Optimized Two-Level Scheduler for Video Traffic in LTE Downlink Framework

Abstract

The most challenging issue in long-term evolution (LTE) mobile networks is the design of an optimal scheduler. The main target of the LTE scheduling strategy is to guarantee a certain level of quality of service (QoS) in order to satisfy users' needs. In this work, a novel queue-optimized two-level scheduling strategy is introduced. The queue-optimized scheduler has the ability to support both real-time (RT) and non-RT (NRT) services in the LTE downlink framework. The first level of the proposed scheduler depends on an optimized queue length threshold that is dynamically adapted to adjust the assigned resources for the video traffic to maintain the throughput around its average value. The proposed scheduling strategy is utilized to efficiently enhance the radio resource allocation process for video traffic by minimizing the allocation of any excess unwanted resources, which consequently facilitates the support of more video connections and enhances the overall network spectral efficiency. The distribution of the resources among users for RT and NRT applications is actually performed at the second level of the scheduling process. The physical resource blocks (PRBs) are assigned according to the enhanced earliest due date first (EEDDF) mechanism. The proposed scheduling strategy prioritizes users according to their service requirements, to improve the QoS provision. The proposed scheduler maximizes the video applications throughput, and generally, minimizes the packet delay and the packet loss rate (PLR) for RT applications.

Full text

RESEARCH ARTICLE
Optimized two-level scheduler for video traffic in LTE
downlink framework
Mohamad I. Elhadad
1,3
| Walid El-Shafai
2,3
| El-Sayed M. El-Rabaie
3
|
Mohammed Abd-Elnaby
4
| Fathi E. Abd El-Samie
3,5
1
Department of Electrical Communication
and Computer Engineering, High
Institute of Engineering and Technology,
Tanta, Egypt
2
Security Engineering Lab, Computer
Science Department, Prince Sultan
University, Riyadh, Saudi Arabia
3
Department of Electronics and Electrical
Communications Engineering, Faculty of
Electronic Engineering, Menoufia
University, Menouf, Egypt
4
Department of Computer Engineering,
College of Computers and Information
Technology, Taif University, Taif, Saudi
Arabia
5
Department of Information Technology,
College of Computer and Information
Sciences, Princess Nourah Bint
Abdulrahman University, Riyadh, Saudi
Arabia
Correspondence
Walid El-Shafai, Department of
Electronics and Electrical
Communications Engineering, Faculty of
Electronic Engineering, Menoufia
University, 32952 Menouf, Egypt.
Email: eng.waled.elshafai@gmail.com
Funding information
Taif University Researchers Supporting
Project Number, Grant/Award Number:
TURSP-2020/147
Summary
The most challenging issue in long-term evolution (LTE) mobile networks is
the design of an optimal scheduler. The main target of the LTE scheduling
strategy is to guarantee a certain level of quality of service (QoS) in order to
satisfy users' needs. In this work, a novel queue-optimized two-level schedul-
ing strategy is introduced. The queue-optimized scheduler has the ability to
support both real-time (RT) and non-RT (NRT) services in the LTE downlink
framework. The first level of the proposed scheduler depends on an optimized
queue length threshold that is dynamically adapted to adjust the assigned
resources for the video traffic to maintain the throughput around its average
value. The proposed scheduling strategy is utilized to efficiently enhance the
radio resource allocation process for video traffic by minimizing the allocation
of any excess unwanted resources, which consequently facilitates the support
of more video connections and enhances the overall network spectral effi-
ciency. The distribution of the resources among users for RT and NRT applica-
tions is actually performed at the second level of the scheduling process. The
physical resource blocks (PRBs) are assigned according to the enhanced earli-
est due date first (EEDDF) mechanism. The proposed scheduling strategy pri-
oritizes users according to their service requirements, to improve the QoS
provision. The proposed scheduler maximizes the video applications through-
put, and generally, minimizes the packet delay and the packet loss rate (PLR)
for RT applications.
KEYWORDS
EEDDF, LTE, PLR, QoS, RT, scheduling, spectral efficiency, video traffic
1|INTRODUCTION
Recently, long term evolution-advanced (LTE-A) has been considered as the most promising cellular network. The
LTE-A cellular system is envisioned by the Third Generation Partnership Project (3GPP).
1
Services such as online gam-
ing, video and voice calls with high quality and Internet browsing with high speed are now available as a consequence
of LTE-A network deployment.
1
Of course, each type of these services requires specific quality of service (QoS) pre-
requisites, in order to be delivered efficiently to the end users. The main task of any packet scheduling strategy is to
allocate the available radio resources among active users taking into consideration different QoS requirements for each
Received: 2 July 2020 Revised: 18 November 2020 Accepted: 24 November 2020
DOI: 10.1002/dac.4704
Int J Commun Syst. 2021;e4704. wileyonlinelibrary.com/journal/dac © 2021 John Wiley & Sons Ltd. 1of15
https://doi.org/10.1002/dac.4704

application.
2
For example, the scheduler tries to minimize both the packet drop rate (PDR) and the packet delay, and at
the same time maximize the delivered throughput for video services.
The 3GPP adopts the orthogonal frequency division multiple access (OFDMA) as a downlink (DL) transmission
technique; however, for the uplink (UL) transmission, the single-carrier frequency division multiple access (SC-FDMA)
technique is utilized.
3
These new techniques enable LTE-A mobile networks to satisfy the necessary requirements
needed to support different applications. The LTE-A systems have the ability to provide higher throughput, good immu-
nity to multipath fading, and spectrum efficiency.
4
Upon the OFDMA technique, a subset of the available bandwidths
is assigned to the services. The minimum subset of the available bandwidths to be assigned to all services is called the
physical resource block (PRB).
5
The PRB is determined in both time and frequency domains. In the frequency domain, it is constructed by 12 sub-
carriers, and each subcarrier has a 15 kHz bandwidth. Therefore, the total bandwidth for each PRB is 180 kHz. How-
ever, for the time domain, the allocation time is 1 ms.
6
So, the available downlink bandwidth in LTE-A systems is
divided into a number of PRBs, and each has an allocation time of 1 ms.
7
Continuing with the time-domain description,
the LTE/LTE-A frame has a 10 ms length, and it is constructed by 10 subframes. Each subframe has two slots to carry
14 OFDM symbols, with the length of each slot as 0.5 ms. The LTE/LTE-A scheduler allocates the available PRBs
among all active services every transmission time interval (TTI) of 1 ms.
8–10
As mentioned previously, different types of services have different QoS requirements. Mainly, there are two types of
traffic: real-time (RT) services and non-RT (NRT) services.
11
In LTE/LTE-A networks, generally, the radio resource
management (RRM) and specifically the packet scheduler are responsible for achieving the required QoS for each traffic
type with the efficient usage of the available PRBs. Each scheduler has two main tasks, firstly, to assign the available
PRBs to the appropriate users, and secondly, to determine how the data will be transmitted in each PRB. The design of
an optimal scheduling strategy plays a vital role to achieve the scheduler main tasks and guarantee the required QoS
for each service with an efficient usage of the available PRBs. Several scheduling techniques have been introduced in
the literature,
12–17
and each technique has its own advantages and disadvantages. Generally, the packet scheduler
working mechanism is similar to an ordinary multiplexer, where packets to be delivered next are accumulated in a
queue system and buffered waiting for the available PRBs to be assigned. Each buffer has several queues, so that each
traffic type is stored in a different queue. Each queue can be defined for instance by source and destination Internet
protocol (IP) addresses.
18–20
Due to the special nature of RT applications, they should be considered as high priority applications.
21,22
Accord-
ingly, a new queue-optimized scheduling strategy for supporting both RT and NRT services is introduced in this paper.
The introduced technique depends on a queue-optimized scheduling strategy, and it is designed to give high priority to
RT packets. An optimized queue length threshold is dynamically adapted around the average bandwidth required for
the video traffic. The proposed scheduling strategy is utilized to efficiently enhance the radio resource allocation process
for video traffic. On the other hand, the allocation process is performed at the second level of the scheduling process.
The PRBs are assigned according to the enhanced earliest due date first (EEDDF) mechanism. The proposed queue-
optimized scheduling strategy supports RT and NRT services. It prioritizes users and gives high priority to the traffic
type with high QoS requirements.
This paper is organized as follows. Section 2 gives a discussion of the most recent related work. The scheduling pro-
cess in LTE cellular networks is reviewed in Section 3. The proposed queue-optimized scheduling strategy is introduced
in Section 4. Section 5 illustrates how to evaluate the performance of the proposed queue-optimized scheduling strategy
and presents the simulation results. The conclusion is given in Section 6.
2|LITERATURE REVIEW
The most recently proposed scheduling techniques in LTE/LTE-A cellular networks are discussed in this section.
Satheesh et al.
23
provided a comparative study of six scheduling techniques utilized in LTE networks. According to
Satheesh et al.,
23
the exponential rule (EXP-rule) scheduling scheme achieves the lowest PLR, the lowest packet delay,
the highest fairness level, the highest throughput, and the highest spectral efficiency. Therefore, they concluded that
the EXP-rule technique is the most suitable scheduler for LTE networks. However, according to Farhana et al.,
24
the
frame level scheduler (FLS) is the most suitable scheduler for LTE networks.
Another comparative analysis of different scheduling techniques was conducted in Jabbar and Abdullah.
25
The con-
sidered algorithms in Jabbar and Abdullah
25
can only support NRT traffic. Grigol and Sergi
26
developed a customized
2of15 ELHADAD ET AL.

adaptive scheduling scheme aiming to maximize the system capacity. In addition, the algorithm in Grigol and Sergi
26
was introduced to enhance the spectral efficiency of the network without any guarantee to achieve the required QoS for
supporting RT applications. Actually, there are several scheduling techniques that were proposed in the literature to
support only NRT applications.
27,28
In this research, RT applications play a vital role in designing an optimal scheduling
strategy. Accordingly, the discussion will be concentrated on scheduling techniques that support both RT and NRT
applications.
Elhadad et al.
29
proposed a scheduling scheme for supporting both RT and NRT services. The assignment of PRBs
to video users is accomplished in two steps. The first step is performed based on an optimized delay threshold, while
the second step is performed according to each packet delay. The scheme in Elhadad et al.
29
ranks the priority of RT
packets according to their head of line (HoL) delay. Hence, the RT packets with the highest HoL delay take the highest
priority. The NRT traffic is scheduled according to the proportional fair (PF) scheduler. Another simple scheduling
technique was proposed in Elhadad et al.
30
The PRBs are distributed among active users based on three main parame-
ters. These parameters are the channel quality for each user, the average data rate delivered for each user until the time
of allocation of the current PRB, and the HoL packet delay. A survey of DL scheduling techniques was presented in
Lawal et al.
31
The authors classified the scheduling schemes into two main types: QoS-aware and QoS-unaware. A
comparative analysis was presented including the operational procedure, strengths, and weaknesses of each technique.
A combined scheduling scheme was proposed in Angri et al.
32
It is designed by combining the metrics of more than
one scheduler in order to gain the advantages of each scheduling technique. Another version of a combined scheduling
scheme was proposed in Angri et al.
33
It is designed for high mobility and dense area scenarios in order to satisfy the
QoS requirements of RT applications. An optimal DL scheduling technique was formalized in Nogueira et al.,
34
in
which the authors tried to implement the scheduling decision on Fifth Generation (5G) radio networks.
According to Capozzi et al.,
35
the modified-largest weighted delay first (M-LWDF) scheduling scheme was designed
to support multiple RT users, each with different QoS pre-requisites. The M-LWDF scheduling technique considers the
HoL packet delay and the characteristics of the channel. In this technique, the available PRBs are assigned to RT packets
based on a delay-weighted metric. On the other hand, for NRT packets, a conventional PF mechanism is utilized.
35
A
vital observation of the M-LWDF scheduler is that its performance changes according to the QoS requirements of
each traffic type. The authors in Christantus et al.
36
proposed an improved version of the MLWDF scheduler. This
algorithm is termed as improved-MLWDF (I-MLWDF), which is assumed to enhance the performance of LTE cellular
system with the ability to support both RT and NRT applications.
36
Unfortunately, regarding the simulation results, the
performance of the I-MLWDF scheduler seems to be approximately like that of the conventional M-LWDF scheduler.
Sadiq et al.
37
designed a practical LTE downlink scheduler and characterized its performance for three traffic sce-
narios, namely, full buffer, streaming video (loose delay constraint), and mixed streaming and live video (tight delay
constraint). They showed that EXP-rule and LOGarithmic rule (LOG-rule) can support a mix of QoS traffic, while
increasing system capacity in terms of the number of users that can be supported. and at the same time, they reduce
resource utilization. A new channel-aware integrated time and frequency-based downlink LTE scheduling (ITFDS)
algorithm was proposed in Tuan et al.
38
The ITFDS algorithm distributes the resources to all users together using
time-based and frequency-based scheduling. It works well for RT flows, and also maintains fairness among users. The
ITFDS algorithm does not have a good performance compared to other algorithms in terms of aggregated throughput.
It is also deployed for 4G LTE, but not for 5G technologies.
According to Ahmed et al.,
39
a new policy was introduced. It simultaneously performs optimal sub-band assignment
and rate allocation, by taking into account channel quality and queue backlogs of each user. The technical novelty of
this technique lies in exploiting a limit theorem on the best SNRs reported by the users and combining them within a
Lyapunov stability framework. A comparative study of different scheduling algorithms and almost blank subframe
(ABS) for LTE heterogeneous networks (HetNets) was presented in Thienthong et al.
40
The study focuses on the com-
parison of the system performance, when operating with different schedulers as well as under different cell range
expansion (CRE) and ABS parameters. Kaleem et al.
41
improved the performance of the conventional fractional
frequency reuse (FFR) scheme by considering the QoS requirements of cell-edge users and efficiently allocating the
non-occupied center-zone frequency bands to those users. The authors claimed that their scheme almost doubles
the cell-edge user throughput and reduces the user packet loss rate. Kaleem et al.
42
improved the performance of the
conventional coordinated multi-point (CoMP) transmission scheme by introducing a QoS priority-based coordinated
scheduling and hybrid spectrum access (QoS-CSaHSA) scheme for downlink CoMP transmission in two-tier networks.
The QoS-CSaHSA scheme dynamically reduces the femtocell power requirements by considering the neighboring cell
interference, and it also balances the macro-cell load by switching the femtocells to hybrid access (HA) mode.
ELHADAD ET AL.3of15

From the study and performance analysis of the previously published scheduling techniques, it is concluded that
most of these techniques suffer from several defects. The most critical weakness points are simply summarized in four
main points as follows: increased complexity of the scheduler, dependency of the scheduler performance on the QoS
requirements for different users, decreased throughput of RT applications, and large number of dropped RT packets.
The design of the proposed queue-optimized scheduling strategy in this paper is conducted considering the above-
concluded weaknesses.
3|LTEPACKETSCHEDULING
As mentioned previously, any scheduler should be capable of assigning the available PRBs among active users
according to a predefined metric. Utilizing the metric, the scheduler decides to allocate a specific PRB in a given TTI to
the appropriate user. In addition, it decides to postpone another user to the next TTI in order to achieve a certain QoS
level, increase the fairness level among active users, enhance the spectral efficiency of the network, and provide service
priorities.
43
Of course, different users have different QoS prerequisites, including different packet delay threshold and
different throughput requirements. An optimum scheduler should take into consideration all these points and
assign PRBs to active users, accordingly. All packets that will be delayed are stored in queues based on their QoS
requirements.
44
In LTE/LTE-A cellular networks, the evolved Node B (eNB) takes all scheduling decisions, whether it is UL or
DL. The scheduling algorithm performs one of the main tasks of the medium access control (MAC) layer at the eNB.
It performs both the DL and UL radio resource allocation.
45
The scheduling algorithm at the eNB decides which
user takes a specific PRB and sends this information as a control signaling through the physical downlink control chan-
nel (PDCCH). On the other hand, the actual DL user traffic is transmitted through the physical downlink shared
channel (PDSCH). This process is a dynamic process with a repetition period of 1 ms (TTI).
46
The scheduler decision is performed with the consideration of the instantaneous radio channel quality for every
user. So, each user calculates the instantaneous radio channel quality in each PRB and sends a status report for
every TTI in the UL direction. The radio channel status report is commonly known as channel quality indicator (CQI).
The CQI values can be estimated from standard reference signals received by every user in the DL. According to 3GPP
standardization, reference signals are embedded in the PRBs. The channel status report plays a vital role in the alloca-
tion of PRBs for downlink transmission.
47
The scheduling process during each TTI can be summarized in the following few points
48
:
•Upon the CQI value, the scheduling algorithm has to make a decision and determine which modulation and coding
scheme (MCS) should be used for data transmission. The keyword in this mechanism is to keep the scheduler
updated with the CQI value and the utilization of the hybrid automatic repeat request (HARQ) technique with rate
adaptation and soft combination.
•All packets to be transmitted are stored in different queues based on their QoS conditions. For example, NRT packets
are insensitive to delay, but require relatively high throughput. On the other hand, the RT packets are delay-
sensitive. Therefore, the LTE/LTE-A scheduler prioritizes active users and even different traffic types for each user to
meet each service QoS conditions.
•The scheduling algorithm at the eNB performs the scheduling process in both DL and UL directions. For a given
TTI, a different transport block (TB) should be generated for each user to carry its data. Any user to be scheduled in
a given TTI should have full information about its allocated PRBs before sending its packets. All needed information
is transmitted to each scheduled user through the control channel, and the massage is named as scheduling control
information.
•LTE/LTE-A cellular networks allow the support of a wide range of services, each with different QoS requirements.
Packets of different services are stored in different queues in the radio link control (RLC) sub-layer. The scheduling
algorithm should be updated with the status of the buffer every TTI, and accordingly, a scheduling decision should
be taken. Based on the scheduling decision, the information about the allocated PRBs in the DL direction for a spe-
cific user is transmitted to the MAC layer. The MAC layer constructs a specific user TB, which can carry the traffic of
that user.
4of15 ELHADAD ET AL.

As mentioned previously, PRBs are assigned to different active users according to a predefined metric. The metric
can be utilized to understand how a certain scheduler works. For instance, assume that m
u,b
is the metric of a certain
scheduling strategy, and it is represented by the following equation
49
:
mu,b¼max
jmu,b
fg
,ð1Þ
where urepresents a specific user and brepresents the PRB under consideration. So, the considered PRB will be allo-
cated to that user only if the user can provide the maximum metric value as compared with other users.
4|THE PROPOSED OPTIMIZED TWO-LEVEL SCHEDULER
From the previous discussion in Sections 2 and 3, an optimal scheduling technique should have the capability to
achieve the following objectives:
•QoS requirements to different traffic types: QoS satisfaction is a major component of LTE/LTE-A cellular networks for
satisfactory service delivery of evolving Internet applications to the end-users and for managing the network
resources. Today's popular mobile Internet applications, such as voice, gaming, streaming, and social networking ser-
vices, have different traffic characteristics, and consequently, they demand different QoS requirements. The data traf-
fic associated with these services must be delivered to the end-users at specific data rates and/or within the specific
delay, packet loss and delay variation bounds. These requirements can be collectively termed as QoS.
50
The proposed
queue-optimized scheduler guarantees delivery of RT packets before being dropped due to exceeding the delay
threshold. This will reduce the PLR, minimize the packet delay, and hence maximize the throughput of RT
applications.
•System spectral efficiency: System spectral efficiency gives a clear indication of the number of users or applications to
be simultaneously served by any mobile network with respect to the utilization of limited bandwidth. It can be calcu-
lated by dividing the maximum throughput of all users in the network by the available bandwidth.
51
The proposed
queue-optimized scheduling strategy can easily guarantee this objective by considering the CQI, while making sched-
uling decisions.
•Fairness level among active users: Fairness level is an essential factor that should be carefully achieved, while per-
forming the scheduling process. It indicates how different users and different traffic types share the available PRBs in
a fair way.
52
A new fairness perspective is tackled here, which is the average packed drop rate (PDR). It is referred to
as PDR fairness, which can be achieved by making sure that all users have almost the same packet drop rate. The
proposed strategy can guarantee fairness in PRB allocation by considering the throughput delivered to each user until
the allocation time. Furthermore, the number of dropped packets will be equally distributed among all users.
The proposed queue-optimized scheduling strategy performs the scheduling process in two levels, in order to guar-
antee the above-explained objectives.
4.1 |First-level working mechanism
The first level of the proposed scheduler concentrates on the video packets, and it has two main objectives:
1. Adjusting the total RBs allocated to video traffic, such that the total throughput delivered to all video users is equal
to the average value. This leads to optimization of the allocated resources to video traffic, minimization of the wasted
allocated resources, and facilities the support of more video users.
2. Identifying all video packets that will be considered for the allocation process in the next level in every TTI. This
leads to efficiently satisfying the instantaneous QoS requirements of video traffic.
To achieve these objectives, we have an optimization mechanism in order to adjust the number of PRBs that will
be allocated to video applications. An optimized queue threshold is proposed to control the total delivered video
ELHADAD ET AL.5of15

throughput to be around the average video throughput required to satisfy video users' needs. The optimization queue
threshold (Q
Threshold
) is dynamically controlled by the average video throughput required to satisfy video users' needs
every TTI. It is calculated by Equation 2 as follows:
QThreshold ¼extraRB
averageRB








,ð2Þ
averageRB ¼ravg
video
ravg
RB
,ð3Þ
extraRB ¼extraRB þallocatedRB averageRB,ð4Þ
where
k•kis the nearest integer.
average
RB
is the overall number of PRBs needed to satisfy video application requirements in the current TTI, and it
is calculated in Equation 3.
ravg
video is the average video throughput required to satisfy video application requirements in the current TTI.
ravg
RB is the average throughput achieved by one PRB in the current TTI.
extra
RB
is the extra PRBs allocated to the video packets in the last TTI over the average
RB
, and it is calculated by
Equation 4.
allocated
RB
is the total number of PRBs that are previously allocated to the video traffic in the previous TTI.
The working mechanism of the first level is performed every TTI, and it can be summarized in the following steps:
•At the start of each TTI, the proposed Q
Threshold
is computed using Equation 2. Q
Threshold
value is assumed to be zero
when there are no extra allocated RBs (extra
RB
=0). As the extra
RB
value is increased, the value of Q
Threshold
is
increased.
•The instantaneous queue length for each video user is calculated at every TTI, and it is compared to the optimized
queue length threshold, Q
Threshold
.
•Only video users that have a queue length greater than Q
Threshold
are considered for the allocation process in the
instantaneous TTI using the next-level scheduler.
As displayed in Figure 1, the value of Q
Threshold
controls the number of video packets that will be scheduled in a cer-
tain TTI, such that only video flows that have queue length greater than the instantaneous value of Q
Threshold
will be
scheduled in the current TTI. However, video flows that have queue length smaller than the instantaneous value of
Q
Threshold
will be postponed from being investigated in the next TTI.
Regarding Equation 4, as the number PRBs that are actually allocated to video traffic becomes greater than the over-
all number of PRBs needed to satisfy video application requirements in the current TTI (allocated
RB
>average
RB
), the
value of extra
RB
is increased. As a result, the value of Q
Threshold
is increased. As a consequence, the overall number of
video users that are scheduled in the current TTI is decreased. This reduces the number of PRBs that are allocated to
video users in the next TTI. This mechanism always guarantees that the number of overall PRBs that are actually allo-
cated to video traffic at any TTI has a value around the overall number of PRBs needed to satisfy video application
requirements.
4.2 |Second-level working mechanism
The main objectives of the second level of the proposed scheduler are as follows:
1. Efficiently distributing the total RBs allocated to video traffic by the first-level scheduler among video UEs to provide
the required QoS for video flow.
2. Guaranteeing the required QoS level for the RT applications by giving priority to the RT applications (i.e., video and
VOIP applications) that have the lowest packet expiration period.
6of15 ELHADAD ET AL.

3. Delivering the NRT applications with a guaranteed high level of fairness considering the shared spectrum to
enhance the overall network spectral efficiency.
Figure 1 represents the proposed queue-optimized scheduler flowchart. As displayed in the figure, video flow with
queue length greater than Q
Threshold
, VOIP packets, and BE packets are scheduled based on EEDDF metric, which is cal-
culated by Equation 5 as follows:
mEEDDF
u,b¼HutðÞ
EutðÞdb
utðÞ
DutðÞ
lutðÞ
LutðÞ
,ð5Þ
where
H
u
(t) is the HoL packet delay of the uthuser at time t.
E
u
(t) is the expiration time of the uthuser HOL packet. It is the difference between the target delay of any packet
and the HoL delay of that packet at time t.
db
utðÞis the throughput delivered to the uthuser at time tif the bthPRB is allocated to that user.

DutðÞis the average throughput delivered to the uthuser until time t.
l
u
(t) is the PDR of the uthuser until time t.
L
u(t)
is the average number of total packets dropped until time t.
Equation 5 has three main terms, which are HutðÞ
EutðÞ
,db
utðÞ
DutðÞ, and lutðÞ
LutðÞ
. The first term (HutðÞ
EutðÞ
) is the dominant term in case of
delivering RT packets. It gives priority to the packet with the lowest expiration period. This leads to a decrease in the
PLR, because the RT packet is delivered before being dropped due to exceeding the delay threshold. As a result, this
FIGURE 1 The proposed scheduler flowchart
ELHADAD ET AL.7of15

increases the throughput of RT applications. This mechanism guarantees the required QoS level for RT applications.
On the other hand, the second term (db
utðÞ
DutðÞ) is the dominant term in case of delivering NRT applications. This guarantees
a high level of fairness regarding the shared spectrum and enhances the overall network spectral efficiency. The third
term (lutðÞ
LutðÞ
) has a vital role in guaranteeing a high fairness level regarding PDR among active users.
It is worthy to mention that the performance of the proposed queue-optimized scheduler is weakly dependent on
the predefined QoS requirements. Therefore, the proposed queue-optimized scheduler has the ability to support a wide
range of predefined QoS requirements. Furthermore, it has the capability to guarantee high throughput for RT applica-
tions as demonstrated in the following section.
5|PERFORMANCE ANALYSIS AND SIMULATION RESULTS
Full performance evaluation of the proposed queue-optimized scheduler is presented in this section.
5.1 |Simulator setup
A widely known simulator, which is commonly used in much researches,
29–40
is utilized to evaluate the performance of
the proposed queue-optimized scheduler. The simulator is known as an LTE simulator,
53
and it is utilized to compare
the performance of the most recent and efficient schedulers with the proposed queue-optimized scheduler.
A 19-cell–clustered framework is utilized to simulate the interference from neighboring cells. The size of each clus-
ter is composed of four cells, each cell with a radius of 500 m. The user density of each cell ranges from 10 to 60 users
per cell. A total of 10 MHz bandwidth is guaranteed to be shared among active users in DL. Therefore, a total of 50 PRBs
are available to be allocated to different active users every TTI based on the proposed queue-optimized scheduler work-
ing mechanism. Two different user speeds are introduced to simulate the movement of different users. A speed of
3 km/h represents pedestrian users and a speed of 120 km/h represents users travelling by vehicles. User mobility
within a certain cell is simulated using a random direction mobility model. Table 1 summarizes the main parameters of
the considered simulator.
53
TABLE 1 The simulation
parameters
29–53
Parameter Value
Simulation time 100 s
LTE band 2 GHz
Bandwidth for DL 10 MHz
Duplex mode FDD mode
Users density per cell 10, 20, …, 60 UEs
Delay threshold of RT packets 100 ms
Transmit power of each eNB 43 dBm
Modulation schemes QPSK, 16-QAM, and 64-QAM
Propagation model Macro-cell urban model
Video packets data rate 128 kbps
eNB antenna configuration 1 Tx, 1 Rx (1 1)
UL delay 3TTI
VoIP streams G.729
Target BLER 10%
Thermal noise density 174 dBm/Hz
Video packets compression scheme H.264
Users speed 120 and 3 km/h
BE streams Infinite buffer source
8of15 ELHADAD ET AL.

The eNBs have an antenna configuration of one antenna for transmission and one antenna for reception. A fre-
quency division duplex (FDD) mode is utilized in this framework. The frame duration is 10 ms. Each frame has 10 sub-
frames, each with 1 ms length. Two time slots (TS) are included in one subframe. The duration of each TS is 0.5 ms.
Normal cyclic prefix (CP) is assumed in this structure. Any active user in the cell receives one video stream, one VoIP
stream, and one best-effort stream at the same time.
5.2 |Performance evaluation
The performance of the proposed queue-optimized scheduler is evaluated regarding the performance of the most
efficient and recent scheduling schemes. I-MLWDF, FLS, EXP-rule, and LOG-rule scheduling algorithms are consid-
ered for performance comparisons. Different key performance indicators (KPIs) such as packet delay, PLR, throughput,
fairness, and network spectral efficiency are utilized to show the competency of the proposed queue-optimized
scheduler. The analysis is conducted with the utilization of different parameters such as changing the number of users
per cell, user speed, and the delay threshold of RT packets.
Starting by analyzing the performance regarding the video packets, Figures 2 and 3 show the video packet PLR as a
function of the number of users per cell at a target delay of 0.1 s and at two different user speeds: 3 and 120 km/h,
respectively. The queue-optimized scheduler achieves the lowest PLR at two values of user speed. Furthermore, as the
user speed increases, the PLR value is also increased. This is a normal result, due to the rapid changes in the channel
FIGURE 3 Video packet PLR as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 120 km/h
FIGURE 2 Video packet PLR as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
ELHADAD ET AL.9of15

quality. As a result, the error rate of selecting a convenient MCS is increased. Hence, the video packet PLR is increased.
Moreover, the video packet PLR is increased as the number of users per cell is increased. This is also a normal result,
due to the increased network load. As a result, the ability of the network to satisfy all user needs is reduced, leading to
an increase in the video packet PLR.
From the two figures, it is easy to know how many users would be supported by the network at a specific PLR. From
Figure 2, for a video packet PLR of 0.001%, and when using the LOG-rule scheduler, the total number of users served in
any cell is less than 30 users. On the other hand, for a video packet PLR of 0.001%, and when using queue-optimized
scheduler, the total number of users served in any cell is increased to be almost 60 users. This means that the queue-
optimized scheduler has the ability to maximize the network capacity as compared by other considered techniques.
Similarly, from Figure 3, the network can only support 25 users per cell, when using the LOG-rule scheduler for a
video packet PLR of 0.01%. On the other hand, by using the queue-optimized scheduler, the total number of users
served in any cell is increased to 50 users for the same PLR value.
Figure 4 presents the video traffic throughput as a function of the number of UEs per cell at a target delay of 0.1 s
and user speed equal to 3 km/h. It is clear from these results that the video traffic throughput is increased as a result of
the increase in the number of UEs per cell. The increase of the video traffic throughput continue to increase, when the
available PRBs are not sufficient to satisfy the needs of all users. It is worthy to mention that the queue-optimized
scheduler achieves the highest video traffic throughput compared with the other considered schedulers.
The queue-optimized scheduler achieves the lowest video packet delay compared with other schedulers. This is
clearly displayed in Figure 5. Generally, any scheduling technique that has the ability to support RT applications drops
FIGURE 5 Video packet delay as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
FIGURE 4 Video traffic throughput as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
10 of 15 ELHADAD ET AL.

any RT packet, which is buffered for a period greater than its target delay. Because the scheduler assumes that it is
meaningless to deliver any RT packet after exceeding its delay threshold. The queue-optimized scheduler has a clear
task, which is to deliver any packet before the violation of its due date.
Figure 6 shows the video traffic fairness index as a function of the number of UEs per cell at a target delay of 0.1 s
and user speed of 3 km/h. The highest fairness level is achieved, when using the queue-optimized scheduler. The VOIP
packet PLR as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h is displayed
in Figure 7. The VOIP packet PLR is increased as the number of UEs per cell is increased. This is a normal result, due
to the increased network load. As a result, the ability of the network to satisfy all user needs is reduced, leading to an
increase in the VOIP packet PLR. It is worthy to mention that queue-optimized scheduler achieves the lowest VOIP
packet PLR compared with other techniques. This is because the main target of the introduced scheduler is to give pri-
ority for RT packets and transmit any RT packet before its due date. This mechanism minimizes the number of VOIP
packets dropped due to exceeding their delay threshold. Furthermore, due to the low source bit rate of VOIP applica-
tions, VOIP packets will have priority to be scheduled before any other packets. This is because the second term in
Equation 5 will be the dominant term and it gives the VOIP packets the first priority regarding their smaller source bit
rate. Consequently, the VOIP packets will have lower PLR than that of the video packets.
The VOIP packet delay as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of
3 km/h is displayed in Figure 8. It is easy to notice that the queue-optimized scheduler achieves the smallest VOIP
packet delay compared with the other scheduling techniques.
FIGURE 6 Video traffic fairness index as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
FIGURE 7 VOIP packet PLR as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
ELHADAD ET AL.11 of 15

For the BE packets, the main KPIs will be the achieved throughput and the fairness level among active users regard-
ing the shared spectrum. Figure 9 displays the BE traffic throughput as a function of the number of UEs per cell at a tar-
get delay of 0.1 s and user speed of 3 km/h. From the figure, it is easy to notice that by considering both LOG-rule and
EXP-rule scheduling techniques, the network achieves higher BE traffic throughput compared to that of the case when
FIGURE 8 VOIP packet delay as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
FIGURE 9 BE traffic throughput as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
FIGURE 10 Cell spectral efficiency as a function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/h
12 of 15 ELHADAD ET AL.

considering both FLS and queue-optimized scheduling techniques. This is due to the bad performance of both LOG-rule
and EXP-rule scheduling techniques with the RT packets. As a result, this bad performance leaves a large number of
available PRBs to be used to satisfy the BE user needs, and hence achieve high BE traffic throughput as compared with
those achieved when considering both FLS and queue-optimized scheduling techniques. Finally, as a normal result, the
fairness level will follow the same behavior as the achieved BE traffic throughput.
A vital KPI considered in the performance evaluation is the cell spectral efficiency. The cell spectral efficiency as a
function of the number of UEs per cell at a target delay of 0.1 s and user speed of 3 km/H is displayed in Figure 10. It is
obvious that as the number of users per cell is increased, the spectral efficiency of the cell is increased. The increase of
the cell spectral efficiency continues to a specific value, when the available PRBs are not sufficient to satisfy the needs
of all users in the cell, and then it is saturated to its maximum value. Finally, the queue-optimized scheduling strategy
significantly enhances the cell spectral efficiency compared with the other scheduling techniques, especially as the
number of UEs per cell is increased.
6|CONCLUSION AND FUTURE WORK
In this work, a novel queue-optimized two-level scheduling strategy was introduced. The queue-optimized scheduler
has the ability to support both RT and NRT services for LTE downlink framework. An optimized queue length
threshold is dynamically adapted around the average bandwidth required for the video traffic. The proposed scheduling
strategy is utilized to efficiently enhance the radio resource allocation process for video traffic. The allocation process is
actually performed at the second level of the scheduling process. The PRBs are assigned according to the EEDDF
mechanism. The proposed queue-optimized scheduling strategy prioritizes users according to their traffic types, to
enhance QoS requirements. The performance of the queue-optimized scheduler is evaluated regarding the performance
of the most efficient and recent scheduling schemes. I-MLWDF, FLS, EXP-rule, and LOG-rule scheduling algorithms
are considered for performance comparisons. Simulation results confirmed the ability of the queue-optimized schedul-
ing strategy to achieve the lowest PLR and packet delay for RT applications and to maximize the achieved throughput,
especially for video applications. Furthermore, the queue-optimized scheduling strategy maximizes cell spectral
efficiency, which is significantly increased as the number of users per cell is increased. In future work, we will consider
more advanced modulation schemes for the proposed scheduling technique. Also, we will concentrate on the adapta-
tion of the proposed scheduler with the new trends of communication networks.
ACKNOWLEDGEMENT
The authors would like to acknowledge the support received from Taif University Researchers Supporting Project Num-
ber (TURSP-2020/147), Taif university, Taif, Saudi Arabia.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
ORCID
Walid El-Shafai https://orcid.org/0000-0001-7509-2120
Fathi E. Abd El-Samie https://orcid.org/0000-0003-4117-3496
REFERENCES
1. Yang Y, Chen Y, Wang W, Yang G. Securing channel state information in multiuser MIMO with limited feedback. IEEE Trans Wirel
Commun. 2020;19(5):3091-3103.
2. Arul R, Raja G, Almagrabi AO, Alkatheiri MS, Chauhdary SH, Bashir AK. A quantum-safe key hierarchy and dynamic security associa-
tion for LTE/SAE in 5G scenario. IEEE Trans Ind Info. 2019;16(1):681-690.
3. Mukhopadhyay A, Das G. Low complexity fair scheduling in LTE/LTE-A uplink involving multiple traffic classes. IEEE Syst J. 2020;
1-12.
4. El-Shafai W, El-Rabaie S, El-Halawany MM, Abd El-Samie FE. Security of 3D-HEVC transmission based on fusion and watermarking
techniques. Multimed Tools Appl. 2019;78(19):27211-27244.
5. Fu Y, Chen H, Zheng Q, et al. An adaptive security data collection and composition recognition method for security measurement over
LTE/LTE-A networks. J Network Comput Appl. 2020;155:1-14, 102549.
ELHADAD ET AL.13 of 15

6. Nashaat H, Refaat O, Zaki FW, Shaalan IE. Dragonfly-based joint delay/energy LTE downlink scheduling algorithm. IEEE Access. 2020;
8:35392-35402.
7. El-Shafai W, El-Rabaie S, El-Halawany MM, Abd El-Samie FE. Encoder-independent decoder-dependent depth-assisted error conceal-
ment algorithm for wireless 3D video communication. Multimed Tools Appl. 2018;77(11):13145-13172.
8. El-Shafai W, El-Rabaie S, El-Halawany MM, Abd El-Samie FE. Efficient hybrid watermarking schemes for robust and secure 3D-MVC
communication. Int J Commun System. 2018;31(4):1-23, e3478.
9. Dahmani S, Gabli M, Mermri EB, Serghini A. Optimization of green RNP problem for LTE networks using possibility theory. Neural
Comput Applic. 2020;32(8):3825-3838.
10. Yi C, Cai J, Su Z. A multi-user mobile computation offloading and transmission scheduling mechanism for delay-sensitive applications.
IEEE Trans Mobile Comput. 2019;19(1):29-43.
11. El-Shafai W. Pixel-level matching based multi-hypothesis error concealment modes for wireless 3D H. 264/MVC communication. 3D
Res. 2015;6(3):1-11, 31.
12. He H, Shan H, Huang A, Lin CAI, Tony X, Quek QS. Proportional fairness-based resource allocation for LTE-U coexisting with Wi-Fi
IEEE access. IEEE Access. 2017;5:4720-4731.
13. Yildiz Ö, Sokullu R. A novel mobility aware downlink scheduling algorithm for LTE-A networks. Ubiq Future Netw (ICUFN). 2017;
300-305.
14. Hsu, Chan-Ching, and Morris Chang, J. (2017). Spectrum-energy efficiency optimization for downlink LTE-A for heterogeneous net-
works. IEEE Trans Mobile Comput, 16(5), 1449-1461.
15. Palit B, Das SS, Kamavaram Y. Multiple QoS provisioning with pre-emptive priority schedulers in multi-resource OFDMA networks.
Wirel Networks. 2020;1-20.
16. Nardini G, Stea G, Virdis A, Sabella D, Caretti M. Resource allocation for network-controlled device-to-device communications in LTE
Advanced. Wirel Netw. 2017;23(3):787-804.
17. Al-Afandy KA, El-Shafai W, El-Rabaie ESM, et al. Robust hybrid watermarking techniques for different color imaging systems. Mul-
timed Tools Appl. 2018;77(19):25709-25759.
18. Ma T, Zhang Y, Wang F, Wang D, Guo D. Slicing resource allocation for eMBB and URLLC in 5G RAN. Wirel Commun Mobile Comput.
2020;2020:1-11.
19. Guo Y, Yang Q, Liu J, Kwak KS. Cross-layer rate control and resource allocation in spectrum-sharing OFDMA small-cell networks with
delay constraints. IEEE T Veh Technol. 2017;66(5):4133-4147.
20. Karachontzitis S, Dagiuklas T, Dounis L. Fair cross-layer scheme for heterogeneous H.264/AVC video streams over LTE-based broad-
band systems 2012 8th International Wireless Communications and Mobile Computing Conference (IWCMC), 2012. https://doi.org/10.
1109/IWCMC.2012.6314343
21. Lee YL, Loo J, Chuah TC, Ayman El-Saleh A. Multi-objective resource allocation for LTE/LTE-A femtocell/HeNB networks using ant
colony optimization. Wirel Personal Commun. 2017;92(2):565-586.
22. Elsayed K, Khattab A. Channel-aware earliest deadline due fair scheduling for wireless multimedia networks. Springer Wireless Pers
Commun. 2006;38(2):233-252.
23. Satheesh B, Sivasubramanian A, SPK, B. A review of MAC scheduling algorithms in LTE system. Int J Adv Sci Eng Inf Technol. 2017;7
(3):1056-1068.
24. Farhana A, Roshanak H, Maruf S, Kumbesan S, Sharmin S. Comparative analysis of downlink packet scheduling algorithms in 3GPP
LTE networks. Int J Wirel Mobile Networks (IJWMN). 2015;7(5):1-21.
25. Jabbar A, Abdullah F. Long term evolution (LTE) scheduling algorithms in wireless sensor networks (WSN). Int J Computer Appl. 2015;
121(10):12-16.
26. Grigol B, Sergi G. Customized packet scheduling algorithm for LTE network. Am Sci Res J Eng Technol Sci (ASRJETS). 2018;40(1):
203-216.
27. Bamidele K. Performance evaluation of scheduling algorithms for 4G (LTE). Commun Network. 2018;10:152-163.
28. Abd-Elnaby M, Elhadad M, El-Rabaie E. Capacity enhancement based on dynamically adapted PF scheduling algorithm for LTE down-
link system. Int Arab J Inf Technol. 2018;15(5):795-801.
29. Elhadad M, Abd-Elnaby M, El-Rabaie E. Optimized delay threshold scheduler for multimedia traffic over LTE downlink network.
Springer, Multimedia Tools Appl. 2018;67(3):1-19.
30. Elhadad MI, El-Shafai W, El-Rabaie ESM, Abd-Elnaby M, Abd El-Samie FE. Enhanced fair earliest due date first scheduling strategy for
multimedia applications in LTE downlink framework. Int J Commun System. 2020;33(6):1-17, e4190.
31. Lawal M, Saidu I, Mohammed A, Sade Y. Downlink scheduling algorithms in LTE networks: a survey. IOSR J Mobile Comput Appl
(IOSR-JMCA). 2017;4(3):1-12.
32. Angri A, Najid A, Mahfoudi M. New combined downlink scheduling algorithm for LTE networks. Int J Commun Antenn Propag. 2017;7
(1):8-17.
33. Angri I, Mahfoudi M, Najid A, EL Bekkali M. Exponential MLWDF (EXP-MLWDF) downlink scheduling algorithm evaluated in LTE
for high mobility and dense area scenario. Int J Electr Comput Eng (IJECE). 2018;8(3):1618-1628.
34. Nogueira, M., Marini, S., Both, C., Rochol, J., Armando, O., Oscar, C., & Luizelli, M. (2016). QoS aware schedulers for multi-users on
OFDMA downlink: optimal and heuristic. 8th IEEE Latin-American Conference on Communications (LATINCOM), Medellin, Colom-
bia, 15-17 Nov. 2016.
14 of 15 ELHADAD ET AL.

35. Capozzi F, Piro G, Grieco LA, Boggia G, Camarda P. Downlink packet scheduling in LTE cellular networks: key design issues and a sur-
vey. IEE Commun Surveys Tutorials. 2013;15(2):678-700.
36. Christantus ON, Chidera LA, Cosmas IA. Improved MLWDF scheduler for LTE downlink transmission. Int J Electr. 2016;103(11):
1857-1867.
37. Sadiq B, Madan R, Sampath A. Downlink scheduling for multiclass traffic in LTE. EURASIP J Wirel Commun Netw. 2009;2009:1-18.
38. Tuan LM, Son LH, Long HV, et al. ITFDS: channel-aware integrated time and frequency-based downlink LTE scheduling in MANET.
Sensors. 2020;20(12):1-20.
39. Ahmed H, Jagannathan K, Bhashyam S. Queue-aware optimal resource allocation for the LTE downlink with best Msubband feedback.
IEEE Trans Wirel Commun. 2015;14(9):4923-4933.
40. Thienthong, P., Teerasuttakorn, N., Nuanyai, K., & Chantaraskul, S. (2019, March). Comparative study of scheduling algorithms and
almost blank subframe for LTE HetNets. In 2019 7th International Electrical Engineering Congress (iEECON) (pp. 1-4). IEEE.
41. Kaleem Z, Hui B, Chang K. QoS priority-based dynamic frequency band allocation algorithm for load balancing and interference avoid-
ance in 3GPP LTE HetNet. EURASIP Journal on Wireless Communications and Networking. 2014;2014(1):1-18.
42. Kaleem Z, Chang K. QoS priority-based coordinated scheduling and hybrid spectrum access for femtocells in dense cooperative 5G cellu-
lar networks. Trans Emerg Telecommun Technol. 2018;29(1):1-17, e3207.
43. Nasralla M, Martini G. A downlink scheduling approach for balancing QoS in LTE wireless networks. IEEE Int Symp Pers Indoor Mob
Radio Commun PIMRC. 2013;1571-1575.
44. Wang Y, Hsieh S. Service-differentiated downlink flow scheduling to support QoS in long term evolution. Comput Networks. 2016;94:
344-359.
45. Ameigeiras P, Navarro-Ortiz J, Andres-Maldonado P, et al. 3GPP QoS-based scheduling framework for LTE. EURASIP J Wirel Commun
Netw. 2016;78(1):1-14.
46. Hendaoui, S., Nawel, Z., & Tabbane, S. (2016). Highway and rate prediction downlink scheduler (HRPDS) for LTE-A real time applica-
tions, in 13th IEEE Annual Consumer Communications & Networking Conference (CCNC).
47. Liu S, Zhang C, Zhou Y, Zhang Y. Delay-based weighted proportional fair algorithm for LTE downlink packet scheduling. Wirel Pers
Commun. 2015;82(3):1955-1965.
48. Rebekka B, Sudheep S, Malarkodi B. An optimal and priority based rate guaranteed radio resource allocation scheme for LTE downlink.
Wirel Pers Commun. 2015;83(3):1643-1661.
49. Ghassan, S., Ali, A., Baba, D., & Mansor, M. (2015). RB preserver downlink scheduling algorithm for real-time multimedia services in
LTE networks, in Computer Application and Industrial Electronics (ISCAIE), 21-26.
50. Skondras, E., Michalas, A., Sgora, A., & Vergados, D. (2015). A downlink scheduler supporting real time services in LTE cellular net-
works, in Information, Intelligence, Systems and Application (IISA), 1-6.
51. Sulthana S, Nakkeeran R. Performance analysis of service based scheduler in LTE OFDMA system. Wirel Pers Commun. 2015;83(2):
841-854.
52. Abu-ali N, Taha A, Member S, Salah M, Hassanein H. Uplink scheduling in LTE and LTE-advanced: tutorial, survey and evaluation
framework. IEEE Commun Surv Tutor. 2014;16(3):1239-1265.
53. Piro G, Grieco L, Boggia G, Capozzi F, Camarda P. Simulating LTE cellular systems: an open-source framework. IEEE Trans Veh
Technol. 2011;60(2):498-513.
How to cite this article: Elhadad MI, El-Shafai W, El-Rabaie E-SM, Abd-Elnaby M, Abd El-Samie FE.
Optimized two-level scheduler for video traffic in LTE downlink framework. Int J Commun Syst. 2021;e4704.
https://doi.org/10.1002/dac.4704
ELHADAD ET AL.15 of 15