2015 International Conference on Computational Intelligence & Networks

A Comparative Analysis of Packet Scheduling

Schemes for Multimedia Services in LTE Networks
B. P. S. Sahoo∗ , Deepak Puthal† , Satyabrata Swain‡ and Sambit Mishra‡
∗ Department of Computer Science and Information Engineering, National Central University, Taiwan (R. O. C)
† Department of Computer Science and Engineering, National Institute of Technology Silchar, India
‡ Department of Computer Science and Engineering, National Institute of Technology Rourkela, India
Email: {biswap.singh, deepak.puthal, satya.swain10, skmishra.nitrkl}@gmail.com

Abstract—The revolution in high-speed broadband network stand and appreciate the benefits of LTE high data rates and
is the requirement of the current time, in other words here services.
is an unceasing demand for high data rate and mobility. Both As one of the primary objective of the LTE network is
provider and customer see, the long time evolution (LTE) could
be the promising technology for providing broadband, mobile to enhance the data-rate so as to cater the range of highly
Internet access. To provide better quality of service (QoS) to demanded services; the radio resources or data channels are di-
customers, the resources must be utilized at its fullest impeccable vided and shared efficiently among different active users while
way. Resource scheduling is one of the important functions for considering a satisfied level of QoS to all active users. To shape
remanufacturing or upgrading system performance. This paper the requirements, the LTE system uses orthogonal frequency
studies the recently proposed packet scheduling schemes for
LTE systems. The study has been concentrated in implication division multiple access (OFDMA) technology in the down-
to real-time services such as online video streaming and Voice link. The OFDMA technology divides the available bandwidth
over Internet Protocol (VOIP). For performance study, the LTE- into multiple narrow-band sub-carriers and allocates a group
Sim simulator is used. The primary objective of this paper is of sub-carriers to a user based on its requirements, current
to provide results that will help researchers to design more system load and system configuration [1]. On that ground,
efficient scheduling schemes, aiming to get better overall system
performance. For the simulation study, two scenarios, one for collisions can happen more often among the active users while
video traffic and other for VoIP have been created. Various per- sharing the same data channel. Hence, it needs to take care
formances metric such as packet loss, fairness, end-to-end (E2E) by the resource allocation scheme. With 3G technologies,
delay, cell throughput and spectral efficiency has been measured simple video services were attempted with attention focused
for both the scenarios varying numbers of users. In the light of on video telephony and multimedia messaging, and were
the simulation result analysis, the frame level scheduler (FLS)
algorithms outperform other, by balancing the QoS requirements somewhat successful. However, these technologies are still
for multimedia services. capacity limited and not able to economically support the
huge mobile video demand that is emerging. This directly
Keywords-Long term evolution (LTE), cellular networks,
packet scheduling schemes, quality-of-service (QoS), multimedia
leads to the desire to use a broadband wireless network, such
streaming. as LTE, for doing the same stuffs with greater speed and
mobility. This can be mostly achievable by efficient resource
allocation technique. Hence, the fast packet scheduling is one
I. I NTRODUCTION
of the important data-rate improvement techniques among
In the growing complexity of today’s wireless communica- Link Adaptation, Multiple-Input and Multiple-Output (MIMO)
tion systems; the cellular data networks are experiencing an Beam forming and Hybrid ARQ in 4G networks [12].
increasing demand for its high data rate and wide mobility. In this paper, we have studied packet scheduling algorithms
In 2008, the Long Term Evolution (LTE) was introduced by that are proposed in the past year for LTE cellular networks
the 3GPP1 (3rd Generation Partnership Project) in order to in implication to multimedia services. A comparison of per-
increase the capacity and speed of wireless data networks. formance indexes such as average system throughput, end-to-
LTE and WiMAX are the two promising standards in the delay (E2E), packet-loss-ratio (PLR) and fairness is reported
current cellular technologies. These two are also marketed for LTE systems over a realistic simulated scenario in multi-
as 4th Generation (4G) wireless services. LTE offers several cell environment. The main goal this study is to provide results
important benefits for the subscribers as well as to the service that will help in the design process of LTE cellular networks,
provider. It significantly satisfies the users requirement by aiming to get better overall performance.
targeting the broadband mobile applications with enhanced The rest of the paper is organized as follows. An overview
mobility. With the introduction of Smartphone, the application LTE of packet scheduler design and challenges are presented in
like HDTV, online gaming, video meetings, etc are certainly Section II. Section III describes the recently proposed packet
become more valuable to the users. Hence, the users under- scheduling algorithms for LTE system. Section IV describes
simulation scenario and examines the reported results. Finally,
1 http://www.3gpp.org Section V concludes the paper.

2375-5822/15 $31.00 © 2015 IEEE 108

110
DOI 10.1109/CINE.2015.30
 maximizing the QoS while maintaining the fairness for these
services. Different scheduling schemes have been proposed to
support real-time (RT) and non real-time (NRT) services. In
this section, we will illustrate the working principle of different
packet scheduling algorithms that are suitable to use in LTE
Systems.
The proportional fair (PF) [3] algorithm provides a good
tradeoff between system throughput and fairness by selecting
the user with highest instantaneous data rate relative to its
average data rate. Its main objective is to provide maximum
throughput while maintaining the fairness. To provide a maxi-
mum throughout, the algorithm attempts to serve each user at
the peak of its channel condition, as a result, the scheduler will
see a drop in channel condition temporarily for tc seconds. A
higher value of tc allows the scheduler to wait longer for a
users channel condition and hence improves overall average
system throughput.
Fig. 1. Simplified model of a packet scheduler [1]
Suppose there are n users, let Ri (t) be the estimated average
rate for user i at slot t, and the current requested rate for
user i is di (t). The algorithm selects a user using following
II. OVERVIEW ON S CHEDULING IN LTE S YSTEMS
computation as:
The mobile broadband is a reality today and growing in very
fast manner. 4G wireless systems such as 3GPP LTE features di (t)
Pi [n] = (1)
high data rate and low end-to-end latency which are the key Ri (t)
requirements of multimedia applications, especially video and The user with the highest Pi [n] out of all n users will receive
VOIP. the transmission at each decision time.
In order to support the rich demand of real-time multimedia The maximum-largest weighted delay first (M-LWDF) [4]
services, it is necessary to ensure the QoS requirements are was originally designed to support multiple real-time data
met and PLR is minimised by keeping it below the applications users in CDMA-HDR systems. The key feature of this algo-
required threshold. Hence, multi-user scheduling is one of rithm is that the scheduling decision depends on both current
the main feature in LTE systems because it is in charge of channel conditions and the states of the queues. The M-LWDF
distributing available resources among active users in order to serves in every TTI to the user with the highest priority. The
satisfy their QoS needs. In a video streaming service environ- priority of user i is computed as [9]:
ment, it is important to maintain the PLR threshold below 1%
[2] such that the QoS requirements of video streaming service Ri [n] Di [n]
Pi [n] = −log(δi ) . . (2)
users are satisfied. λi [n] Ti
A simplified model of, how the main radio resource man- where Pi [n] denotes the priority of user i at the TTI number
agement (RRM) modules interact with the downlink packet n, Di the head of line (HOL) packet delay of user i, Ri the
scheduler is presented in Figure 1. Packet scheduling is one per-TTI maximum supportable data rate, Ti expresses the
of the RRM functions and it is responsible for intelligent discard timer parameter for the user i, λi is an estimation of
selections of users and transmissions of their packets such that the average user throughput, and δi denotes a QoS parameter
the radio resources are efficiently utilized and the users QoS that allows to differentiate between users with different QoS
requirements are satisfied [11]. The whole process shown in priorities. The HOL packet delay Di [n] is defined as queuing
the figure is repeated, in each transmission time interval (TTI) delay dij of the packet pij that is located at the front of the
duration. The detailed procedure of, how the packet scheduler queue Qi at the TTI n.
determines which users are to be scheduled is presented in [1]. The above discussed PF [3] scheduler is a suitable scheme
The above outlined working flow slightly differs in the for NRT traffic where as Exponential Proportional Fairness
uplink direction as the eNodeB2 does not need any additional (EXP/PF) [5] was developed to support real-time multimedia
information on the uplink channel quality. applications. This algorithm has been designed to increase the
priority of real time flows with respect to non-real time ones.
III. S CHEDULING S TRATEGIES FOR LTE S YSTEMS
The resources are allocated to users based on the following
The desire QoS for multimedia services is of growing im- metric [11]:
portance for research and study LTE is the major step towards
the 4th generation of cellular networks. Hence, the packet ⎧
⎨exp( ai Wi (t)−a
√ i W (t) ) ∗ ( Rri (t) ), i ∈ RT
scheduling schemes to be used in LTE network should aim at i (t)
Pi [n] = 1+ aW (t) (3)
⎩ w(t) r i (t)
, i ∈ NRT
2 The term ’eNodeB’ is used to refer to base stations in 3GPP LTE. M (t) Ri (t)

111
109
and: The Optimized Logarithmic rule (LOG-Rule) [8] scheduler
was designed to provide a balance in QoS metrics in terms
1  of mean delay and robustness. It allocates service to a user
aw(t) = ai Wi (t) (4)
NRT in the same manner as EXP rule to maximize the current
i∈RT
 system throughput, assuming the traffic arrival and channel
w(t − 1) − ε, Wmax > τmax statistics are known. The author claims that the LOG rule is for
w(t) = ε (5)
w(t − 1) + k, Wmax < τmax a practical solution but, is not provably mean-delay optimal.
For the LOG rule, the preference function is calculated as:
Where M (t) is the average number of packets at the
eNodeB’s buffer at time t, k and ε are constants, Wmax is mi,k i
LOG−Rule = bi .log(c + αi DHOL,i )Γk (8)
the HOL, the packet delay of RT services, and τmax is the where bi , c, and ai are tuneable parameters; Γik represents the
maximum delay of RT service users. The EXP/PF algorithm spectral efficiency for the ith user on the kth resource block.
prioritizes RT traffic users over the NRT traffic users when A range of downlink packet scheduling algorithms are
their HOL delays are approaching the delayed deadline. classified according to their design, principles and target
The frame level scheduler (FLS) [6] is a two-level schedul- achievements are presented in Table I.
ing scheme with one upper level and lower level. Two different To demonstrate the effectiveness of the above-studied
algorithms are implemented in these two levels. A low com- schemes while considering several network conditions, sce-
plexity resource allocation algorithm based on discrete time narios with real-time multimedia flows are being analyzing on
linear control theory is implemented in the upper level. It the impact of PLR, throughput, delay and fairness. Results
computes the amount of data that each real-time source should will emphasize the effectiveness of these schemes describing
transmit within a single frame, to satisfy its delay constraint. how it is able to fully respect QoS requirements of multimedia
The PF [3] algorithm is implemented in the lower level to flows with respect to other ones.
assign radio resources to the user, to ensure a good level
of fairness among multimedia flows. The following equation IV. S IMULATION F RAMEWORK AND R ESULTS
calculates the amount of data to be transmitted. In this section, we will discuss the simulation setup which is
used in this paper to evaluate different scheduling schemes in
ui (k) = hi (k) ∗ qi (k) (6)
multimedia traffic scenarios over LTE networks. The reported
where ui (k) corresponds to the amount of data that is trans- results have been studied and presented a comparison between
mitted during the kth frame; ’*’ operator is the discrete time video flows and VoIP flows. To compare the performance of
convolution. The above equation tells that the amount of data the above-discussed scheduling algorithms, the LTE-Sim [10],
to be transmitted by the ith flow during the kth LTE frame an open source simulator for LTE networks was used.
is obtained by filtering the signal qi (k) (i.e., the queue level) A. Simulation Scenario
through a time-invariant linear filter with pulse response hi (k).
The Optimized EXP rule [7] was proposed to provide QoS In real world scenarios, performance evaluation of a well-
guarantees over a shared wireless link. The author claims designed network model and the model itself carries signifi-
that the proposed rule is throughput-optimal. This scheduling cant importance. To perform the test, we defined a multicell
scheme explicitly uses the information on the state of the scenario with a fixed eNodeB at the centre of the cell. The
channel and queues and it ensures the queues stability without users are moving in a vehicular propagation environment in
any prior knowledge of arrival and channel statistics of traffic. a bounded region of radius equal to 1 km with a speed
of 60-120 kmph. Mobility of each mobile station or user
This scheme defined two exponential rules for service a
equipment (UE) travelling cells is described with the random
queue: the EXP (Queue length) rule (EXP-Q) and the EXP
direction model. The number of connections to an eNodeB at
(Waiting time) rule (EXP-W). It chooses either EXP-Q or
a particular time varies from 20-200 UEs. Each UE receives
EXP-W rule to serve a queue with a fixed set of positive
one video flow and one VoIP flow at the same time. Table
parameters such as: β, η ∈ (0, 1) and γi , ai , i ∈ N
II gives the details simulation parameters used in the created
simulation scenario for this paper.
ai Qi (t)
i ∈ i(S(t)) =arg max γi μi (t) exp( β+[Q(t)]η )
i (7)
ai Wi (t) B. Traffic Model
i ∈ i(S(t)) =arg max γi μi (t) exp( β+[W (t)]η
)
i
The video and VOIP streaming is treated as a real-time
where μi (t) is the number of  user served from the multimedia service. The real time video streaming traffic
queue at time t, Q(t) = ( N1 ) iai Qi (t), and W (t) = model can be described as continuous occurrence of video
( N1 ) iai Wi (t). frames in a specific time interval. Video traces can be em-
To provide a balance between mean-delay and throughput, ployed to simulate video traffic in a wide range of network
the EXP rule maintains a lower delay by compromising the simulations [18]. Video traces are used to generate video traffic
throughput, and eventually the mean delays and tails for almost workloads as well as to estimate the video related performance
all users. metrics, such as the frame starvation probability. In the studies

112
110
 TABLE I
C LASSIFICATION OF S CHEDULING S CHEMES IN LTE

Scheduling Strategy Scheduling Type Service Target Service Type

PF [3] Opportunistic Based Maximize throughput while maintaining fairness Non Real Time
M-LWDF [4] Opportunistic and Delay Based Maximize throughput by ensuring delay Real Time
EXP-PF [5] Opportunistic Based Maximize throughput by ensuring delay Real Time
FLS [6] Fairness Based Ensure delay and a good level of fairness Real Time
EXP-Rule [7] Throughput Based Maintains lower delay compromising throughput Real Time
LOG-Rule [8] Throughput Based Maximize throughput and balanced QoS Real Time
Round-Robin (RR) [15] Fairness Based Maximize fairness (but throughput degrades) Both
Max-Rate [16] Throughput Based Maximized throughput (but results in low fairness) Both

TABLE II
S IMULATION PARAMETERS
outperforms all other schemes, meeting the target packet loss
Parameters Values
Simulation Duration 100 Sec of around 2% by supporting 120 UEs. It is worth noting that
Frame Structure FDD VoIP flows experience significantly smaller packet loss than
Radius 1 KM video services.
Bandwidth 10 MHz
Slot Duration 0.5 ms
2) Fairness: Fairness is a major requirement that should be
Scheduling Time (TTI) 1 ms taken into account to guarantee a minimum performance to the
Sub-career Spacing 15 kHz cell-edge users. Figs. 2(b) and 3(b) show the fairness achieved
Maximum Delay 0.1 ms for video and VoIP flows, respectively. It is possible to observe
Video bit-rate 242 kbps
User Speed 60-120 KM that the fairness decreases with increasing number of UEs, due
Traffic Model H.264, Random Direction to the higher network load. As the PF algorithm provides a
good tradeoff between system throughput and fairness; when
compare to all other schemes, the PF scheme still performs
a poor. It is possibly due to its NRT oriented design. The
reported here, in order to obtain a realistic simulation of an FLS and EXP-Rule scheme performs significantly well when
H.264 SVC video streaming, we have used an encoded video compared to all other schemes in lower number of UEs and
sequence “foreman.yuv3 “, which is publicly available. The gradually produce equal results in higher number UEs but FLS
encoded spatial resolution CIF 352x288 with 300 frames per remains in the higher performing scheme for both scenario
second has been used for the entire simulation. because the PF algorithm in lower level ensures a a good
The design and performance analysis of VoIP QoS tech- level of fairness among multimedia flows. Hence, it can be
niques requires adequate traffic models. For VoIP streaming, concluded that FLS is the best candidate for both flows in
this paper simulates a multiplexed packet-level VoIP traffic that considering the fairness, the major performance metric.
lead to models for the packet arrival process of the inbound 3) End-to-End Delay: The End-to-End delay is the time
IP traffic from an IP gateway. The resulting traffic is a good required for a packet to be traversed from source to destination
approximation of the live traffic for the purpose of QoS traffic in the network and is measured in seconds. For quality voice
engineering. and video service application such as Skype and Google+
Hangout, the packet E2E delay should not exceed 150 ms
C. Performance Metrics to evaluate that the quality of the created VoIP calls is
The real-time multimedia performance under the above- accepted [13]. The result obtained from the video and VOIP
discussed schemes is evaluated with below-listed metrics. All quality assessment for delay variations are plotted in Fig. 2(c)
the metrics are related to QoS requirements except the cell and 3(c) respectively. As the PF [3] scheme was originally
throughput, which is a network related performance metric. designed for NRT applications, the reflected result is beyond
Detailed descriptions of each metrics are provided as follows: the target E2E delay of 150 ms. Hence, the PF scheme is not
1) Packet Loss Ratio: Packet loss refers to the failure of one suitable for real-time multimedia applications. An interesting
or more transmitted packets to reach their destination across observation is seen among all schemes that E2E is almost
a network. For real-time video service application, the packet uniform within the cell in both scenarios. Since there is very
loss should be maintained at a threshold less than 1% [2]. tiny difference in delay, which algorithm can best support
From Fig. 2(a), it can be concluded that all the schemes failed VOIP services in the downlink LTE system is inconclusive.
to meet the QoS requirements of video streaming service. In 4) Cell Throughput: The cell throughput performance for
other hand, a VoIP user is satisfied if more than 98% of its video and VOIP flows with increasing number of users is
voice packets are delivered successfully [13]. From Fig. 3(a), shown in Figs. 2(d) and 3(d) respectively. With increasing
it can be observed that less than 60 and 80 UEs can be number of users, PF remains the worst performing scheme
supported by PF and M-LWDF, EXP/PF, EXP-Rule, LOG- despite the tradeoff between fairness and throughput. As the
Rule schemes respectively. In other hand, the FLS scheme FLS computes the amount of data that each real-time source
should transmit within a single frame, which, in other words,
3 The video sequences are available at http://trace.eas.asu.edu/yuv/ take care to assign radio resources to users. The FLS scheme

113
111
 (a) Packet Loss Ratio Vs. Number of Users (a) Packet Loss Ratio Vs. Number of Users

(b) Fairness Vs. Number of Users (b) Fairness Vs. Number of Users

(c) Delay Vs. Number of Users (c) Delay Vs. Number of Users

(d) Cell Throughput Vs. Number of Users (d) Cell Throughput Vs. Number of Users

Fig. 2. Video Streaming Performance - Varying Number of Users Fig. 3. VOIP Streaming Performance - Varying Number of Users

114
112
 stops by the conclusion that, so far, FLS is the best scheduler
supporting video and VOIP streaming over LTE systems.

R EFERENCES
[1] Capozzi, Francesco, Giuseppe Piro, Luigi Alfredo Grieco, Gennaro Bog-
gia, and Pietro Camarda. ”Downlink packet scheduling in lte cellular
networks: Key design issues and a survey.” Communications Surveys &
Tutorials, IEEE 15, no. 2 (2013): 678-700.
[2] Janevski, Toni. Traffic analysis and design of wireless IP networks. Artech
House, 2003.
[3] Jalali, A., R. Padovani, and R. Pankaj. ”Data throughput of CDMA-
HDR a high efficiency-high data rate personal communication wireless
system.” In Vehicular Technology Conference Proceedings, 2000. VTC
2000-Spring Tokyo. 2000 IEEE 51st, vol. 3, pp. 1854-1858. IEEE, 2000.
Fig. 4. Spectral Efficiency Vs. Number of Users [4] Andrews, Matthew, Lijun Qian, and Alexander Stolyar. ”Optimal utility
based multi-user throughput allocation subject to throughput constraints.”
In INFOCOM 2005. 24th Annual Joint Conference of the IEEE Computer
clearly has a very good cell throughput when compared to and Communications Societies. Proceedings IEEE, vol. 4, pp. 2415-2424.
IEEE, 2005.
other schemes for video flows. But the worrying factor for [5] Rhee, Jong-Hun, Jack M. Holtzman, and Dong-Ku Kim. ”Scheduling
VOIP flows which reflects in an unprecedented manner. In of real/non-real time services: adaptive EXP/PF algorithm.” In Vehicular
general throughput decreases with increasing number of UEs Technology Conference, 2003. VTC 2003-Spring. The 57th IEEE Semi-
annual, vol. 1, pp. 462-466. IEEE, 2003.
but here this is not the case. Hence, it is inconclusive for VOIP [6] Piro, Giuseppe, Luigi Alfredo Grieco, Gennaro Boggia, Rossella Fortuna,
flows. and Pietro Camarda. ”Two-level downlink scheduling for real-time multi-
5) Spectral Efficiency: Successful usage of radio resources media services in LTE networks.” Multimedia, IEEE Transactions on 13,
no. 5 (2011): 1052-1065.
is one of the principle objectives to be accomplished. The [7] Shakkottai, Sanjay, and Alexander L. Stolyar. ”Scheduling for multiple
spectral efficiency (SE) is regarded as the performance mea- flows sharing a time-varying channel: The exponential rule.” Translations
sures for the whole cell. Fig. 4 examines the performance in of the American Mathematical Society-Series 2 207 (2002): 185-202.
[8] Sadiq, Bilal, Seung Jun Baek, and Gustavo De Veciana. ”Delay-optimal
terms of SE for both scenario. The FLS approach achieves opportunistic scheduling and approximations: The log rule.” IEEE/ACM
the lowest SE with varying number of users from 20 to 200 Transactions on Networking (TON) 19, no. 2 (2011): 405-418.
while the EXP/PF approach outperform all other schemes. [9] Andrews, Matthew, Krishnan Kumaran, Kavita Ramanan, Alexander
Stolyar, Phil Whiting, and Rajiv Vijayakumar. ”Providing quality of
When compared with EXP/PF scheme, the LOG-Rule scheme service over a shared wireless link.” IEEE Communications magazine
performs slightly better than the EXP/PF scheme on 20 39, no. 2 (2001): 150-154.
number of UEs but performs slightly lower 20 users onwards [10] Piro, Giuseppe, Luigi Alfredo Grieco, Gennaro Boggia, Francesco
Capozzi, and Pietro Camarda. ”Simulating LTE cellular systems: An
while maintaining a good level of performance. The M-LWDF open-source framework.” Vehicular Technology, IEEE Transactions on
scheme can also be regarded as the competing approach when 60, no. 2 (2011): 498-513.
compared to EXP/PF and LOG-Rule scheme. [11] Ramli, Huda Adibah Mohd, Riyaj Basukala, Kumbesan Sandrasegaran,
and Rachod Patachaianand. ”Performance of well known packet schedul-
V. C ONCLUSION ing algorithms in the downlink 3GPP LTE system.” In Communications
(MICC), 2009 IEEE 9th Malaysia International Conference on, pp. 815-
This paper evaluates the performance of few promising 820. IEEE, 2009.
downlink packet scheduling algorithms for single carrier wire- [12] B. Sahoo, ”Performance Comparison of Packet Scheduling Algorithms
for Video Traffic in LTE Cellular Network”, International Journal of
less systems, highlighting the work flow of a packet scheduler. Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3,
The study was analysed with performance matrices such as PP 09-18, June 2013
packet loss, fairness, packet delay, system throughput and [13] Shen, Ng Liang, Noraniah Abdul Aziz, and Tutut Herawan. ”End-to-
finally spectral efficiency over multimedia streaming traffic in End delay performance for VoIP on LTE system in access network.”
In Information and Communication Technology, pp. 162-172. Springer
a vehicular environment. The study explains the benefits and Berlin Heidelberg, 2013.
limitations of the packet scheduling schemes and its impact [14] Sahoo, B. P. S., et al. ”A Comparative Analysis of Time Stamped Based
on multimedia traffic. During our studies, as obvious, we Packet Scheduling Algorithms.” All India National Seminar on Comput-
ing, Communication and Sensor Network, CCSN - 2010. Institution of
found that video streaming services required more resources, Engineers, Rourkela, India, 2010.
resulting poor performance compared to VOIP traffic analysis. [15] Dahlman, Erik, Stefan Parkvall, Johan Skold, and Per Beming. 3G
The simulation results show that, FLS scheduling scheme evolution: HSPA and LTE for mobile broadband. Academic press, 2010.
[16] Tsybakov, B. ”File transmission over wireless fast fading downlink.”
outperforms other scheduler with lesser packet loss, provid- Information Theory, IEEE Transactions on 48, no. 8 (2002): 2323-2337.
ing better streaming quality and guaranteeing fairness at a [17] Seeling, Patrick, Martin Reisslein, and Beshan Kulapala. ”Network
satisfactory level too. As PF scheme was not designed for performance evaluation using frame size and quality traces of single-layer
and two-layer video: A tutorial.” Communications Surveys & Tutorials,
multimedia service, it has the worst performance in terms of IEEE 6, no. 3 (2004): 58-78.
delay. The FLS provides a higher throughput among all other [18] Seeling, Patrick, and Martin Reisslein. ”Video transport evaluation with
scheduler for video streaming application while all scheduler H. 264 video traces.” Communications Surveys & Tutorials, IEEE 14, no.
4 (2012): 1142-1165.
behaves abnormally for VOIP streaming. Therefore, this study

115
113