Fuzzy RED to Reduce Packet Loss in Computer network

Khulood. A. Nassar Ammar. A. Abdullah

Dept. of Computer Science College of Science University of Basrah

Abstract :
Congestion is the most important and greatest challenge that facing the transmission of data packets process. There
are many sites, it is possible that the network congestion occurs as network devices (e.g. routers and switches, etc...) as
well as transmission medium between network devices. There are a numerous of mechanisms to control congestion in
routers. Random early detection algorithm (RED) is one of the most known mechanisms of control on the routers queues
in order to reduce the proportion of the loss of packets. Random Early Detection algorithm is considered one of the active
queue management. This paper uses random early detection algorithm to control the queues routers where the queues has
been controlled by one of intelligent methods (fuzzy logic). The proposed method has used four parameters, which are
the length of the queue rate, the rate of change in the queue and the rate of delay as inputs for fuzzy system and the
dropping probability as output. The practical part has been done using object-oriented programming

language C ++ within OMNET++ environment.

Keywords: Active Queue Management (AQM), Congestion Control, Fuzzy Logic Control, RED algorithm, Fuzzy
RED.

1. Introduction
In communication networks, Intermediate nodes example the congestion that occurs in
(such as a routers), and transmission bandwidth intermediate nodes [1, 2].
play a mainly role during data packets
As is well known, the Internet offers a lot of
transmission. As a result of that computer services to users on the network, as well as share
communications networks consist of limited network resources among themselves, therefore
resources, which lead to increased competition occur competition between Internet users and
existing between users on computer networks could lead to the congestion collapse. One of the
resources, and thus increasing the greatest and most famous transport layer
competitiveness as well as network resources protocols of the Internet is the transport control
become more limited. This leads to an overload protocol(TCP) for several reasons variety such as
on the network, leading to deterioration of the multipath routing, route fluttering, and
efficiency of the network. One of the greatest retransmissions, packets belonging to the same
problems occur during the transmission of data flow may arrive out of order destination and
packets are data loss which is caused by a congestion control. TCP is oriented connection
problem of congestion. Congestion problem reliable protocol. Where all the TCP's processes
occurs when the overload on the network is is a part of the transport layer [3].
greater than the limited capacity of the network Essentially, internet depend on transport control
and thus cause this pro protocol therefore that an urgent necessity to
improve the quality of service through the use
blem to increase high in loss of data, and overload mechanisms of and Queuing management
frequently buffer and high delay in the transfer of algorithms which controls the router queue, where
data packets. The collapse of congestion occurs the packets are accepted or dropping it [2, 4].
when dropping data packets before it is delivered
to its planned destination, due to the occupancy
network sources other transfer data packets, for
2. Related Work generates congestion. Our scheme selects packets
This section will briefly clarify the previous that will be dropped in the bottleneck link, and
works, which controls the congestion that occurs drops them early in the edge router. We
in computer networks. In 2003, C. Chrysostomou conducted simulation to measure the
et al, present the results of a fuzzy logic control performance, and it shows that the proposed
approach to the implementation of RED – Fuzzy- scheme improves the efficiency of core networks
RED, they believe that with fuzzy logic we are [10].
able to use linguistic knowledge to implement
better understood nonlinear probability discard 3. Congestion control
functions, achieve better differentiation for packet Congestion is one of the greatest and most
discarding behaviors for aggregated flows, and so important issues in packet switching networks,
provide better quality of service to different kinds Congestion may occur in communications
of traffic whilst maintaining high utilization [5]. networks in the case if the number of packets sent
In 2008, Mohammad Hussein Yaghmaee et al. is greater than the capacity of the network that
proposed Fuzzy AQM computes the packet drop can handle with a limited number of data packets,
probability according to preconfigured fuzzy Congestion may occur in the connection link or
logic using the instantaneous queue length and intermediate nodes due to carry a lot of data
number of packets dropped in a period of time as packet which lead to the deterioration of the
input variables [6]. In 2010, Weiyan Liu et al. quality of service (such as high delay, low
proposed an effective fuzzy congestion control throughput, packet loss, etc..). Figure1 illustrates
algorithm based on fuzzy logic to calculate packet the concept of congestion [11, 12].
loss by using the queue length and the buffer There are two different terms in the concept of
usage ratio [7]. In 2012, Shilpa N. Ingoley et al. congestion must differentiate between them, the
use of fuzzy logic (FL) in controlling congestion first term is called congestion control and the
in computer network. Discussed FL based second term is called congestion avoidance,
methods show improvement in increased in Congestion control execution are similar as
throughput, reduction in delays and packet loss. curative thing and whereas the avoidance
Fuzzy Logic can be effectively use in controlling execution are similar as preventative thing, a
congestion at core as well as at bottleneck router. congestion control scheme tries to bring the
Output of one fuzzy system can be given as input network back to an operating state, while a
to other fuzzy system. It can be also efficiently congestion avoidance scheme tries to keep the
network at an optimal state. Without congestion
use in priority base fuzzy algorithms, which also control a network may cease operating whereas
gives better Quality of Service [8]. In 2014, networks have been operating without congestion
Baklizi et al. proposed technique uses the average
queue length and the delay rate as input linguistic
variables for a fuzzy logic system, he utilized avoidance for a long time. The point at which a
congestion control scheme is called upon depends
fuzzy logic system produces a single output that on the amount of memory available in the routers,
represents a packet dropping probability , which whereas, the point at which a congestion
in turn control and prevent congestion in early avoidance scheme is invoked is independent of
stage , the performance of proposed technique has the memory size. A congestion avoidance scheme
been evaluated and compared with regard to may continuously oscillate slightly around its
various performance measures , which are : mean goal without significant degradation in
queue length , throughput , average queue delay , performance, whereas, congestion control scheme
packet lost and packet dropping probability [9]. In tries to minimize the chances of going above the
2015, Younghoon Kim et al. propose Cooperative limit [11].
Random Early Detection (CoopRED) to avoid
inefficiency in core routers when incoming traffic Generally congestion control is classified into two
main types that are host centric algorithms and
router centric algorithms. Where the host centric
algorithms divided to open loop mechanisms and
close loop mechanisms [11, 13]. RED is active queue management algorithms.
One of the biggest problems with TCP’s
congestion control algorithm over drop tail
queues is that sources reduce their transmission
rates only after detecting packet loss due to queue
overflow. Since a considerable amount of time
may elapse between the packet drop at the router
and its detection at the source, a large number of
packets may be dropped as the senders continue
transmission at a rate that the network cannot
support. RED alleviates this problem by detecting
incipient congestion early and delivering
congestion notification to the end-hosts, allowing
them to reduce their transmission rates before
queue over flow occurs [2, 15, 16, 17].
Van Jacobson and Sally Floyd first introduced the
Figure1: concept of congestion RED algorithm in August of 1993. RED has been
1. Random Early Detection (RED) designed with the objective to minimize packet
The queue management algorithm, which is loss and queuing delay, avoid global
applied to a router, plays an important role to synchronization of sources, maintain high link
improve Quality of Service (QoS). Queuing utilization and remove biases against bursty
management algorithm is responsible for sources [15, 16].
accepting the arriving packets or not accepting
them and consequently it directly affects the The RED algorithm manages the queue in a more
packet loss quantity parameter [4]. active manner by dropping packets randomly with
From the point of dropping packets, queue increasing probability as the average queue size
management can be classified into two categories increases; the packet drop rate increases linearly
which are Passive Queue Management (PQM) from zero, when the average queue size is at the
and Active Queue Management (AQM). For RED parameter minthresh (denoted by minth), to a
example Drop Tail algorithm is a representative drop rate of when the average queue size reaches
maxthresh (denoted by maxth). One of RED’s
PQM algorithm which only sets a maximum main goals is to use this combination of queue
length for each queue at the router. When the length averaging (which accommodates bursty
queue length is smaller than the maximum length, traffic) and early congestion notification (which
all packets are accepted, and if the queue reaches reduces the average queue length) to
its maximum length all subsequent incoming simultaneously achieve low average queuing
packets are dropped until queue length decreases delay and high throughput (see Figure2).
to be less than the maximum length. It was that
under heavy load conditions, Drop Tail routers
cause global synchronization, a phenomenon in
which all senders sharing the same bottleneck
router link shut down their transmission windows
at almost the same time whereas Active queue
management (AQM) techniques provide
mechanisms to control the queue length (i.e. the
number of packets in a router's buffer) by actively
discarding arriving packets before the router's
buffer becomes full [3, 4, 12, 14].

Figure2: RED operation

2. Proposed Method respectively. The Average Queue Length is
classified into 4 linguistic variables that are
In this section, we present an intelligent approach Conservative, Middle, Aggressive and
(fuzzy logic system) to control the congestion that VeryAggressive of membership function as
occurs inside the router by dropping packets shown in Fig 3. Also the Queue Rate of Change is
dynamically. Our proposed approach is based on classified into 4 linguistic variables that are Few,
RED algorithm, which drops the packets by Medium, Alot and VeryALot of membership
dropping probability. Our system based on function as shown in Figure 4. And finally the
congestion measure's variable for RED algorithm Delay Rate is classified into 4 linguistic variables
such as average queue length (AQL). The that are Low, Med, high and VeryHigh of
proposed technique aims at obtaining more membership function as shown in Figure 5.
satisfactory performance measure results in terms Whereas, there are one output of this system,
of packet loss in the event of heavy congestion. Drop Probability. The Drop Probability classified
Fuzzy logic (FL) is aimed at a formalization of into 5 linguistic variables that are VeryLittle,
modes of reasoning which are approximate rather Little, Moderate, Much and VeryMuch of
than exact. It also provides an alternative solution membership function as shown in Figure 6.
to non-linear control because it is closer to real
world. Non-linearity is handled by rules,
membership functions, and the inference process
which results in improved performance. Fuzzy
Controller: A fuzzy controller works similar to a
conventional system: it accepts an input values,
performs some calculations and generate an output
value. The basic structure of Fuzzy System consist
of four main components. A Fuzzifier which
translates crisp (real valued) inputs into fuzzy
values. An Inference Engine that applies a fuzzy Figure3: AQL input variable (in packet)
reasoning mechanism to obtain a fuzzy output. A
Defuzzifier which translates this latter output into
a crisp values. A Knowledge Base which
contains both an ensemble of fuzzy rules known
as the rule base, and an ensemble of membership
functions, known as the database [8].
The proposed Fuzzy RED is based on an AQM
approach, which implements a drop probability
function, And try as much as possible to make
packets loss is much less. The principal aim of the
proposed Fuzzy RED is to achieve Low packet Figure4: QRC input variable (in second)
loss rate leads to increase in throughput and
decrease in delay thereby increasing the quality of
service. The proposed Fuzzy RED is compatible
with changing network and traffic conditions.
When the incoming traffic is Intense, the current
internet router buffers fail
to control congestion effectively whereas
proposed Fuzzy RED gives better performance.

In this system there are three inputs, Average

Queue Length, Queue Rate of Change and Delay
Rate which are denoted by AQL, QRC and DR Figure5: DR input variable (in second)
 The practical part has been done using
OMNET++ to simulate proposed method where
uses computer network that consist of 7 nodes As
shown in Figure7. Each nodes computing
probability of packet dropping based on proposed
Fuzzy RED

Figure6: DP output variable

Table1: The Rule base of Fuzzy Logic System

Figure7: Computer Network form

6. Result

This section presents the results for

Proposed Fuzzy RED algorithm. The results are
shown in Table 2. The Table 2 illustrates
probability of dropping for Proposed Fuzzy RED
algorithm. For example, as shown in Table 1 that
the Average Queue Length which represents the
first input value equal to 0 where this value
Stimulates membership function which is a
linguistic term is Conservative as shown in
previous

The fuzzy controller has a rule base,

according to Table 1, which makes decisions on
the basis of the rules in it. For example if the
Average Queue Length is Conservative, the
Queue Rate of Change is VeryALot and the
Delay Rate is VeryHigh, then the inference
engine searches the rule base and finds the Rule
15 of Table 1.
 Table2: The Result of Proposed Fuzzy RED
Algorithm

Figure8: Drop Probability of Fuzzy RED and

Original RED

7. Conclusion
This paper uses fuzzy logic (FL) in
controlling congestion in whole computer
network. Fuzzy Logic can be effectively use in
controlling congestion of network as well as at
bottleneck router. Proposed Fuzzy RED algorithm
gives better quality of service because it gives
very good evaluation to packet loss for router.
Then it leads to better throughput. Where the use
of the sensitive Proposed Fuzzy RED algorithm
Fig3, As for the second input Queue Rate of parameters has a very significant effect on
Change equal to 1.98 where this value Stimulates reducing the loss of packets such as average
membership function which is a linguistic term is queue length where it has been chosen as a first
VeryAlot as shown in previous Fig4, and the input to proposed fuzzy RED algorithm because it
third input Delay Rate equal to 1.8975 where this is considered mainly indicator of overflow queue
value Stimulates membership function which is a which represents the number of packets in the
linguistic term is VeryHigh as shown in previous queue where whenever the value of average
Fig5. As queue length is increased this indicates that the
number of packets in the queue in a significant
increase. And also Queue Rate of change has
a result, the value of the Drop Probability that been chosen as the second input to proposed
represent the output of the system will be equal to fuzzy RED algorithm, because is a sensitive and
3 where this value Stimulates membership accurate indicator of the rate of change for the
function which is a linguistic term is Little as packets in the queue whether an increase or
shown in previous Figure6. decrease in the number of packets in the queue at
In figure 8 illustrated the deference of dropping a given time period. While the third input that is
probability between original RED and our the delay rate, which refers to the accumulation of
proposed fuzzy RED. the packets in the queue for a long time and not to
leave more quickly leading to congestion inside
the router, where we conclude from these
effective parameter as a measure for delaying
packets in the queue thus gives an important
notification for these delays and therefore their
utility prevent congestion due to stay the packets
within the queue for a long period of time. So that
 conclude from this parameter as a real measure of
[6] Mohammad Hossein Yaghmaee, "A Modified
the speed or slow change the size of the queue
Random Early Detection Algorithm: Fuzzy Logic
leading to give notification accurate and thus lead Based Approach", Communications and
to an increase in the speed of processing packets Networks, Journal of 7.3 (2005): 337-352.
in router, And also giving an accurate alert before
the queue fills by packets which lead to [7] Liu, Weiyan, Shunyi Zhang, Mu Zhang, and
congestion. As well as the number of effective Tao Liu. "A fuzzy-logic control algorithm for
active queue management in IP networks."
parameters has a very significant effect in
Journal of Electronics (China) 25, no. 1 (2008):
controlling congestion happening in the network, 102-107.
where determine to number of the appropriate
parameters as inputs to the fuzzy logic system [8] Shilpa N. Ingoley, Madhu Nashipudi, "A
have a significant impact on the performance of Review: Fuzzy Logic in Congestion Control of
the network. In the proposed Fuzzy RED Computer Network", (IJCA) (0975 – 8887),
Shilpa N. Ingoley, Madhu Nashipudi, (ICRTITCS
algorithm has been used three parameters instead
- 2012).
of using two parameters has also been used in
other previous work in order to become a [9] Baklizi, Mahmoud, Hussein Abdel-Jaber,
notification process that there is congestion will Ahmad Adel Abu-Shareha, Mosleh M. Abualhaj,
happen more accurate. and Sureswaran Ramadass. "Fuzzy logic
controller of gentle random early detection based
on average queue length and delay rate."
8. References International Journal of Fuzzy Systems 16, no. 1
(2014): 9-19.
[1] Sridhar Madipelli, David Raj Gillella,
Sudhakar Devaraya, "The RED Algorithm [10] Jeong, Seongyun, Younghoon Kim,
Averaged Queue Weight Modeling for Non Donghyeok An, and Ikjun Yeom. "CoopRED:
Linear Traffic", Blekinge Institute of Technology Cooperative RED for software defined networks."
November 2009. In Information Networking (ICOIN), 2015
International Conference on, pp. 307-308. IEEE,
[2] Haydar Abdulameer Marhoon, "Performance 2015.
Evaluation of RED and REM Algorithms in
TCP/IP Networks", The First Scientific [11] Sapna Gupta, Nitin Kumar Sharma, K.P.
Conference the Collage of Sciences ,Karbala Yadav, "A Survey on Congestion Control &
University, 2013. Avoidance", VSRD-IJCSIT, Vol. 2 (9), 2012,
790-797.
[3] Alshimaa H. Ismail, Ayman EL-Sayed &
Zeiad, Ibrahim Z. Morsi, "Enhanced Random [12] Kiran Chhabra, Manali Kshirsagar, A. S.
Early Detection (ENRED)", International Journal Zadgaonkar, "Effective Congestion Indication for
of Computer Applications, Volume 92 – No.9, Performance Improvement of Random Early
April 2014. Detection", International Journal of Innovative
Technology and Exploring Engineering (IJITEE),
[4] Arash Dana, Ahmad Malekloo, " Performance ISSN: 2278-3075, Volume-3, Issue-2, July 2013.
Comparison between Active and Passive Queue
Management", IJCSI International Journal of [13] Alexander K¨uchler, Matthieu-Patrick
Computer Science Issues, Vol. 7, Issue 3, No 5, Schapranow, "Congestion Control", university
May 2010. potsdam, 2005.

[5] C. Chrysostomou, A. Pitsillides, L. Rossides, [14] K. Tabash, M. A. A. Mamun, A. Negi, " A

A. Sekercioglu, " Fuzzy logic controlled RED: Fuzzy Logic Based Network Congestion Control
congestion control in TCP/IP differentiated Using Active Queue Management Techniques", J.
services networks", Soft Computing 8.2 (2003): Sci. Res. 2 (2), 273-284 (2010).
79-92.
[15] Serhat ÖZEKES, "EVALUATION OF
ACTIVE QUEUE MANAGEMENT
ALGORITHMS" ,İstanbul Ticaret Üniversitesi
Fen Bilimleri Dergisi Yıl:4 Sayı:7 Bahar 2005/1
s.123-140.
 ‫‪[16] Omid Seifaddini, Azizol Abdullah, Hamid‬‬ ‫‪[17] Dolly Kalav, Sudha Gupta, "Congestion‬‬
‫‪Vosough,‬‬ ‫‪"RED,‬‬ ‫‪GRED,‬‬ ‫‪AGRED‬‬ ‫‪Control in Communication Network Using RED,‬‬
‫‪CONGESTION CONTROL ALGORITHMS IN‬‬ ‫‪SFQ and REM Algorithm", International‬‬
‫‪HETEROGENEOUS‬‬ ‫‪TRAFFIC‬‬ ‫‪TYPES",‬‬ ‫‪Refereed Journal of Engineering and Science‬‬
‫‪Proceedings of the 4th International Conference‬‬ ‫‪(IRJES), Volume 1, Issue 2 (October 2012).‬‬
‫‪on Computing and Informatics,ICOCI 2013 28-30‬‬
‫‪August, 2013 Sarawak, Malaysia. Universiti‬‬
‫‪Utara Malaysia.‬‬

‫االكتشاف الوبكر العشوائي الوضبة لتقليل فقذاى حزم البياناث في شبكاث الحاسوب‬

‫عوار عبذالهادي عبذهللا‬ ‫خلود احوذ نصار‬

‫جاهعت البصرة ‪ /‬كليت العلوم ‪ /‬قسن علوم الحاسوب‬

‫الوستخلص ‪:‬‬
‫يعذ انتضاحى يٍ اهى واعظى انتحذيبث انتي تىاجه عًهيت تشاسم حضو انبيبَبث‪ .‬حيج اٌ هُبنك انعذيذ يٍ انًىالع انتي يٍ انًًكٍ اٌ يحذث‬
‫فيهب انتضاحى كأجهضة انشبكت (عهً سبيم انًخبل انًىجهبث وانًبذالث‪..‬انخ) وكزنك ايضب وسبئظ انُمم بيٍ اجهضة انشبكت‪ .‬هُبنك انعذيذ يٍ انيبث‬
‫انسيطشة عهً انتضاحى في انًىجهبث‪ .‬خىاسصييت االكتشبف انًبكش انعشىائي(‪ )Random Early Detection‬هي يٍ اشهش انيبث انسيطشة عهً‬
‫طىابيش انًىجهبث ورنك نتمهيم َسبت فمذاٌ انحضو‪ .‬تعتبش خىاسصييت االكتشبف انًبكش انعشىائي‬

‫احذي خىاسصييبث اداسة انطىابيش انفعبنت(‪.)Active queue management‬‬

‫نمذ تى في هزا انبحج استخذاو خىاسصييت االكتشبف انًبكش انعشىائي نهسيطشة عهً طىابيش انًىجهبث حيج تى تطىيشهب بإحذي انطشق انزكيت‬
‫(انً ُطك انًضبب)‪ .‬حيج اَه تى استخذاو حالحت يعهًبث هي يعذل طىل انطببىس‪ ،‬يعذل انتغييش في انطببىس ويعذل انتأخيش كإدخبالث نهُظبو‬
‫انًضبب واخشاجب واحذا يتًخال ببحتًبنيت اسمبط حضو انبيبَبث‪ .‬تى بشيجت انجبَب انعًهي ببستخذاو نغت انبشيجت انكبئُيت انًُحً ‪ C++‬ضًٍ بيئت‬
‫بشَبيج يحبكبة انشبكبث ‪.OMNET++‬‬