A CQI based Novel Shared Channel Downlink Scheduler
for LTE Networks
Mandeep Singh Ramdev1[0000-0001-6389-1718] and Rohit Bajaj2] and Ruchika Gupta3
1
Chandigarh University, Punjab, India
2
Chandigarh University, Punjab, India
3
Chandigarh University, Punjab, India
[Link]@[Link]
Abstract. Long Terrn Evolution (LTE) can be called as the new generation of
High Speed Cellular Communication. LTE networks serves as back bone for
4G networks delivering high data transmission speeds and support for Quality
of Service (QoS). It also ensures the availability of high speed connection, HD
Calling, more security and extended support for streaming of HD multimedia
content which includes audio and video content. With this much development
in the field of mobile communication, another term was coined QoE (Quality of
experience) which refers to the overall degree of acceptability of the multimedia
content as perceived by the end users. In this paper we introduce a CQI based
algorithm to improve the overall QoE while it is being applied on downlink
scheduling. Simulation runs proves that CQI has better results as compared to
other algorithms based upon parameters such as throughput, SnR and fairness.
Keywords: QoS, Scheduling, CQI, LTE, Downlink.
1 Introduction
In view of the recent trends, it is quite clear that wireless communication has emerged
as the communication type registering exponential growth in data volume as well as
user base. LTE architecture is basically the brainchild of 3rd Generation Partnership
Project popularly known as 3GPP which provides high speed Wireless Broadband
access upto speeds of 1 Gbps via IP based packet switched network.
The high data speeds of LTE networks is very important as the user is churning out
extra money for this connection, secondly according to survey done by Accenture [1]
90% of the unsatisfied users, change the ISP instead of complaining about the low
data speeds. Now this has quite an impact on the revenues of ISP‟s offering LTE ser-
vices. Therefore it is quite important for them to maintain required QoE level.
The efficient design and implementation of scheduling algorithms has been the main
point of research as it is one of the way to maximize QoE. The target of this work is
to make optimum usage of network resources so as to provide high speed data to end
users. This optimum usage of network resources is calculated on various parameters
to ensure that the data received by the end user satisfies his need of agreed QoS. The
various parameters used are throughput, delay and fairness.
2
2 Issues in LTE Networks
There are many challenges faced by LTE and LTE-A systems hen we discuss about
resource management. The first issue is the Multiuser Problem. As the LTE system is
spread over a large geographical area and there are several UE‟s in that area, the ma-
jor challenge is to serve all the User Equipments (UE) with minimum latency in that
cell.
The resource allocation in LTE is done via OFDMA. The users needs to share the
bandwidth allocated to the UE in the area to which it is connected. Here the role of
scheduling algorithms came into play. Scheduling algorithms increase the efficiency
of the system by dealing with each type of traffic differently.
2.1 Aspects impelling the Scheduling Algorithms
One of the main challenges faced while developing scheduling algorithm is the
changing nature of the network. There are several variations in the network at any
given instance of time which may ultimately result in fluctuating channel conditions,
conclusively it is very hard to uphold same modulation process for transmission of
data to all UE and LTE system has routed to Adaptive Modulation and Coding Tech-
nique (AMC) to overcome this hurdle [2]. It is also significant to recollect that spec-
tral efficiency is not the only factor that needs deliberation while designing schedul-
ing algorithms. The other parameters that require special attention are as follows:
1. Fairness
2. Complexity
3. Signal to Interface plus Noise Ratio
4. Channel Conditions
5. Packet Delay
6. Compatibility support of the algorithm with hardware
Now we will discuss all these parameters in detail to get an idea about the signifi-
cance of these parameters and their overall impact on the system.
2.1.1 Fairness: Fairness as a parameter is very difficult to explain and understand
as in homogenous LTE system, there are different types of traffic and each type of
traffic has its special need for resources. The challenge lies in the fact that each
traffic flow should get required resources and none of the traffic flow should ever
come in starvation stage. This definition of fairness does not hold good when we tal
about heterogeneous systems. According to some authors, there are two types of
fairness. [2]
A) Partial Fairness: Partial Fairness may be defined as the degree of fairness
in the same domain. By domain we mean that he similar types of flows. For
example if we are taking in consideration only real time flows like audio or
video streaming, If we are only concerned about a single type of traffic flow
and we are doing justice to that flow, it may be defined as partial fairness.
B) Total Fairness: Total Fairness is the measure of fairness if we consider all
traffic flows at a single instance of time. For example, in a network, traffic of
3
various classes is transmitting at an instance like video streaming, VoIP, text
and audio which can also be termed as heterogeneous traffic; maintaining fair-
ness in terms of resource allocation in this case may be termed as total fair-
ness.
2.1.2 Service Type: LTE has two service types viz. real time and non-real time.
This is one of the major advantages of LTE as this coarse classification helps reduc-
ing overhead while transmission of data. Real time data has the higher priority as
compared to non-real time data. Delay is also a factor which differentiates between
real time and non-real time data. Real time data includes VoIP, Video conferencing
and streaming. Non-real time data includes web browsing and email.
2.1.3 Channel Conditions: Channel condition is a very important factor that needs
special attention while making scheduling decisions. There are a lot of factors that
affect quality of channel like path fading, shadow fading, the ever changing distance
between UE and eNodeB and congestion. Sometimes external weather conditions lead
to packet loss and hence quality of channel is degraded.
2.1.4 Packet Delays: Packet delay is one of the major reasons for performance
degradation of the channel. It has a direct effect on real time services like video
streaming and VoIP and ultimately the required QoS levels are not achieved. The
ideal packet scheduling algorithm should ensure that the packet delay is minimum.
2.1.5 Complexity: Complexity of an algorithm defines that how much time the al-
gorithm will take in determining and sending the packet to its destination. More com-
plex algorithms will lead to more overhead in sending data and will result in unneces-
sary packet delays.
The algorithm designed should be simple enough so that the output overhead is
minimum. In LTE, resource allocation is estimated to be a function of TTI which is
normally 1ms so we expect that an optimized algorithm should take less time than
TTI.
3 Literature Review
As we all know that OFDMA (Orthagonal Frequency Division Multiple Access) is
the base of LTE systems. The main target of all algorithms designed for scheduling is
to take advantage of all the qualities of OFDMS for maximization of resource utiliza-
tion and all the decisions pertaining to resource allocation are taken by centralized
scheduler. There are lots of scheduling algorithms which are better and optimized in
their own way. We will be discussion some of the scheduling algorithms which have
left a great impact on scheduling in LTE. For the sake of clarity, we have classified
schedulers according to the type of functionality they offer.
In [3] author discussed Proportional fair scheduling algorithm which is implement-
ed in several high speed wireless networks. The author insisted that this algorithm is
an excellent choice for non-real time traffic. The resource allocation is done to User
Equipment (UE) strictly based upon channel quality. The aim of this algorithm is
ensuring fairness as well as maximizing throughput.
4
Proportional fair (PF) schedulers were also used by Kwan et al. [4]. Pf is the most
commonly used algorithm in 3G networks, so it was thought that is would also work
well with 4G systems as well. But the results proved otherwise. The basic reason was
the extensive support for real time data in 4G networks and the clause of time sensi-
tivity because of this it generally also termed non elastic service flow. With rigorous
experimentation, the authors proved that PF holds good for non-real time data but
desired QoS standards are not achieved for real time data and applications.
Fig 1: PF Algorithm Flow Chart
Basukala et al. [5] worked on exponential proportional fairness algorithm (EPF)
which has been exclusively designed to support multimedia applications. This algo-
rithm works according to the demand of users. It uses adaptive modulation and coding
which adapts according to the resources demanded by the users. It means that this
adaptive algorithm provides higher priority to real time applications as compared to
non-real time applications. One disadvantage of this algorithm is that it does not hold
5
good while handling buffer delays and its performance is same as of traditional PF
algorithm. So as a result of time delay this algorithm is not very extensively used for
4G networks.
Another algorithm was proposed by Park et al. [6] which was basically delay based
algorithm which was designed especially for real time applications. In this algorithm a
concept of Head of Line (HOL) value was also introduced which was very effective
in handling real time data. There was only one disadvantage of HOL that when pack-
ets exceeded the threshold they got discarded and ultimately resulting in unwanted
retransmission and packet loss and hence degradation of QoS.
Ameigeiras et al. [7] proposed Maximum-Largest Weighted Delay First Algorithm
which supported multiusers in 3G and 4G networks. The main focus of this algorithm
was real time data keeping in consideration the current channel quality. This algo-
rithm aimed to support multiple applications with different QoS. This algorithm does
not hold good for non-real time data.
Sadiq et al. [8] proposed a scheme which proportionally balanced real time and
non-real time applications. The scheduling decisions were based upon the length of
the queue. But due to very high computational overhead, this scheme was not very
successful. Authors used packet delay based max-weight scheduling scheme. The
outputs were more optimized as compared to other schemes.
Leinonen et al. [9] and Varadarajan et al. [10] came up with a scheduling scheme
using Round-Robin scheduler. Round Robin is a very traditional technique which has
minimum computation overhead. Round Robin technique takes processes in a line
and maintains a time quantum in which every process should run and then go to wait-
ing state. Waiting time for processes is reduced to a great extent using Round Robin.
On the other side, throughput is degraded by a great extent in this algorithm. It is
because of the inability of the algorithm to take into consideration the current channel
conditions. The required QoS standards cannot be maintained in an environment
where each type of service like VoIP, email, video streaming has its own specific QoS
requirements.
The scheduling scheme proposed by Niu et al. [11] works on multiuser diversity
and frequency. The status of channel is obtained by using CSI (Channel State Infor-
mation). User classification based upon CSI was used to develop a scheduling algo-
rithm which gets the status of frequency for use in high mobility users and status of
multi-user diversity for low mobility users. During experimentation they came to the
conclusion that results were better with respect to throughput and PF but it does not
handle buffer delays well.
A scheme proposed by Prasad et al. [12] was a multi user scheduling scheme which
supported multiple users. It was based on the physical layer interface of OFDMA. The
results hence obtained were also improved. The main disadvantage of scheme was the
computational overhead due to complex algorithms used.
A scheduler was proposed by Donthi et al. [13] which was best suited for synthetic
networks. It was a feedback based scheme which used subband level feedback and
user selected subband feedback techniques as described in LTE. The disadvantage of
the algorithm was that it often lead to wrong determination of rate for some resource
blocks by the scheduler.
6
Work done by Gotsis et al. [14] focused on machine to machine communication.
M2M is a challenge in itself as here are n number of devices each of different make
and use. These devices have different range, frequencies and area of application. In
LTE networks, the heterogeneity of these devices has increased to a great extent
which poses a direct degradation in QoS in case the communication is not happening
as it should be. To cater these problems, the devices were broadly categorized and
each pair of communicating devices was treated with a flexible scheduling algorithm.
But this resulted in signaling overhead.
Capozzi et al. [15] in his research gave an overview of key design and scheduling
issues in LTE networks with special reference to QoS provisioning. Authors also very
keenly studied and tried to optimize the power costs incurred, by letting the radio
devices sleep while maintaining the required amount of QoS on the parameters of
bitrate, packet loss ratio and packet delay.
Nyugen et al. [16] proposed a multiple component carrier based PF algorithm. This
property of LTE systems enables it to boost data rates exponentially. Authors also
found that majority of scheduling algorithms were unable to gain full advantage of
this feature of LTE. The algorithm proposed by authors provided better packet sched-
uling than existing algorithms.
Another algorithm was discussed by Tsai et al. [17] which took into consideration
the QoS required by every service class. This algorithm was made using divide and
conquer scheme. First part known as dynamic priority adjustment gave highest priori-
ty to users which required real time data by dynamically adjusting frame value in line
with QoS requirement of the user. The second part known as Priority based Greedy
Algorithm used Greedy approach in order to reach required QoS.
Xu et al. [18] proposed an uplink transmission scheme to exploit the very less ex-
plored parts of LTE-A systems. This algorithm evaluated LTE systems on Host to
Host and Machine to Machine scenarios. This algorithm was also capable of handling
power and resource block allocation dynamically. Simulations established a major
performance augmentation.
4 Model and Simulation Set-up
This work focusses on implementing CQI scheduling algorithm. In the past there are
several works which implemented CQI scheduling algorithm and the results they got
were also on the positive side. The main advantage of CQI is its simplicity, less over-
head and the quality of providing feedback. This helps in smooth data transmission
because of availability of channel. The earlier works proposed CQI scheduler to be
implemented in eNodeB whereas we will be implementing this algorithm at Downlink
Shared Channel also known as LTE DL-SCH.
Figure 2 shows a representation of DL-SCH. Here we have installed a scheduler
whose function is to schedule the incoming data to various streams so that they can be
processed simultaneously. This scheduler will be coded with CQI algorithm as CQI
has proven track record of providing satisfactory QoS levels as compared with other
7
algorithms. This scheduler will divide the data stream into many channels according
to the priority, the type of traffic and the channel status. Each data will first be
brought under CRC check which checks data for any inconsistencies. Then the data is
forwarded to code block segmentation and after that it goes to Turbo Coding block.
After that rate matching is done, the data is again segregated and forwarded to the
channel with respect to the type of data guaranteeing satisfactory levels of QoS.
Figure 2: DownLink Shared Channel Processing
4.1 Parameters Setting
Table 1 shows the parameters used and the simulations settings for the experimental
setup.
Table 1: Simulation Parameters
Parameter Value
Number of Users 100
Number of eNodeB 1
Bandwidth 20MHz
Channel type Pedestrian and Vehicular
Simulation Length 100 sub-frames
Scheduling Algorithm CQI
Transmission Scheme MIMO
4.2 Results and Discussions
This simulation was performed on a proprietary simulator and the here UE data is
transmitted from the co-operating eNodeB and based upon Channel quality reports
from the shared channel.
8
Figure 3: SnR Vs Throughput
In this experiment, we have considered four types of data which needs to be transmit-
ted which is audio streaming, video streaming, simple text transmission or web
browsing and non- real time data which may include some graphic content. As shown
in figure 2, scheduler has divided data into two channels. Now according to the type
of data received, the CQI scheduler checks the quality of the channel and forwards the
data to the channel which is comparatively free
9
Figure 3: Amplitude Constellation
.
Here in this case when the data has been transmitted, the throughput started from 30%
and has shown a gradual increase until the SnR is 1 which is more optimized as com-
pared to other algorithms. The constellation hence formed also gives a fine view
about the scattering of data in the channel which is very even in our case.
From the above experiments we have come to the conclusion that CQI is best suited
for LTE when applied at shared downlink channel and it does justice to almost all
type of data received.
References
1. T. Ghalut, H. Larijani, A. Shahrabi, “QoE-aware optimization of video stream
downlink scheduling over LTE networks using RNNs and generic algorithm”,
Procedia Computer Science 94 (2016) 232–239.
2. Nagarajan. N “Bandwidth Allocation and Scheduling for quality of service in
LTE Systems” Thesis. Anna University, Chennai, 2015. Shodhganga. Web 16
march 2017.
< [Link]
3. Kwan,R, Leung, C & Zhang, J 2009, „Proportional fair multiuser scheduling in
LTE‟, IEEE Signal Processing Letters, Vol. 16, no.6, pp. 461-464.
4. Borst, S 2003, „User-level performance of channel-aware scheduling algo-
rithms in wireless data networks‟, IEEE International Conference on Comput-
er and Communications Conference (INFOCOM), Vol. 1,no.1, pp. 321-331.
10
5. Basukala, R, Mohd Ramli, H & Sandrasegaran, K 2009, „Performance analysis
of exp/pf andmlwdf in downlink 3gpp LTE system‟, IEEE First Asian Hima-
layas Conference, vol. 1, pp. 1-5.
6. Park, W, Cho, S & Bahk, S 2006, „Scheduler design for multiple traffic classes
in ofdma networks‟, IEEE International Conference Communications (ICC),
vol. 2, pp. 790-795.
7. Ameigeiras, P, Wigard, J & Mogensen, P2004, „Performance of the m-lwdf
scheduling algorithm for streaming services in hsdpa‟, IEEE Vehicular Tech-
nology Conference (VTC), vol. 2, pp. 999-1003.
8. Sadiq, B & de Veciana, G 2008, „Optimality and large deviations of queues
under the pseudo-log rule opportunistic scheduling‟, IEEE Allerton Confer-
ence, Communication, Control, and Computing, pp. 776-783.
9. Leinonen, J, Hamalainen, J & Juntti, M 2009, „Performance analysis of down-
link ofdma resource allocation with limited feedback‟, IEEE Transactions on
Wireless Communications, Vol. 8, no.6, pp. 2927-2937
10. Varadarajan, Chen, R, Onggosanusi, EN, II Han Kim & Dabak, AG 2009, „Ef-
ficient channel quality feedback schemes for ofdma systems with different
schedulers‟, IEEE Vehicular Technology Conference (VTC), pp 1-5.
11. Jinping Niu, Daewon Lee, Xiaofeng Ren, Li, GY & Tao Su 2013, „Scheduling
Exploiting Frequency and Multi-User Diversity in LTE Downlink Systems‟,
IEEE Transactions on Wireless Communications, vol.12, no.4, pp.1843-1849.
12. Prasad, N, Honghai Zhang, Hao Zhu & Rangarajan, S 2014 „Multiuser Sched-
uling in the 3GPP LTE Cellular Uplink‟, IEEE Transactions onMobile Com-
puting, vol.13, no.1, pp.130-145.
13. Donthi, SN & Mehta, NB 2011, „Joint Performance Analysis of Channel Qual-
ity Indicator Feedback Schemes and Frequency-Domain Scheduling for LTE‟,
IEEE Transactions on VehicularTechnology,vol.60, no.7, pp. 3096-3109.
14. Gotsis, AG, Lioumpas, AS & Alexiou, A 2012, „M2M Scheduling over LTE:
Challenges and New Perspectives‟, IEEE Magazine on Vehicular Technolo-
gy‟, vol.7, no.3, pp.34-39.
15. Capozzi, F, Piro, G, Grieco, LA, Boggia, G & Camarda, P 2013, „Downlink
Packet Scheduling in LTE Cellular Networks: Key Design Issues and a Sur-
vey‟, IEEECommunications Surveys & Tutorials, vol.15, no.2, pp. 678-700.
16. Nguyen, SC& Sandrasegaran, K 2012, „Optimised proportional fair algorithm
for long term evolution-advanced system with multiple component carriers‟,
IET Communications,vol.6, no.11, pp.1579-1586.
17. Tsai ,CF Chang, CJ, Ren, FC & Yen, CM 2008, „Adaptive radio resource allo-
cation for downlink OFDMA/SDMA systems with multimedia traffic‟, IEEE
Transactions on Wireless Communication, vol. 7, no. 5, pp. 1734–1743.
18. Xu, H, Tian, H, Feng, Y, Gao, Y & Zhang, P 2008, „An efficient resource
management scheme with guaranteed QoS of heterogeneous servicesin
MIMO-OFDM system‟, IEEE Wireless Communication and Networking Con-
ference, pp. 1838-1843.