DL-RED: a RED-based algorithm for routers

Full text

244 Int. J. Computer Applications in Technology, Vol. 70, Nos. 3/4, 2022
Copyright © 2022 Inderscience Enterprises Ltd.
DL-RED: a RED-based algorithm for routers
S.O. Hassan*
Department of Mathematical Sciences,
Olabisi Onabanjo University,
Ago-Iwoye, Nigeria
Email: samuel.hassan@oouagoiwoye.edu.ng
*Corresponding author
A.U. Rufai
Department of Computer Sciences,
University of Lagos,
Akoka, Lagos, Nigeria
Email: arufai@unilag.edu.ng
C. Ajaegbu
School of Computing Engineering Sciences,
Babcock University,
Ilishan-Remo, Nigeria
Email: ajaegbuc@babcock.edu.ng
F. Ayankoya
Computer Sciences Department,
Babcock University,
Ilishan-Remo, Nigeria
Email: ayankoyaf@babcock.edu.ng
Abstract: Keeping a small-size average queue size (which will in turn, offer a minimised delay
performance) is regarded an important goal of Active Queue Management (AQM) algorithms
implemented in internet routers. The long-established Random Early Detection (RED) algorithm
is unable to achieve this desired objective. Herein, we present an enhanced RED algorithm,
named Double Linear RED (DL-RED) which utilises a linear packet dropping function for
light- and moderate- network traffic load scenarios and another linear packet dropping function
for a heavy network traffic load scenario. The effectiveness of DL-RED was evaluated and
compared with two other well-researched algorithms using the ns-3 simulation platform.
Experimental results proved that DL-RED performed better than both prior algorithms with
reference to delay and throughput. Importantly, a little effort is required to amend the packet
dropping probability profile of RED’s algorithm implementation with DL-RED.
Keywords: AQM; congestion control; delay; DL-RED; simulation.
Reference to this paper should be made as follows: Hassan, S.O., Rufai, A.U., Ajaegbu, C. and
Ayankoya, F. (2022) ‘DL-RED: a RED-based algorithm for routers’, Int. J. Computer
Applications in Technology, Vol. 70, Nos. 3/4, pp.244–253.
Biographical notes: S.O. Hassan received his MSc and PhD degrees in Computer Science from
Obafemi Awolowo University, Ile-Ife, Nigeria. Currently, he is a Lecturer in the Department of
Mathematical Sciences (Computer Science Unit), Olabisi Onabanjo University, Ago-Iwoye,
Nigeria. He is a Member of Computer Professionals (Registration Council) of Nigeria (CPN),
International Association of Engineers (IAENG). His research interests spans computational
mathematics, computer networks and communications, mathematical modelling and simulation,
and internet congestion control.
A.U. Rufai is a Senior Lecturer at the Department of Computer Sciences, University of Lagos,
Akoka, Lagos, Nigeria. His areas of interests include software engineering, cognitive computing
and cyber security.
C. Ajaegbu is a Senior Lecturer in Babcock University, Ilishan-Remo, Nigeria. His research
interest includes wireless networking, humanised computing and information technology. He is
also an associate editor and reviewer of some high ranking journals.

DL-RED: a RED-based algorithm for routers 245
F. Ayankoya bagged a First Class Honours Degree in Computer Science from Bells University of
Technology, Ota. She obtained a PhD degree (in Computer Science) from Federal University of
Technology, Akure, Nigeria in 2016. She is presently a Senior Lecturer in the Computer Science
Department of Babcock University, Ilishan-Remo, Nigeria.
1 Introduction
Congestion, a critical concern in computer networks occurs
when the rate of data traffic transmitted over a network exceeds
the available buffer capacity of the network’s resource, that is
the router (Abu-Shareha, 2019; Kaur and Singhai, 2019;
Adamu et al., 2021). Avoiding congestion is considered an
essential issue in order to ensure an improved quality of service
(QoS) offered in network systems (Adamu et al., 2021; Adamu
et al., 2020; Hamdi et al., 2021). Congestion control remains a
persistent problem even in today’s Internet due to ever
increasing number of users (Adamu et al., 2021; Abdel-Jaber,
2020; Hamadneh et al., 2019; Abbasov and Korukoglu, 2009;
Zhou et al., 2006). The traditional queue management
algorithm – Drop-Tail used in routers lacked mechanism to
detect congestion early before the buffer gets full (Adamu et
al., 2021). Drop-Tail algorithm is challenged with problems
such as high packet loss rate, buffer overflow, network
deadlock, large delay in data delivery, and global
synchronisation (Braden et al., 1998; Brandauer et al., 2001).
To address the problems associated with Drop-Tail
algorithm, the Internet Engineering Task Force (IETF) strongly
recommends Active Queue Management (AQM) algorithms
for deployment in Internet routers as an approach for
controlling congestion (Braden et al., 1998).
AQM algorithms aimed at achieving two important goals.
First, to keep the average queue size small, which will in turn,
leads to an improve delay performance for interactive services.
Second, to increase throughput performance by reducing the
rate of packet dropped and avoiding lock-out phenomenon.
The Random Early Detection (RED) AQM algorithm
proposed by Floyd and Jacobson (1993) is perhaps the most
prominent among several AQM algorithms that exist in
literature (Abdel-Jaber et al., 2019; Adamu et al., 2020). RED
algorithm uses the average queue size (denoted avg
q) as an
indicator for congestion and is updated at every packet arrival
into the router. qavg is calculated as follows:
1 Suppose the queue is empty, qavg is computed according
to:

=1 '
m
avg q avg
qwq (1)
where '
avg
q refers to avg
q value computed previously
and q
w (which lies between 0 and 1) is a pre-defined
weight parameter for computing qavg.
In equation (1), m which refers to the idle time
parameter is computed as follows:
=( )m f time q idle time

 (2)
where ()
f
t represents a linear function of time t;
q idle time
 indicates the start of queue idle time and
time represents the current time.
2 However, suppose the queue is non-empty, avg
q is
calculated using which is an EWMA (exponential
weighted moving average), a low-pass filter according to:
=(1 ) ' ( )
avg q avg q
qwqwq

 (3)
where q indicates the current queue size.
Therefore, the packet dropping mechanism for RED is
explained as follows:
1 When
avg
q is lower than minTh (that is, the minimum
threshold for the queue), the packet is enqueued. That is,
initial packet dropping probability function ( b
p) is
computed as:
=0
b
p (4)
2 When
avg
q is between minTh and maxTh (that is, the
maximum threshold for the queue), b
p is computed
according to a linear function that changes from 0 to
max
p (that is, the maximum packet dropping probability
at maxTh ) as follows:
()
=max avg
b
pqminTh
pmaxTh minTh

 (5)
3 When avg
q exceeds maxTh , the incoming packet is
dropped with a probability of 1.
Mathematically, the dropping function for RED is expressed
as:
0, <
()
=,<
1,
avg
max avg
bavg
avg
qminTh
pqminTh
P minTh q maxTh
maxTh minTh
maxTh q







(6)
Thus,
1
=1
ab
b
pp count p

(7)
where a
p represents the final packet dropping probability
function and count represents the number of arrived packets
since the last dropped.

246 S.O. Hassan et al.
Studies on RED algorithm have become more prevalent
and prominent with each passing year. However, the
aggressiveness and suitability of an unaccompanied linear
drop function used by RED has brought about a tremendous
increase in enthusiasm among researchers both in the industry
and academia. Unlike several enhancements to RED
algorithm (which will be unveiled in Section 2), we propose a
minimal adjustment to the packet dropping mechanism for
the queue region between minTh and maxTh. In particular, a
two-linear packet drop functions utilised by the proposed
Double Linear - Random Early Detection (DL-RED)
algorithm aimed at improving the delay performance by
stemming the growth in average queue size.
The rest of the paper is arranged as follows: Section 2
gives related works. In Section 3, we give a description of the
proposed DL-RED algorithm. Next, in Section 4, we present
the simulation experiments and results. Lastly, Section 5
concludes the paper.
2 Related works
Floyd (2000) proposed a modified variant of RED algorithm
named Gentle RED (GRED). In GRED, a linear drop
function that changes from 0 to max
p is utilised when avg
q is
between minTh and maxTh , while another linear drop
function that changes from max
p to 1 is utilised when avg
q is
between maxTh and doubly maxTh . GRED obtained an
increased throughput performance.
MRED, an enhanced GRED algorithm was proposed by
Zhang et al. (2012) which utilised a quadratic drop function
that increases from 0 to max
p when avg
q is between minTh
and maxTh . MRED achieved an increased throughput
performance.
The idea of Stabilised Dynamic GRED (SDGRED in
short) proposed by Baklizi (2019) is to dynamically adjust
two important thresholds of GRED namely: minTh and
doubly maxTh . The algorithm attempted to stabilised avg
q
between minTh and doubly maxTh thresholds and achieved
an improved performance in terms of both queuing delay and
packet loss rate.
Hassan (2022) developed the RED-Linear Exponential
(RED-LE) algorithm. Specifically, RED-LE employs a linear
together with an exponential drop functions to increase the
effectiveness of RED algorithm. RED-LE obtained a shorter
delay and a lesser average queue size performance.
Similar to SDGRED, Baklizi (2020) dynamically adjust
two thresholds namely: maxTh and doubly maxTh and
stabilises avg
q between minTh and maxTh thresholds. The
algorithm improved the packet loss rate and delay
performance metrics.
The important idea of Congestion Control RED (CoCo-
RED) algorithm developed by Suwannapong and Khunboa
(2019) is to drop packets with a linear function (similar to
RED) when avg
q exist between minTh and maxTh
thresholds. However, an exponential function (according to
equation (8)) is utilised particularly when avg
q is between
maxTh and the maximum buffer capacity (represented by
K
). CoCo-RED addresses the high number of packet drop
drawback of RED.
=qavg
b
pab (8)
where
(1/ )
=
()
max
ln pmax
maxTh
KmaxTh
p
a
e
(9)
and
(1/ )
=
ln pmax
K
maxTh
be
 (10)
Again, Suwannapong and Khunboa (2021) proposed the
Enhanced CoCo-RED (EnCoCo-RED) algorithm whereby
the two thresholds: minTh and maxTh were adjusted based
on the level of congestion in order to improve the throughput
performance.
Paul et al. (2016) introduced a mid-point threshold called
Target (computed according to equation (11)) between
minTh and maxTh in RED to distinguish between a low and
a high network load conditions. The modified algorithm was
named Smart RED (SmRED).
=( )
2
maxTh minTh
Target minTh 
 (11)
In SmRED, a (nonlinear) quadratic drop function is utilised
for low load condition, while a square root drop function is
utilised for high load condition respectively. The SmRED’s
packet dropping function is expressed as follows:
2
0<
<
=
<
1
avg
avg
max avg
b
avg
max avg
avg
qminTh
qminTh
p minTh q Target
maxTh minTh
p
qminTh
p Target q maxTh
maxTh minTh
maxTh q













(12)
Paul et al. (2017) proposed SmRED-i as an extension for
SmRED. In SmRED-i, the packet dropping function is
expressed as follows:
1/
0<
<
=
<
1
avg
i
avg
max avg
bi
avg
max avg
avg
qminTh
qminTh
p minTh q Target
maxTh minTh
p
qminTh
p Target q maxTh
maxTh minTh
maxTh q

















(13)
where the parameter =i2, 3, 4, 5, ...

DL-RED: a RED-based algorithm for routers 247
The authors aimed at achieving a trade-off between delay
and throughput by using different values of parameter i.
Patel and Karmeshu (2019) aimed at improving network
performance at heavy congested traffic especially when
<
avg
minTh q maxTh by utilising an improved drop function
as follows:
1
1
()
=1 1
b
log p
pp
count






(14)
where
1=avg
max
qminTh
pp maxTh minTh





(15)
Zhou et al. (2006) proposed a (nonlinear) quadratic packet drop
function in lieu of a linear drop function utilised by RED. The
algorithm was called Nonlinear RED (NLRED in short).
NLRED obtained an improved throughput performance.
Zheng and Atiquzzaman (2000) described low throughput
as the key weakness of the RED algorithm. To address this
issue, a double linear packet dropping functions was
proposed, named Double Slope RED (DSRED).
The RED E (RED Exponential) is another RED-based
algorithm proposed by Abdel-Jaber (2020). In RED

E, an
exponential drop function is utilised that changes from 0 to 1
and eliminates the need for pmax parameter. At heavy traffic
load, RED E achieved an improved delay performance. The
drop function for RED_E is given as follows:
0, <
=,<
1,
avg
qminTh
avg
bavg
maxTh minTh
avg
qminTh
ee
PminThqmaxTh
ee
maxTh q










(16)
Another variant of RED named MultiRED (MRED) was
proposed by Fgee et al. (2017). This algorithm deploys a
classifier to splits packets into two virtual queues depending
of the traffic application protocol type (TCP or UDP). The
virtual queues utilises RED algorithm. MRED obtained an
improved packet loss performance especially for traffic
flows emanating from delay-sensitive applications.
Hassan et al. (2022) suggested the RED-Quadratic Linear
(RED-QL) algorithm. RED-QL takes into consideration two
packet dropping functions namely, quadratic and linear.
Under varying traffic load conditions, RED-QL achieved a
lesser end-to-end packet delay and a smaller average queue
size.
In an effort to the increase the effectiveness of RED,
Danladi and Ambursa (2019) proposed a modification to the
setting of maxTh threshold. Using the enhanced algorithm
named DyRED, maxTh is dynamically obtained in accordance
to the traffic condition. DyRED achieved an increased
throughput and a reduced packet loss rate.
Jamali et al. (2013) proposed the queue size growth
velocity as a congestion measure instead of the usual qavg
utilised by RED. The modified algorithm was named VRED
and obtained an improved performance in terms of utilisation.
The HERED algorithm proposed by Abbasov and
Korukoglu (2009) utilised a hazard packet drop function in
lieu of the linear function deployed by RED. HERED
obtained an improved throughput performance.
Feng et al. (2017) proposed the TRED (Three-Section
RED) algorithm which improved RED by dividing the
queue domain between minTh and maxTh into three
sub-sections in order to distinguish between low, moderate,
and high traffic load scenarios. At low traffic load
condition, a nonlinear drop function is utilised. At moderate
traffic load condition, a linear drop function is utilised.
At high traffic load condition, a nonlinear drop
function is utilised. TRED achieved a trade-off between
delay and throughput. An increased throughput was
reported to be obtained at low traffic load scenario, while an
improved delay performance was reported for high traffic
load scenario.
A modified RED algorithm named Q-learning RED
(QRED) was proposed by Su et al. (2018). In QRED, max
p
parameter was dynamically adjusted based on Q-learning in
order to obtain an improved performance in terms of
throughput and packet loss rate.
Abdel-Jaber et al. (2015) proposed two analytical models:
RED-Exponential and RED-Linear. Both models are based
on RED. However, instantaneous queue length is utilised as
the congestion measure.
The DcRED (Delay-Controller RED) algorithm
proposed by Abu-Shareha (2019) estimates the dropping
probability by calculating a delay parameter for every
incoming packet.
In this paper, a Double Linear - Random Early Detection
(DL-RED) algorithm is proposed. Unlike RED and some of
its variants which attempts to obviate the problem of large
delay by relying on an unaccompanied drop function, DL-
RED employs a mixture of two (i.e., linear and another
linear) dropping functions and utilises a target threshold in
order to distinguish between a light- and moderate- network
traffic loads and a heavy traffic load.
3 The proposed DL-RED algorithm
As depicted in Figure 1, the proposed Double Linear –
Random Early Detection (DL-RED) algorithm utilises two
linear packet dropping functions such that the operating
region of RED’s linear dropping function between minTh
and maxTh is divided into two segments. A new queue
threshold namely, Target (computed according to equation
(17)) is introduced between minTh and maxTh so as to
distinguish between light and moderate network traffic load
conditions, and a heavy- network traffic load.
=2 3
maxTh minTh
Target minTh





(17)

248 S.O. Hassan et al.
Figure 1 DL-RED’s packet dropping probability
After qavg is computed according to RED mechanism in
equations (1)–(3), a decision is made whether to accept or
drop an incoming packet based on the dropping probability as
follows:
1 When
avg
q is found between the operating region of
0– minTh , the packet will be accepted. That is,
=0
b
p (18)
2 When
avg
q is found between the operating region of
minTh –Target , the network is believed to operate
under light – and moderate – traffic load conditions.
Hence, the packet dropping probability for this domain is
calculated according to:
3( )
=2( 2 )
max avg
b
pqminTh
pmaxTh minTh

 (19)
The linear function in equation (19) changes from 0 to max
p.
3 Similarly, when avg
q is found between the operating
region of Target –maxTh , the network is believed to
operate under heavy traffic load condition. Hence, the
packet dropping probability for this domain is calculated
according to:
3(1 ) ( )
=max avg
bmax
pqTarget
pp maxTh minTh

 (20)
The linear function in equation (20) changes from max
p to 1
4 Lastly, if
avg
q exceeds maxTh , then the packet will be
dropped. That is,
=1
b
p (21)
The pseudocode for DL-RED algorithm is presented in
Algorithm 1.
4 Simulation experiments and results
We evaluate DL-RED AQM algorithm and compare it with
RED and NLRED algorithms using performance metrics such
average queue size, delay, and throughput under light,
moderate, and heavy network traffic loads in ns-3 simulator.
The network topology configuration is presented in Table 1.
A small buffer size of 25 packets and a mean packet size of
1000 bytes were used. AQM algorithms are deployed on the
router. The values set for both minTh and maxTh were
suggested by Abdel-Jaber (2020), while the values set for
both max
p and q
w were suggested by Floyd and Jacobson
(1993). Simulation time is 100 seconds.

DL-RED: a RED-based algorithm for routers 249
Table 1 Simulation setup
Parameter Value
Simulation topology Single router dumbbell
Capacity of bottleneck link 10 Mbps
Bottleneck delay 10 ms
Capacity of non-bottleneck link 100 Mbps
Non-bottleneck delay 5 ms
minTh 3 packets
Target 5 packets
maxTh 9 packets
TCP New Reno
max
p 0.1
q
w 0.002
4.1 Simulation scenario 1:
light TCP network traffic load
This simulation scenario which consists of 5 TCP flows
determines how the trio algorithms (RED, NLRED, and
DL-RED) handle light traffic congestion.
Figure 2(a) depicts the change in average queue size over
time. All trio algorithms reach an initial height of 7.08 and all
perform similar by bringing down the queue size. However,
we can observe from Table 2 that DL-RED obtained a
reduced and more stabilised control of the queue size when
compared with RED and NLRED.
Table 2 Comparison of network performance under light
traffic load
AQM Algorithm
Criterion
Average Queue
Size (Packets)
Delay
(ms)
Throughput
(Mbps)
RED 2.67 0.50 9.98
NLRED 2.66 0.48 9.98
DL-RED 1.85 0.47 9.99
Figure 2(b) compares the delay performance for the trio
algorithms. Analysis presented in Table 2 shows that
DL-RED performed better than RED and NLRED.
Figure 2(c) compares the throughput performance for the
trio algorithms under consideration. From the analysis shown
in Table 2, it can be observed that DL-RED performed well
when compared with RED and NLRED.
Figure 2 Average queue size, delay and throughput under light traffic condition
(a) Average queue size (b) Delay
(c) Throughput

250 S.O. Hassan et al.
Figure 3 Average queue size, delay and throughput under moderate traffic condition
(a) Average queue size (b) Delay
(c) Throughput
4.2 Simulation scenario 2: moderate TCP network
traffic load
This simulation scenario which consists of 20 TCP flows
determines how the trio algorithms (RED, NLRED, and
DL-RED) handle moderate traffic congestion.
Figure 3(a) presents the variations in instantaneous
average queue size over time. RED has reaches an initial
height of 6.96, NLRED has an initial height of 6.93, while
DL-RED reaches an initial height of 6.89. All trio algorithms
perform similar by bringing down the queue size. However, it
can be observed from Table 3 that DL-RED obtained a much
lower and a stabilised queue size when compared with RED
and NLRED.
Figure 3(b) compares the delay performance for the trio
algorithms under consideration. Analysis presented in
Table 3 confirms that DL-RED continues to outperform
both RED and NLRED.
Figure 3(c) depicts the throughput performance for the
trio algorithms. It can be seen in Table 3 DL-RED performed
optimally just like the other two algorithms.
Table 3 Comparison of network performance under moderate
traffic load
AQM Algorithm
Criterion
Average Queue
Size (Packets)
Delay
(ms)
Throughput
(Mbps)
RED 5.42 2.14 10.02
NLRED 5.28 2.12 10.02
DL-RED 3.73 1.97 10.01
4.3 Simulation scenario 3: heavy TCP network
traffic load
This simulation scenario which consists of 50 TCP flows
determines how the trio algorithms (RED, NLRED, and
DL-RED) handles heavy traffic congestion. Most times, the
effectiveness of any congestion control algorithm is
determined at heavy traffic scenario.
Figure 4(a) compares the change in average queue size
over time for all the three algorithms under consideration. It
can be observed RED obtained an initial height of 6.55,

DL-RED: a RED-based algorithm for routers 251
NLRED reaches to 6.02, while DL-RED reaches to 4.62. The
trio algorithms perform similar by bringing down the queue
size; however, DL-RED again demonstrated a better control of
the queue size that stabilises faster when compared with RED
and NLRED. Analysis presented in Table 4 further confirms
this improved performance of DL-RED. This is because at
heavy network traffic load when average queue size is near
maxTh , DL-RED increases the packet dropping probability
faster in order to ease congestion than both RED and NLRED.
Figure 4(b) compares the delay performance among the trio
algorithms. Analysis from Table 4 again confirms that DL-
RED performed better than RED and NLRED by achieving the
lowest delay. This is because at heavy traffic load when
congestion is serious, DL-RED aggressively drop packets
Figure 4(c) depicts the performance of RED, NLRED and
DL-RED in terms of throughput. As shown in the analysis of
Table 4, DL-RED continues to offer an optimal result.
Table 4 Comparison of network performance under heavy
traffic load
AQM Algorithm
Criterion
Average Queue
Size (Packets)
Delay
(ms)
Throughput
(Mbps)
RED 6.15 5.38 10.15
NLRED 5.78 5.29 10.15
DL-RED 4.42 5.23 10.07
Figure 4 Average queue size, delay and throughput under heavy traffic condition
(a) Average queue size (b) Delay
(c) Throughput

252 S.O. Hassan et al.
5 Conclusion
In this paper, an improved RED-oriented active queue
management algorithm named DL-RED which modifies the
packet dropping probability of RED has been presented.
Accordingly, DL-RED deploys a linear drop function for a
light and moderate traffic congestion conditions and another
linear drop function for a heavy-load traffic condition. DL-
RED obtained a surpassing network performance from the
aspect of obviating the challenges of increased the average
queue size and delay. Thus, DL-RED controls congestion
more effectively than RED and NLRED algorithms in all
simulation experiments considered. Going forward, it would
be interesting to evaluate the performance of DL-RED
algorithm alongside other mentioned AQM algorithms using
a network topology with a large buffer capacity.
References
Abbasov, B. and Korukoglu, S. (2009) ‘An active queue
management algorithm for reducing packet load rate’,
Mathematical and Computational Applications, Vol. 14, No. 1,
pp.65–72.
Abdel-Jaber, H. (2020) ‘An exponential active queue management
method based on random early detection’, Journal of Computer
Networks and Communications, Vol. 2020, pp.1–11,
DOI:10.1155/2020/8090468
Abdel-Jaber, H., Shehab, A., Barakat, M. and Rashad, M. (2019)
‘Igred: an improved gentle random early detection method
for management of congested networks’, Journal of
Interconnection Networks, Vol. 19, No. 2, pp.1–20,
DOI:10.1142/S021926591950004X
Abdel-Jaber, H., Thabtah, F. and Woodward, M. (2015) ‘Modeling
discrete-time analytical models based on random early
detection: exponential and linear’, International Journal of
Modeling, Simulation, and Scientific Computing, Vol. 6, No. 3,
pp.155028-1–155028-22, DOI:10.1142/S1793962315500282
Abu-Shareha, A.A. (2019) ‘Controlling delay at the router buffer
using modified random early detection’, International Journal
of Computer Networks and Communications, Vol. 11, No. 6,
pp.63–75.
Adamu, A., Shorgin, V., Melnikov, S., and Gaidamaka, Y. (2020)
‘Flexible random early detection algorithm for queue
management in routers’, LNCS, Vol. 12563, pp.196–208.
Adamu, A., Surajo, Y. and Jafar, M.T. (2021) ‘SARED: self-
adaptive active queue management scheme for improving
quality of service in network systems’, Journal of Computer
Science, Vol. 22, No. 2, pp.253–267.
Baklizi, M. (2019) ‘Stabilizing average queue length in active queue
management method’, International Journal of Advanced
Computer Science and Applications, Vol. 10, No. 3, pp.77–83.
Baklizi, M. (2020) ‘Weight queue dynamic active queue
management algorithm’, Symmetry, Vol. 12, No. 12, pp.1–16,
DOI:10.3390/sym12122077
Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering, S., Estrin,
D. et al. (1998) ‘Recommendations on queue management and
congestion avoidance in the internet’, RFC 2309.
Brandauer, C., Iannaccone, G., Diot, C., Ziegler, T., Fdida, S. and
May, M. (2001) ‘Comparison of tail drop and active queue
management performance for bulk-data and web-like internet
traffic’, Proceedings of the Sixth IEEE Symposium on
Computers and Communications, pp.122–129.
Danladi, S.B. and Ambursa, F.U. (2019) ‘DyRED: an enhanced
random early detection based on a new adaptive congestion
control’, Proceedings of the 15th International Conference on
Electronics, Computer and Coputation (ICECCO), pp.1–5,
DOI:10.1109/ICECCO 48375.2019.9043276
Feng, C-W., Huang, L-F., Xu, C. and Chang, Y-C. (2017)
‘Congestion control scheme performance analysis based on
nonlinear RED’, IEEE Systems Journal, Vol. 11, No. 4,
pp.2247–2254, DOI:10.1109/JSYST.2014.2375314
Fgee, E., Smeda, A. and AbooElgaseem, K. (2017) ‘MRED: an
algorithm to insure high QoS IP networks’, Journal of
Communications, Vol. 12, No. 4, pp.200–206.
Floyd, S. (2000) Recommendation on using the “gentle” variant of
RED. Available online at: https://icir.org/floyd/red/gentle.html
Floyd, S. and Jacobson, V. (1993) ‘Random early gateway for
congestion avoidance’, IEEE/ACM Transactions on
Networking, Vol. 1, No. 4, pp.397–413.
Hamadneh, N., Obiedat, M., Qawasmeh, A. and Bsoul, M. (2019)
‘HRED: an active queue management algorithm for TCP
congestion control’, Recent Patents on Computer Science,
Vol. 12, No. 3, pp.1–6.
Hamdi, M.M., Mahdi, H.F., Abood, M.S., Mohammed, R.Q., Abbas,
A.D. and Hamid, A. (2021) ‘A review on queue management
algorithms in large networks’, Proceedings of the 2nd
International Scientific Conference on Engineering Sciences,
pp.1–12, DOI:10.1088/1757-899X/1076/012034
Hassan, S.O. (2022) ‘RED-LE: a revised algorithm for active queue
management scheme’, Journal of Telecommunications and
Information Technology, Vol. 2, pp.91–97.
Hassan, S.O., Nwaocha, V.O., Rufai, A.U., Odule, T.J., Enem, T.A.,
Ogundele, L.A. and Usman, S.A. (2022) ‘Random early
detection-quadratic linear: an enhanced active queue
management algorithm’, Bulletin of Electrical Engineering and
Informatics, Vol. 11, No. 4, pp.2262–2272.
Jamali, S., Alipasandi, B. and Alipasandi, N. (2013) ‘VRED: an
improvement over RED algorithm by using queue length
growth velocity’, Journal of Advances in Computer Research,
Vol. 4, No. 1, pp.31–38.
Kaur, N. and Singhai, R. (2019) ‘Congestion control scheme
using network coding with local route assistance in mobile
adhoc network’, International Journal of Computer
Applications in Technology, Vol. 60, No. 3, pp.242–253,
DOI:10.1504/IJCAT.2019.100298
Patel, S. and Karmeshu (2019) ‘A new modified dropping for
congested networks’, Wireless Personal Communications,
Vol. 104, No. 1, pp.37–55, DOI:10.1007/s11277-018-6007-8
Paul, A.K., Kawakami, H., Tachibana, A. and Hasegawa, T. (2016)
‘An AQM based congestion control for ENB RLC in 4G/LTE
network’, Proceedings of the 2016 IEEE Canandian
Conference on Electrical and Computer Engineering
(CCECE), pp.1–5.
Paul, A.K., Kawakami, H., Tachibana, A. and Hasegawa, T. (2017)
‘Effect of AQM-Based RLC buffer management on the eNB
scheduling algorithm in LTE network’, Technology, Vol. 5,
No. 59, pp.1–7.

DL-RED: a RED-based algorithm for routers 253
Su, Y., Huang, L. and Feng, C. (2018) ‘QRED:
a Q-Learning-based active queue management scheme’,
Journal of Internet Technology, Vol. 19, No. 4,
pp.1169–1178.
Suwannapong, C. and Khunboa, C. (2019) ‘Congestion control in
CoAP observe group communication’, Sensors, Vol. 19,
No. 3433, pp.1–14, DOI:10.3390/s19153433
Suwannapong, C. and Khunboa, C. (2021) ‘EnCoCo-RED:
enhanced congestion control mechanism for CoAP observe
group communication’, Journal of Ad Hoc Networks, Vol. 112,
pp.1–10, DOI:10.1016/j.adhoc.2020.102377
Zhang, Y., Ma, J., Wang, Y. and Xu, C. (2012) ‘MRED: an
improved nonlinear RED algorithm’, Proceedings of the 2011
International Conference on Computer and Automation
Engineering (ICCAE 2011), Vol. 44, No. 2, pp.6–11.
Zheng, B. and Atiquzzaman, M. (2000) ‘DSRED: an active queue
management scheme for the next generation networks’,
Proceedings of the 25th Annual IEEE Conference on Local
Computer Networks, IEEE Computer Society, pp.242–251.
Zhou, K., Yeung, K.L. and Li, V.O.K. (2006) ‘Nonlinear RED: a
simple yet efficient active queue management scheme’, Journal
of Computer Networks, Vol. 50, pp.3784–3794.