan network
[8], as is shown in Fig. 1, an example of a typical LEO
satellite network with five
orbital planes and six satellites in each orbital plane is
given. Each satellite in the
network is connected to four adjacent satellites by optical
ISLs. However, in
addition to the model shown in Fig. 1, there are also several
different logical
connection architectures studied by researchers, corresponding
to different
schematics of the network.
-
8
When a network task, also called connection request, arrives at
the
LEO satellite network, a sequence of ISLs which starts from the
source satellite
and ends in destination satellite is generated and assigned to
this connection
request. Each generated sequence of the ISLs is a specific route
to the
corresponding connection, and each specific route needs a
customized
wavelength to meet the communication requirement.
A practical problem is that the amount of available wavelengths
is finite in the
network, which means the number of connections simultaneously
operating on
the network is limited. Some network tasks arriving at the
network have to wait
for an accessible route along with an unoccupied wavelength in
this route to
accomplish communication mission.
On this occasion, the time spent on the waiting is called
queuing delay.
Obviously, queuing delay is associated with the network
capacity, routing
principle, temporal distribution of network tasks and some other
network
parameters.
Theoretical analysis of network traffic parameters :-
Since queuing delay in the optical space network is relevant to
network
capacity, clearly are expression of network capacity should be
given.
Considering a LEO satellite network with L orbital planes and M
satellites in
each orbital plane, total number of ISLs in the network can be
written as
-
9
In number of ISLs, the average connection length count between
network node
pairs can be expressed as
where pi is the connection length count in number of ISLs
between the reference
satellite and ith satellite in the network. What should to be
point out is that the
average connection length count Pav shown in Eq. (2) is computed
with assigning
the shortest path to each node pair.
The network capacity can be measured by the number of
connections operating
on the network simultaneously. With average connection length
shown in Eq. (2)
-
10
and total number of ISLs shown in Eq. (1), the theoretical
network capacity NTL
can be depicted as
where n is the number of available wavelengths in the optical
satellite network.
Practically, not all of the ISLs and their wave-lengths can be
utilizing at the same
time, and then the practical network capacity can be written
as
where P(t) stands for the number of occupied ISLs at time t in
the network.
In general, in the care of identical temporal distribution of
the network tasks,
increasing network capacity can decrease the number of queuing
connections. On
the other hand, under the condition of certain network capacity,
increasing
network task generation rate will lead to a boost of the queuing
delay.
Distribution of the arrival time determines the amount of
connection requests per
unit duration, while service time distribution affects the
operating speed of
connection requests.
To build a reasonable temporal distribution model of network
tasks, arrival time
and service time of connection requests generated on certain
distributions should
be considered. Actually, plentiful approaches have been proposed
in order to
model large network flows as well as their superposition
properties [9].In this
-
11
paper, arrival times with Poisson distribution [10] and service
times with
lognormal distribution [11] were chosen both for their
simplicity and because
they provide rather general and realistic representation of
large network systems.
Under these assumptions, the mean arrival time can be define das
_ and mean
service time is assumed as _. Since the expression of network
capacity is depicted
and network model is built, queuing delay of the network in unit
duration can be
written as
In order to estimate the average queuing delay for every
connection requirement,
the total number of connection requirements should be considered
and the
average queuing delay in unit duration can be expressed as
From Eqs. (6) and (7) it can be seen that the queuing delay of
the entire network
must be a function related to the Dun or Dunav. To understand
and quantify the
performance of queuing delay in an optical space network, a
series of
experiments carried out.
-
12
Numerical results and analysis
Simulation setup
To investigate the relation between temporal distribution of
connection requires
and queuing delay on LEO optical satellite net-work, the
parameters for the
simulation process are set as follows: the satellites orbital
altitude is 1200 km, the
period of satellite is6565 s, the inclination of orbital plane
is 55, the number of
orbits is 5 and every orbit have 6 satellites, the same logical
connections as shown
in Fig. 1. Unless otherwise mentioned, these parameters are used
for all the results.
At the beginning of optical space network process simulation,
connection
requirements were generated randomly with Poisson distribution
arrival time and
lognormal distribution service time. Following the arrival time
schedule, the
connection requirements arrived at the network system in
chronological order
and became activated. The satellite system will work out a
specified light path
with a dedicated wavelength for every activated connection
requirement.
The shortest path routing method and first fit wavelength
assignment [12] are
employed through the network. All wave-lengths are numbered from
1 to n.
Subsequently the network with n wavelengths in each link is
decomposed into n
layered networks. Each of which has the same topology but one
wave-length
capacity in each link. For a connection requirement, when the
light path was
worked out by the satellite system, check every numbered network
layer if the
corresponding light path is free. The wavelength with the lowest
number is
selected from the available wavelengths.
However, a connection requirement should wait if every
wave-length of the
corresponding light path is occupied, and the waiting time will
be added to the
network queuing delay. With network time going on, one
connection requirement
will be performed immediately when its assigned light path and
wave-length are
free. All the waiting time will be summed up at the end of the
network time, and
ISL utilization ratio at every satellite time is also recorded
.As is mentioned above,
-
13
the light path for new connection requirements follows the
shortest path routing
method, and the wavelength assignment for a new connection
requirement is
deter-mined both by the required light path and first fit
wavelength assignment
mechanism.
Results and analysis :-
The simulation result of ISLs utilization ratio for different
arrival time and service
time distributions is shown in Fig. 2.
Fig. 2 plots the ISLs utilization ratio in various temporal
distribution statuses of
connection requirements. A larger ratio, which is the key
feature of the connection
requirements distribution, implies a heavier traffic on the
network as the system is
going to deal with more connection requirements simultaneously.
It is
-
14
in Fig. 2 that ISL utilization ratio becomes larger with
increasing of the ratio,
because more ISLs are occupied by increasing network traffic.
Contrasting the Fig.
2(b) with (a), it can be seen that optical space network with
more available
wavelengths is more capable in dealing with larger connection
requirements at
the same time and the ISL utilization ratio is steadier than
with fewer
wavelengths. Fig. 3 shows the average ISLs utilization ratio
varying along with the
-
15
ratio. When ratio less than theoretical network capacity NTL,
average ISLs
utilization ratio is increasing linearly along with the ratio.
However, the increasing
speed is lower when ratio above NTL. It has been demonstrated
that raising ratio
is leading to increasing network traffic and a indicates the
used ISL resources.
When the increasing is no match for corresponding ratio
increasing, a
conspicuous rising of queuing delay will be generated in the
network. The result
represented in Fig. 3 is helpful in designing an optical space
network. On one
hand, if the number of satellites is previously settled by other
reasons, in order to
improve the utilization ratio of network ISLs, a proper ratio
should be considered.
On the other hand, if the network traffic demand is the core
factor of the
designed network, the designer should assign reasonable
satellites and ISLs to
meet the network requirements. The plot in Fig. 4 depicts the
average waiting
time performance at different network traffic level, and the
average waiting time
is the mean value of queuing delay for each network connection
requirement.
Average waiting time, also called average queuing delay, is less
than 200 s when
the ratio is below the value of theoretical
-
16
Network capacity NTL. And the results in Figs. 3 and 4 indicate
that the
performance of optical satellite network with different
avail-able wavelengths is
similar. To investigate the similitude character of optical
satellite network with
different available wavelengths, a normalized network traffic
level should be
utilized.
As shown in Fig. 5, when the traffic level in different network
(mainly
distinguished by number of available wavelengths n_) are
normalized by the
transformation of , their performances of average queuing delay
are nearly the
same.
This result implies that the analysis of network queuing delay
can be unified by
the normalized network traffic level rather than dealing with
many different
situations of n. On the other hand, that also suggests the
normalized network
traffic level in optical space network depend on the temporal
distribution of the
network tasks and theoretical network capacity.
Eq. (8) describes the relationship between the average waiting
time and
normalized network traffic level under several given conditions.
These conditions
include network logical topology, light path routing method, and
the wavelength
assignment mechanism. And parameters describes in Eq. (8) vary
along with the
above condition changes.
-
17
A contrast of theoretical and experimental average waiting time
is represented in
Fig. 6. Results show that the value obtained numerically from
Eq. (8) and through
simulations match closely. Curves expressed in Figs. 5 and 6
prove that average
queuing delay in optical space network could be written as a
function of
normalized network traffic level . If queuing delay in a
designing network can be
modeled, the predesign of this network will benefit.
FUTURE WORK :-
In this paper, arrival times with Poisson distribution and
service times with
lognormal distribution were chosen both for their
simplicity.
So better model can be used because they can provide rather
general realistic
representation of large network systems.
-
18
Conclusion:-
In this paper, queuing delay in the optical space network on LEO
satellite
constellations was analyzed with the consideration of the
network logical
topology, routing and wavelength assignment method and temporal
distribution
of the connection requires. To research the performance of
queuing delay, a
model of the optical space network on LEO satellite
constellations with shortest
path routing method and first fit wavelength assignment
mechanism was built.
Distributions of connection requirement such as arrival time and
service time
were chosen reasonable. Arrival time with Poisson distribution
and service time
with lognormal distribution were chosen to simulate connection
requirements in
large network systems. Under these preconditions, the
experimental results
show that the performance of optical satellite networks with
different available
wavelengths is similar, and the normalized network traffic level
in optical space
networks depends on the temporal distribution of network tasks
and theoretical
network capacity. Through all the theories and simulation
results, the
performance of ISLs utilization ratio and the average queuing
delay in a given
optical space network has been proved describable as function of
normalized
network traffic level. The results of the study will help to
improve optical space
net-work design and to advance the performance of networking in
optical space
networks.
-
19
REFERENCES:-
[1] J. Walfisch and H.L.Bertoni, " Analysis of queuing delay in
optical
space network on LEO satellite constellations ", IEEE Trans.
Antenn. &
proD., vol. AP-36, no. 12, pp. 1788-1796, 2013
[2] R.J. Leopold, K.M. Peterson, CELESTRITM Ka-band sharing,
Proc. IEEE
4 (1998)553560.
[3] D.P. Patterson, Teledesic: a global broadband network, Proc.
IEEE 4
(1998)547552.
[4] Y. Koyama, E. Morikawa, H. Kunimori, K. Shiratama, Y.
Yasuda, Com-
ponents development for NeLS optical terminal, Proc. SPIE 5712
(2005)217
224.[
5] Q.L. Yang, L.Y. Tan, J. Ma, Doppler characterization of laser
inter-
satellitelinks for optical LEO satellite constellations, Opt.
Commun. 282
(2009)35473552.
[6] R. Suzuki, S. Motoyoshi, A study of constellation for LEO
satellite
network,in: 22nd AIAA International Communications Satellite
Systems
Conference &Exhibit, 2004, pp. 32363245.[
7] Y. Karafolas, S. Baroni, Optical satellite networks, J.
Lightwave Technol.
18(2000) 17921806.
[8] R. Suzuki, Y. Yasuda, Study on ISL network structure in LEO
satellite
communi-cation systems, Acta. Astronaut. 61 (2007) 648658.[
9] C. Cavdar, M. Tornatore, F. Buzluca, B. Mukherjee,
Shared-Path protection
withdelay tolerance (SDT) in optical WDM mesh networks, J.
Lightwave
Technol. 28(14) (2010) 20682076.
[10] G.X. Zhang, S. Ferrari, C.H. Cai, A comparison of
information functions
andsearch strategies for sensor planning in target
classification, IEEE Trans.
Syst.42 (1) (2012) 216.