Document Sample
IJETTCS-2013-04-18-124 Powered By Docstoc
					   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

                   CONGESTION CONTROL
                                      Yogini Bazaz1, Sudesh kumar2 and Sanjay Anand3
                                   Computer science & Engineering, Central University of Rajasthan,India
                                          Computer Science, IGNTU, Amarkantak, (M.P.) India

Abstract: Multimedia transmission over the communication             controlling congestion are really important and work
network and computer has gained importance from many                 really well but we need an easy, quick and effective
years. Real time multimedia applications have come into              mechanism for controlling congestion and that
existence like audio, video etc. So the congestion problem is        mechanism is fuzzy logic [8]. The concept of Fuzzy Logic
also increasing .Congestion is really a complex problem to           (FL) was given by LotfiZadeh, a professor at the
define. It arises when the users tries to access the same            University of California at Berkley. It is a way of
resource. Congestion is a complex problem. There are many
                                                                     processing data by allowing partial set membership rather
types of techniques to control congestion but we want time
sensitive, easier, fast, effective technique to control
                                                                     than crisp set membership or non-membership and it is
congestion.     To control the problem of congestion many            presented as a control methodology. This set theory was
algorithms were designed but fuzzy logic is very easy, simple        never applied to control systems until the 70's due to
and effective way to solve the problem of congestion. For            having small-computer capability of that time. Professor
multimedia application RTCP is used with RTP and uses TCP            Zadeh reasoned that people are capable of highly adaptive
for bidirectional client server connection. The rapid internet       control and do not require precise, numerical information
growth and the demand of usage of internet for time sensitive        input. If controllers could be programmed to accept
streaming applications is increasing the utilization and             imprecise, noisy data they would be much more effective
design of effective congestion control techniques. This paper        and perhaps easier to implement.
takes the fuzzy logic to control congestion problem. It’s
totally easy and faster and is based on human thinking. Fuzzy
is based on the logic that takes the values between 0 and 1.It       2. METHODOLOGY
does not take the crisp values. Fuzzy logic toolbox is used in       Before making a system we must first built and consider
the matlab to make the effective fuzzy system. Congestion            rules and define all these terms we plan for using and the
control has become an application for the fuzzy logic. In this
                                                                     adjectives that describe them. Fuzzy inference system is a
paper we are proposing a mechanism to control congestion in
                                                                     method that interprets the values in input vector and
streaming media applications by using fuzzy logic. Then a
model has been generated by the fuzzy logic controller to            based on some set of rules assigns values to the output
control the congestion.                                              vector [24]. Rules used for controlling congestion in our
Keywords: Streaming media, Congestion, Congestion                    work for fuzzy are as follows:
Control, Fuzzy Logic, Matlab
                                                                           If (available bandwidth is low) and (change rate is
1. INTRODUCTION                                                               decreasing) then (send rate is low).
                                                                           If (available bandwidth is low) and (change rate is
The internet stability is on risk if the video and voice                      decreasing) then (send rate is high).
traffic continues to increase.UDP is the most suited
                                                                           If (TCP response is low) then (send rate is low).
protocol for streaming media[9][22] but TCP also work
                                                                           If (available bandwidth is high) then (TCP
good for maximum applications but still they have some
                                                                              response is high) then (send rate is high).
problem[15]. More over TCP is a reliable protocol and
                                                                           If (available bandwidth is high) then (send rate is
faces the problem of congestion.UDP works well with
streaming applications but is not reliable and also will not
                                                                            If (available bandwidth is low) then (send rate is
give the problem of congestion. Congestion is really a
complex problem. It leads to the degradation of
performance of network. Congestion arises when the user              There are three inputs in our system. The inputs used in
tries to use the same resource. Congestion control is the            our system are TCPresponse, change rate and available
term used for controlling congestion[5][6]. So it is                 bandwidth. So based on that we are going to describe
accepted that the network congestion control remains a               their membership functions. Membership function is a
critical issue and a high priority because of the growing            curve that defines how each point in input space is
demand, size and speed i.e. Bandwidth of the network                 mapped to a membership value (or degree of
.So for streaming media applications we are going to                 membership) between 0 and 1.There are eleven types of
control congestion using Fuzzy logic. Although the                   membership functions but these membership functions
algorithms [21] which are previously designed for                    come under four major sub categories
Volume 2, Issue 2 March – April 2013                                                                                Page 313
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

       Piece wise linear function                           method and OR method. In our work of congestion
       Sigmoid curve                                        control we have used the methods AND and OR.For AND
       Gaussian distribution function                       method we have used min and for OR method we have
       Quadratic and cubic polynomial curves.               used max. Rules weighting is done by implication method
                                                             .we are using weight 1 for the implication process.There
trapmf membership function comes under piece wise            are mainly three types of aggregation methods and these
linear function This membership function curve is used       three types are Max (maximum),Probor (Probabilistic
for output i.e for send rate.this is known as triangular     OR) and Sum (sumof each rules).In this work, we are
membership function.dsigmf and psigmf membership             using max aggregation method to do the aggregation .So
function comes under Sigmoid curve. This membership          that we can get the certain number of outputs.There are
function curve is used for input i.e for available           five types of defuzzification methods and these are
bandwidth.This       are     known      as    sigmonidal     Centroid, Bisector,Middle of max,Largest ofmax and
functions.gaussmf membership function comes under            Smallest of max.In this work, we are using centroid
Gaussian Distribution curve. This membership function        defuzzification process which gives the centre of area of
curve is used for input i.e for Tcp response.This are        the curve. Some tools used to built fuzzy system are:
known as Gaussian distribution curves. pimf membership
function comes under polynomial based curves This                   Fuzzy inference system(fis) editor:
membership function curve is used for input i.e for          Here FIS editor uses the input variables as available
Change Rate.These are known as Gaussian distribution         bandwidth,tcp response,change rate and output is send
curves.So the membership curves are created.these curves     rate.
are used because of their advantages.Some curves are                Membership function editor:
used because of having the advantage of simplicity like      Here in our work curves are made up of all the input
piece wise linear function.some are having the advantage     varaiable    and     output    available
of smoothness like gaussian curves.Fuzzy inference is the    bandwidth,change rate,tcp response and send rate.
process of deciding the mapping..The process of fuzzy               Rule editor:
inference is the combination of membership                   Here in our work we have made six rules while adding ,
functions,rules and logical operators.There are two types    deleting and putting rules.Also we have put operators to it
of fuzzy inference system :                                  (AND ,OR). Weights are also given in it.
                                                                    Rule viewer:
     Mamdani type fuzzy inference system
                                                             We have used six rules to seeits working in rule viewer.
     Sugeno type fuzzy inference system.
                                                                    Surface viewer:
The difference between these system is their                 Here in our work we use the surface viewer to see the
output..Mamdani fuzzy inference system is the first          output surface.It shows how the output of one is
commonly and properly used system.It was proposed by         dependent on inputs of two.
Ebrahim mamdani. Mamdani inference system gives the
output mambership function to be fuzzy sets as compared      For our work we are saving this fuzzy file as named
to sugeno type system can be used to make any inference      send.fis. Whenever we open it, we open it with command.
system in which the output membership functions are          Fuzzy send.fis When FIS editor opens up we will name
either linear or constant.In this work of congestion         three input variables available bandwidth, Tcp response,
control we are using mamdani fuzzy inference system          change rate and output as send rate.We are using
because we want the output fuzzy not linear not constant     mamdani fuzzy inference system. For AND method we
like in sugeno which gives linear or constant                are using min, OR method we are using max, for
results.Fuzzy inference system is divided into five parts:   implication method we are using min and aggregation we
                                                             are using max and for defuzzification we are using
       Fuzzification                                        centroid defuzzification method.Now it’s time to map
       fuzzy operator                                       curves for all the input and output and gives names to
       Implication                                          them and set their range.For input available bandwidth
       Aggregation                                          we are setting the range of 0 to 1 with sigmonidal curves
       Defuzzification.                                     where 0 represents low bandwidth and 1 represents high
For fuzzification our inputs are available bandwidth
which are set up to the range of 0 represents low
bandwidth and 1 represents high bandwidth,then TCP
response which are set up to the range of 0-1 where 0
represents low and 1 represents high,then change rate
which are set up to the range of -1-1 where -1 represents
decreasing and increasing and the output which are set up
to the range of 0-1 where 0 represents low and 1
represents high.The operators used in our work are AND

Volume 2, Issue 2 March – April 2013                                                                         Page 314
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

                                                             For input change rate we are setting the range of 0 to 1
                                                             with polynomial curves with display range -1 to 1 where -
                                                             1 represents and 1 represents decreasing change rate and
                                                             increasing change rate.

                                                                                     Figure 4

                                                             For input send rate we are setting the range of -1 to1with
                                                             piece wise linear function i.e. triangular curve.
                          Figure 1

                        Figure 2

For input tcpresponse we are setting the range of 0 to 1                             Figure 5
.curves used in it are Gaussian curves. Where 0 represents
low and 1 represents high.

                                                                                     Figure 6.

Volume 2, Issue 2 March – April 2013                                                                        Page 315
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

Here in our work we have made six rules which we can           viewer.This evaluation is done by filling all the X input,Y
add , delete and putting rules.Also we have put operators      input and Z output.after that whole sysem evaluation is
to it (AND ,OR). Weights are also given in it like 1.Rule      done. The inputs which we select for suface viewing are
viewer helps to view the rules of the system.It shows how      available bandwidth,tcpresponse and             output as
each rule behave in the system.It shows the output with        sendrate.With the help of menu’s like file ,edit and view
aggregation done and lastly the defuzzification output is      we can save edit and view our system rule by rule.This
displayed with the vertical bold line.we have used six         surface viewing grid helps us to see our system in actual
rules to see its working in rule viewer.                       form and we can also command its working.After surface
                                                               view we save this whole file. (send.fis)
Surface viewer helps to show the whole surface in micro
form.It also helps to see the system in x,y,z axis.The two
inputs selected are available bandwidth and tcp response
and output as sendrate. We can see all the three in three
domensionnal form..For smoothness we will give plot
value.By default its 101.By clicking evaluate all the
calculations will be done.For changing X axis and Y axis
after the suface is in view,change the input field and press
enter..If there will be more inputs then we use Ref. Input
field.Now this whole FIS file is saved in fuzzy logic
controller [3] block so that it can take the inputs like
available BW, tcp response, change rate to produce one
output send rate .Now the file which have been saved and
put it in controller will give us the output in the form of
model. This model will take all the inputs of the system
and named as such inputs name like available bandwidth
to input MF, then Tcp response and last change rate and
output as send rate. Rules are applied to the entire rule
column which produces the aggregation method of
maximum. Aggregation gives the range of outputs but the
defuzzification takes the centroid defuzzification and               Figure 8- Model for Controlling Congestion
produces one output and gives one output. The whole
mechanism is used to control congestion. This fuzzy            3. CONCLUSION
system helps to control congestion .in faster, easier and
better way.                                                    Fuzzy inference system is the mapping of input space to
                                                               output space. Mamdani fuzzy inference system is used to
                                                               control congestion. Here the Fuzzy Inference system file
                                                               send.fis is created in which there are input values like
                                                               available bandwidth, Tcp response and change rate but
                                                               this whole system is based on six rules. This fuzzy system
                                                               consists of inputs, membership function curves which
                                                               describe the curves of inputs and also the output. We can
                                                               view all our system through Rule viewer and Surface
                                                               viewer .Rule viewer helps us to see the whole functioning
                                                               of the system and the Surface viewer helps us to see the
                                                               whole system in micro form where change rate is set to
                                                               zero because computer system can’t show the surface in
                                                               more than X, Y and Z axis. Then a model is generated by
                                                               using fuzzy logic controller and send.fis file which takes
                                                               all inputs, rules, operators, implication method(min)
                                                               ,aggregation method(max) and gives the output send rate
                                                               .We can also view this system file in code form which
                                                               gives the summary of the send.fis file.

                                                               If we want this system should work in live then we need
                         Figure 7                              to generate code in C language and then call the
Through surface viewer we can now see our system in            controller so that it can run in any other environment or
micro form which was not possible with the help of rule        in live environment. We can also precede this test not
                                                               only for controlling congestion but also for Tcp
Volume 2, Issue 2 March – April 2013                                                                           Page 316
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

friendliness, Tcp smoothness and Multimedia’s Quality              Internet: Approaches and Directions” IEEE
Of Service (QOS) [6][10][18] like delay, jitter and losses.        Transactions on Circuits and Systems for Video
More over fuzzy inference system remains an active                 Technology, Vol. 11, No. 1, 2001.
research area. More applications can be designed for          [14.] Mani       Zarei,      Amir        MasoudRahmani,
fuzzy logic like DC motor; water level control etc and             RaziehFarazkish,     Sara     Zahirnia     ,“Fairness
also great improvements can be done for the previous               Congestion Control for a distrustful wireless sensor
applications which already exist.                                  network using Fuzzy logic” 10th         International
                                                                   Conference on Hybrid Intelligent Systems , 2010.
REFERENCES:                                                   [15.] Y.-G. Kim, J. W. Kim, and C.-C. J. Kuo, “TCP-
                                                                   friendly Internet video with smooth and fast rate
[1.] Jon C. Ervin, Sema E. Alptekin, “Fuzzy Logic                  adaptation and network-aware error control,”IEEE
     Control of a Model Airplane”, International                   Trans. Circuits Syst. Video Technol., vol. 14, page
     Conference on Systems, Man and Cybernetics, IEEE              no. 256–268, Feb. 2004
     1998.                                                    [16.] S. Mercy Shalinie, G. Preetha, S. Dina Nidhya, B.S.
[2.] Chrysostomou, A. Pitsillides, G. Hadjipollas, M.              Kiruthika Devi, “Fuzzy Adaptive Tuning of Router
     Polycarpou, A. Sekercioglu, “Fuzzy Logic Control for          Buffers for Congestion Control” International
     Active Queue Management in TCP/IP networks”,                  Journal of Advancements in Technology ,Vol 1,
     12th IEEE Mediterranean Conference on Control and             Pages 85-94, June 2010.
     Automation (MED’04), Kusadasi, Aydin, Turkey, 6-         [17.] Emmanuel Jammeh, Martin Fleury, and Mohammed
     9 June 2004.                                                  Ghanbari, “Delay-based Congestion Avoidance for
[3.] PuminDuangmanee         and      PeerapongUttansakul,         Video Communication with Fuzzy Logic Control”,
     “Implementation of Real time video streaming with             Packet Video, IEEE 2007.
     fuzzy logic controller”, International Conference on     [18.] Ditze. “Resource Adaptation for Mobile AV Devices
     Wireless and Signal processing, IEEE 2010.                    in the UPnP QoS Architecture. In Journal of
[4.] SomchaiLekcharoenand Chun Che Fung, “An                       MobileMultimedia, vol4 ,2006.
     Adaptive Fuzzy Control Traffic Shaping Scheme            [19.] Michael Ditze, Matthias Grawinkel, “Fuzzy Logic
     over Wireless Networks”, Proceedings of Asia-Pacific          Based Admission Control for Multimedia Streams in
     Conference on Communications, Pg no 177-180,                  the UPnP QoS
     2007.                                                    [20.] hitecture” 21st International Conference on
[5.] S. Mercy Shalinie, G. Preetha, S. Dina Nidhya, B.S.           Advanced Information Networking and Applications
     Kiruthika Devi, “ Fuzzy Adaptive Tuning of Router             Workshops IEEE ,2007
     Buffers for Congestion Control” International            [21.] Danny H. K. Tsang, BrahimBensaou, Shirley T. C.
     Journal of Advancements in Technology ,Vol 1,                 Lam, “Fuzzy-Based Rate Control for Real-Time
     Pages 85-94, June 2010.                                       MPEG Video”, IEEE Transactions no. 4 on Fuzzy
[6.] Runtong Zhang, Xiaomin Zhu, “Congestion Control               system, VOL. 6 1998
     Using Fuzzy Logic in QoS Networks” IEEE. vol ,           [22.] BehrouzSafaiezadeh, Amir MasoudRahmani and
     2006 .                                                        EbrahimMahdipour, “A New Fuzzy Congestion
[7.] ChuenChien lee, “Fuzzy-Logic in Control systems:              Control Algorithm in Computer Networks”, IEEE
     Fuzzy logic controller Part –I”, IEEE Transactions            International Conference on Future Computer and
     on systems and cybernetics, vol.20, no 2 1990.                Communication”, page no 314-317 April 03-05,
[8.] Fuzzy Logic Toolbox User’s Guide for using                    2009
     MATLAB.                                                  [23.] Michael, “Streaming Media Demystified” McGraw
[9.] Franc Kozamernik “Media streaming over the                    Hill, 2002.
     internet an overview of delivery technologies” EBU            bin/pbg/007138877X?
     Technical Department, page no 1- 15, oct 2002.
[10.] Yazeed A. Al-Sbou,“Fuzzy Logic Estimation System                     Ms. Yogini Bazaz received M.Tech. degree
     of Quality of Service for Multimedia Transmission”,                   in Computer Science & Engineering with
     International Journal of QoS Issues in Networking,                    specialization in Information Security from
     Vol. 1, No. 1, December 2010.                                         Central      University    of     Rajasthan,
[11.] VeselinRakocevic“congestion control for multimedia      Bandarsindri, India in 2012. She has been lecturing at the
     applications in the wireless internet”.                  SLITE University and current research area is
[12.] Rahul Malhotra, TejbeerKaur, “Dc Motor Control          Information security and Fuzzy logic.
     Using Fuzzy Logic Controller”, International Journal
     Of     Advance      Engineering       Sciences   And
     Technologies, Vol No. 8, Issue No. 2, 291 – 296,                      Mr.Sudesh Kumar received the M.Sc.
     2011.                                                                 degree in mathematics from Bikaner
[13.] Dapeng Wu, Yiwei Thomas Hou, Wenwu Zhu, Ya-                          University, India in 2005, and ME degree in
     Qin Zhang, Jon M. Peha, “Streaming Video over the                     Computer Science & Engineering from

Volume 2, Issue 2 March – April 2013                                                                         Page 317
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 2, Issue 2, March – April 2013                                          ISSN 2278-6856

Thapar University, Patiala, India in 2009. Presently, he is
working as Assistant Professor in IGNTU, Amarkantak
and his current research area is Number Theory
application for cryptography, Advance DSA and Fuzzy
Logic & Application.
             Mr. Sanjay Kumar Anand received the
             M.Sc. degree in IT and M.Tech. Degree in
             Information Technology from C-DAC,
             Noida, India in 2008. Presently, he is
working as an Assistant Professor in Central University
Rajasthan, India and his area of interest is NLP and Data

Volume 2, Issue 2 March – April 2013                                                Page 318

Shared By: