IRJET-A Survey on Red Queue Mechanism for Reduce Congestion in Wireless Network
- 1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 01 | Jan-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 99
A SURVEY ON RED QUEUE MECHANISM FOR REDUCE CONGESTION IN
WIRELESS NETWORK
Kachhad Krishna Chandulal
Department of Computer Engineering, Marwadi Education Foundation – Faculty of PG Studies & Res. in
Engg. & Tec. Rajkot, Gujarat, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Now a day’s use of wireless networks for access
the internet is increasing. It shows that use of the main
transport layer protocol TCP will be increase in near future.
The number of applications running over computer networks
has been increasing tremendously, which ultimately results in
congestion. Congestion increases both delay and packet loss.
Network congestion can be controlled by several methods,
such as Random Early Detection (RED) which is the most well
known and widely used queue mechanism to avoidcongestion.
Analysis of various RED mechanism base on that Fuzzy Logic
RED (FLRED) uses fuzzy logic to avoid linearity and
parameterization problem in RED and Early Congestion
Control (ECC) mechanism is dynamically changesthe value of
the window field from TCP header where as Explicit Non-
Congestion Notification (ENCN) is notifying non-congestionin
the router and Three section Random Early Detection (TRED)
is divided into three section light, moderate and high load for
calculate packet dropping probability also the Efficient LAL
Random Early Detection (ELALRED) is founded on the
principles of the RED mechanism with LAL philosophy but the
Hemi-Rise Cloud Model (CRED) was proposed non linear
packet loss strategy was used. Based on observation CRED
effectively controls the oscillation of the averagequeuelength.
Key Words: Congestion control, TCP, Active queue
management (AQM), Random early detection (RED),
Hemi-rise cloud (CRED), Fuzzy logic random early
detection (FLRED), Early congestion control (ECC),
Explicit non-congestion notification (ENCN), Nonlinear,
Three section random early detection (TRED).
1. INTRODUCTION
Wireless network is the transfer of informationbetweentwo
or more points that are not connected physically. A wireless
network is a flexible data communications system, which
uses wireless media such as radio frequency technology to
transmit and receive data over the air. Wireless means
transmitting signals using radio waves instead of wires.
Wireless networks use electromagnetic waves to
communicate information from onepointtoanotherwithout
any physical connection. The Transmission ControlProtocol
(TCP) is a transport layer protocol that provides reliable
data transfer, connection oriented service, flow control and
congestion control [3]. Transmission control protocol is
mostly used protocol in currentcommunicationnetworkand
over the internet. TCP also controls the rate of transmission
from the sender node across the network with end to end
feedback, through packet. The protocol supports reliable
data transport by establishing a connection between the
transmitting and receiving ends. The TCP protocol attains
congestion control through an end-to-end algorithm which
computes the appropriate sender congestion window by
means of the estimated traffic conditions [4]. Nowadays, the
rapid growing of network technology and applications,
network congestion has been increasingly serious, and the
congestion control is becoming more and more urgent.
Congestion control strategy based on intermediate node
(router) have been proposed and gotten attention to
compensate for the lack of the source end TCP [1]. Results of
congestion include a high delay, wasted resources, and even
global synchronization. The aim then would be to control
congestion or, more ideally, avoid congestion. Congestion
control techniqueshave been developed calledActiveQueue
Management (AQM) [3-5]. A congestion control scheme
based on active queue management has become a research
hot spot in the industry.
1.1 Active Queue Management (AQM)
Congestion in Internet occurs when the link bandwidth
exceeds the capacity of available routers which results in
buffer bloat problem. This results in long delay in data
delivery and wasting of resources due to lost or dropped
packets. To resolve these mentioned congestion problems
two approachesare identified. The mostbasicAQMapproach
is the Drop Tail (DT) scheme where packets arriving in a
queue are dropped with probability one when the queue is
full. However, DT can cause that all TCP flows through the
congested queue reduce theirtransmission ratesat thesame
time, because each TCP sourcereducesitswindoweverytime
it detects a packet loss. Also, in each congestion episode,
there is a high chance of dropping a packet from each active
flow. This phenomenon is known as global synchronization
[7]. Drop Tail queue suffer the problem of global
synchronization in which queue is over utilized and
underutilized at alternative period of time.
To overcome such inconveniences several other AQM
mechanismshave been proposed,in whichinsteadofwaiting
for the queue to be full to act they respond in advance to
congestion by dropping or marking packets. Accordingly,
Floyd and Jacobson[7]proposedanalgorithmcalledRandom
Early Detection (RED) which basically detects congestion by
estimating the average occupation of the queue. The RED
algorithm functions by detecting incipient congestion and
- 2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 01 | Jan-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 100
notifying the Transmission Control Protocol (TCP) by
probabilistically dropping packets before the queue when a
router fillsup.Briefly,the algorithm worksbymaintainingan
average queue size. Asthe average queue sizevariesbetween
the minimum(minth) and maximum(maxth) thresholds,the
packet dropping probability linearly changes between zero
and maximum drop probability Pmax. Thus, the packet
dropping probability function is linear to the change of the
average queue size (ave). If the average queue size exceeds
the maximum threshold, all arriving packets are dropped.
Since the packet droppingmechanism isbasedonthemoving
average algorithm, RED can control the transient congestion
by absorbing arrival rate fluctuations. Although RED is a
significant improvement over simple Drop Tail that simply
drops all incoming packets when a Drop Tail queue is full,
RED is particularly sensitive to the traffic load and the
parameters of the scheme itself [5].
RED algorithm forecasts buffer queue length first and
takes measures according to values predicted in advance
such as adjusting sending rate, which can avoid network
congestion to a certain extent and effectively solve the
problem of the global synchronization [1].
1.2 Overview of RED (Random Early Detection)
Floyd and Jacobson [7] proposed an algorithm called
Random Early Detection (RED) which basically detects
congestion by estimating the average occupation of the
queue. RED also known asRandom early drop andRandom
early discard. RED algorithm solve the problemof the global
synchronization but RED is sensitive to parameters and
traffic load which exists uncertainty. RED is highly sensitive
to its parameter settings namely minimum threshold
(maxth), minimum threshold (minth), maximum packet
dropping probability (Pmax) and weighting factor (Wq) [1].
RED uses a mechanism early detection of packet drop
without waiting to queue overflow. When congestion will
happen, router discards the arriving packets with certain
probability. This can inform the sender to adjust size of
sending window before congestion happen. RED gets the
average queue length of router to predict the network
congestion. RED algorithm maintains two parameters
maximum threshold and minimum threshold. RED uses the
weighted average function to predict the average queue
length and compares the results with the minimum and
maximum threshold predefined. If it is less than the
minimum threshold, the arrival packet will be forwarded
normally; if estimated result is greater than the maximum
threshold, all the arriving groupswillbediscardedotherwise
the packet reaching will be randomly discarded or tagged
according to certain probability [1].
The RED scheme drops packets with a probability by
computing the average queue length (ave) to notify traffic
sources about early stage of network congestion.Where Wq
is the weighting factor, ave is the average of the queue
length. In addition, the average queue length is expressed as
ave = (1 − Wq)ave + Wq q (1)
In the aforementioned formula, q is the instantaneous
queue length, and Wq ∈ [0, 1] is the weighting factor.
RED has three more parameters, i.e., minimum threshold
minth, maximum threshold maxth, and the maximum
dropping probability Pmax at maxth. If the average queue
length is below minth, RED dropsno packets. However,ifthe
average queue length increases above minth but is below
maxth, RED drops incoming packets with a probability
proportional to the average queue length linearly. When the
average queue length exceedsmaxth, all the arrivingpackets
are dropped. The packet dropping probability (Pb)beingthe
function of the average queue length is calculated by
0, ave∈[0, minth]
Pb= maxp(ave-minth /maxth-minth),ave∈[minth,maxth](2)
1, ave∈[maxth, +∞]
RED is an improvement over simple DropTail.REDavoids
the TCP starvation problem and global synchronization but
compared with Drop Tail, RED exhibits a lower delay and a
higher throughput and packet loss [2].
2. LITERATURE SURVEY
The Hemi-Rise Cloud Model (CRED) [1] was proposed a new
RED algorithm for congestion control and used non linear
packet loss strategy. RED algorithm avoids the global
synchronization but RED performance is sensitive to traffic
load andparametermeans randomnessand fuzziness.Hemi-
rise is typical clouds model that solving fuzziness and
randomness of RED algorithm. In CRED combine drop tail
algorithm with RED algorithm. Packet dropping strategy of
CRED is described below.
(1)Average queue length is calculated for each packet
arriving.
(2)If queue is not empty then
Lavg = (1 - Wq) * Lavg + Lnow * Wq (3)
Where Lnow is the currentlength ofrouter queue,Lavgisthe
average of the queue length and Wq is constant. If Lavg is
below the Lmin then packet is not drop. If Lavg increases
above Lmin but is below Lbuffer, CRED drops incoming
packets with a probabilityof averagequeuelength.According
to CRED Lbuffer is the Lmax.
P = Ptmp / (1 – count * Ptmp) (4)
Where Ptmp is the transitional drop loss rate. When Lavg
exceeds Lbuffer, all arriving packets are dropped. CRED
algorithms proposed effectively improve problems of
parametersand load sensitivity in RED and ARED algorithm.
CRED algorithm is also good for stability of the network.
- 3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 01 | Jan-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 101
Advantages: CRED algorithmis improve the stabilityandget
better performance than the RED. Also improve the
uncertainty of parameters. CRED effectively controls the
oscillation of the average queue length and stably operates
under various network environments. Quality of service is
greatly improved such as robustness and stability.
Disadvantages: It is complex queue mechanism because
necessary the cloud model propertiesEx (expectationvalue),
En (entropy) and He (ultra-entropy) for finding dropping
probability.
The Fuzzy Logic RED (FLRED) [2] was proposed a new
congestion control method which extends RED. FLRED uses
fuzzy logic to avoid the linearity and parameterization
problem in RED. FLRED uses two congestion indicators that
are aql (average queuelength) andDspec(delayspeculation).
At a time only one indicator is good, either aql is good or
Dspec is good. These two indicators are used to calculate the
Dp (dropping probability) for incoming packets within a
Fuzzy Inference Process(FIP). In FIP basedmethod,FLREDis
implemented in four sequential steps that are fuzzification,
rule evaluation, aggregationanddefuzzification.Thefirststep
of the process is fuzzification, in which triangular is used to
formulate the input and output linguistic variables. The
second step of the FIP process is rule evaluation, which
evaluates the rules. The third step of the FIP process is
aggregation, the probabilities of the output linguistic terms
obtained by the applied rules.Thelast stepof the FIP process
is defuzzification, in which the output linguistic variable
values are produced based on fuzzy set. For defuzzification
employed Center of Gravity (COG). In short forfindingDpget
one common point through aql and Dspec indicators.
Advantages: FLRED queuemechanismdecreasingbothdelay
up to 1.5 to 4.5% and packet lossup to 6 to 30%underheavy
congestion than RED and effective RED (ERED).
Disadvantages: Under light congestion, increase delay than
RED and ERED.
The Explicit Congestion Control (ECC) [3] presents a new
approach to improve the performance of TCP in ad hoc
networks. The ECC is based on adjusting the window field of
TCP segments. The ECC mechanism is dynamically changes
the value of the window field from TCP headers according to
the utilization of the router queue. When queue is higher
than the threshold Q*l the value of the window field of the
TCP header is reduced to the percentage (Q – q / Q) of bytes
available in the queue. If queue is not higher than the
threshold Q*l the value ofthewindow field ofthe TCPheader
is same.
w = Q – q / Q *w q > Q*l (5)
ECC operates in the same way as drop tail but one
characteristic is added to reduce the value of the window
field of the TCP header.
Advantages: In ECC approach the good put is 8.28% higher
than the Drop Tail and RED. Also ECC reduce the packet loss.
Comparing ECC approach is reduce an averagedelay 28.40%
less than Drop Tail and 36.21% smaller than RED.
Disadvantages: ECC approach only suitable for ad-hoc
networks and grid topology.
TheExplicit Non-CongestionNotification(ENCN)[4]presents
a new approach for Active Queue Management (AQM)
technique. Most of works are based on congestion
notificationmeans Explicit CongestionNotification(ECN)but
the absence of congestionnotifications still notresearch.The
ENCN approach overcome unwanted empty queue
phenomenon and take advantage ofthe non-congestionstate
of the queue. The TCP action with AQM action, every time
transmitter receives a marked ACK and it will modify its
congestion window. If instantaneous queue length (Qinst) is
lower than a threshold (Qmin), router markwithprobability
one (ENCN = 1). At transmitter side if TCP is in slowstart, it
will calculate its new congestion window. If TCP is in
congestion avoidance the sender will increase (W + 1) its
congestion window by one packet size per round trip
time(RTT). If instantaneous queue length (Qinst) is greater
thanor equal to a threshold (Qmin),router is unmark (ENCN
= 0) that time transmitter side if TCP is in slowstart it will
continue in slowstart and if TCP is in congestion avoidance
the sender will decrease (W – 1) its congestion window by
one packet size.
Advantages: ENCN approach gives better throughput
compared with other AQM techniques because empty queue
phenomenon causesperformancelossintermsofthroughput
and ENCN overcome the empty queue phenomenon.
Disadvantages:ENCNapproachgivesbetterthroughputonly
for TCP Reno and TCP sack variants than other variants.
TheThree section RandomEarly Detection (TRED) [5]based
on nonlinear RED presents a minimal adjustment to RED.
TRED aimed at solving RED’s link underutilization and large
delay in low andhigh traffic load scenario problems.InTRED
average queue length betweentwo thresholdsis dividedinto
three sections: light, moderate and high load. If average
queue size between the minth and minth + Δ, allow the
incoming packets with low load, where Δ = (maxth –
minth)/3. If average queue size between the minth + Δ and
minth + 2Δ, allow the incoming packets with moderate load
and average queue size between the minth + 2Δ and maxth,
allow the incoming packets with high load. TRED find the
packet dropping probability based on curve. Based on curve
formula finally get the packet dropping probability. Through
the simulation the performance of TRED solves RED’s low
bandwidth utilization in low load scenario andlarge delay in
high load scenario.
Advantages: Very little work needs to be done to migrate
from RED to TRED. Also TRED queue mechanisms reduce
delay 4.8ms at high load compare to RED.
- 4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 01 | Jan-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 102
Disadvantages: TRED queue mechanism not used Explicit
Congestion Notification (ECN).
The Efficient Learning Automata Like Random Early
Detection (ELALRED) [6] presentsa algorithm theconceptof
a Learning Automata Like (LAL) mechanism devised for
congestion avoidance in wired networks. ELALRED is the
operations of the existing RED congestion avoidance
mechanisms, augmented with an LAL philosophy. Total five
action used in ELALRED algorithm: Forced_drop,
Minimum_exceed, Unforced_drop, No_drop and
Maximum_exceed. Forced_drop is chosen when the average
queue size is above the maximum threshold (avg > maxth).
Minimum_exceed is chosen when the average queue size
exceeds the minimum threshold and average just crosses
minth (minth< avg < maxth). Unforced_drop is chosen when
the average queue size lies between the minimum threshold
and maximum threshold (minth < avg < maxth). No_drop is
chosen when the average queue size liesbelowtheminimum
threshold (avg < minth). Maximum_exceed is chosen whem
the average queue size exceeds the maximum threshold and
average just crosses maxth (avg <= maxth). ELALRED
maintain a maximum likelihood estimate of how profitable
action has been andthis is inferredbyexaminingtheestimate
vector d(t). The algorithm also maintain and update
probability vector p(t). The ELALRED mechanisms avoid
congestion in the network and maximize the number of
packets.
Advantages: ELALREDalgorithmreducesthepacketdropsat
the gateways compared to the LALRED and RED. Using
ELALREDmore packets are acknowledgedtothesender.Also
ELALRED scheme is superior to LALRED and RED because
suitable even networks with 100 nodes.
Disadvantages: ELALRED scheme is avoid congestion for
wired networks.
Table -1: Comparison of Various Queue Mechanisms
Queue
Name
Dropping
Probability
TCP
Variants
Queue
Size
ECN
Notification
Complexity
CRED[1] Yes TCP 300
Packet
Yes Yes
FLRED[2] Yes TCP 20 Packet No Yes
ECC[3] No TCP
NewReno
97KB No No
ENCN[4] No TCP
NewReno &
TCP Sack
17 Packet Yes No
TRED[5] Yes TCP
NewReno
120
Packet
No Yes
ELALRED[6] Yes TCP 100 &
1000
Packet
No Yes
3. CONCLUSIONS
The study and research of queue mechanism for congestion
control continues to be an active area for the researchers. In
this survey paper, first studied about the Random Early
Detection (RED) queue mechanism and their many variants.
After that surveyed the earlier queue management
techniques such as Drop Tail and RED. Then, came to the
active queue management techniques such asCRED, FLRED,
ECC, ENCN, TRED and ELALRED mechanism. This paper
gives an idea for various variants of RED queue mechanism
and variousdropping probability for controlling congestion
and each has its own limitations.
Many study and experimental results show that CRED
gives better performance in comparison with other
techniques such as FLRED, ECC, ENCN, TRED and ELALRED
because it maintains stable queue length. Overall CRED
effectivelycontrolsthe oscillationoftheaveragequeuelength
and stably operates under various network environments
also quality of service isgreatly improved such asrobustness
and stability.
ACKNOWLEDGEMENT
I acknowledge to Marwadi Education Foundation, my
parents, and journal team for their support in my research
work.
REFERENCES
[1] Y. Zhao, Z. Ma, X. Zheng and X. Tu, "An Improved
Algorithm of Nonlinear RED Based on Membership
CloudTheory," in Chinese Journal of Electronics, vol.26,
no. 3, pp. 537-543, 5 2017.
[2] Abualhaj, Mosleh M., Ahmad Adel Abu-Shareha, and
Mayy M. Al-Tahrawi. "FLRED: an efficient fuzzy logic
based network congestion control method." Neural
Computing and Applications: 1-11.
[3] M. Talau, M. Fonseca, A. Munaretto and E. C. G. Wille,
"Early congestion control: A new approach to improve
the performance of TCP in ad hoc networks," 2016 7th
International Conference on the Network of the Future
(NOF), Buzios, 2016, pp. 1-6.
[4] L. M. A. Sup, R. M. de Moraes and A. Bauchspiess,
"Explicit non-congestion notification: A new AQM
approach for TCP networks," 2017 13th International
Wireless Communications and Mobile Computing
Conference (IWCMC), Valencia, 2017, pp. 1239-1244.
[5] C. W. Feng, L. F. Huang, C. Xu and Y. C. Chang,
"Congestion Control Scheme Performance Analysis
Based on Nonlinear RED," in IEEE Systems Journal, vol.
PP, no. 99, pp. 1-8.
[6] S. G. Mahajan, "Efficient LALRED for Congestion
Avoidance Using Automata-Like Solution," 2015
International Conference on Emerging Information
Technology and Engineering Solutions, Pune, 2015, pp.
1-6.
- 5. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 01 | Jan-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 103
[7] S. Floyd and V. Jacobson, "Random early detection
gateways for congestion avoidance," in IEEE/ACM
Transactions on Networking, vol. 1, no. 4, pp. 397-413,
Aug 1993.
BIOGRAPHIES
Krishna Kachhad is currently a second-
year M.E. (Computer) student at
Marwadi EducationFoundation–Faculty
of PG Studies & Res. in Engg. & Tec.
Rajkot, Gujarat, India. Her research
interests primarily include Congestion
control in wireless network.