Professional Documents
Culture Documents
AbstractLong Term Evolution (LTE) uses Single Carrier Power Control (PC) to adjust the maximum Physical Resource
Frequency Division Multiple Access (SC-FDMA) as the uplink Blocks (PRBs) allocable to a UE within a Transmission Time
transmission scheme. The Quality of Service (QoS) provision is Interval (TTI). The uplink scheduler is expected to be aware of
one of the primary objectives of the wireless network operators. the PC functionality. The scheduler should also be able to
In this paper, the end-to-end QoS performance of Bandwidth and
serve multi-bearer UEs and fulfill QoS requirements of each
QoS Aware (BQA) scheduler for LTE uplink is evaluated in
heterogeneous traffic environment. The BQA scheduler is traffic type. The resources should be allocated to UEs
designed to provide efficient allocation of radio resources to users proportional to data in the UE buffers.
according to the QoS requirements of various traffic classes and
the instantaneous channel conditions. The user QoS provision is II. STATE OF THE ART
ensured by using dynamic QoS weights. Additionally, the delay LTE packet scheduling algorithms available in literature
sensitive traffic is facilitated by employing delay thresholds. The mostly focus on downlink scheduling (e.g. [2,3] and others).
BQA scheduler algorithm supports multi-bearer users. The end- The work related to uplink scheduling rarely addresses the QoS
to-end QoS performance of the scheduler is analyzed in several
provision to multiclass traffic. Seldom have the QoS based
simulation scenarios. The results show that the proposed
scheduler guarantees provision of QoS to users.
uplink scheduling algorithms been designed to support multi-
traffic type users; to our knowledge, none has discussed this
Keywords-component; SC-FDMA; uplink; scheduling; issue.
bandwidth; QoS In the scheduling schemes in [4,5], the QoS requirements of
users in heterogeneous traffic environment are fulfilled by
I. INTRODUCTION employing target data rate based metric for each traffic class. In
Long Term Evolution (LTE) standard is currently being [6,7], the traffic classes are served by utilizing target delay rate
developed by the Third Generation Partnership Project (3GPP) based metric. In [8,9], the traffic classes are served on the basis
with targets of increasing peak user throughput, enhancing of target data loss rate. These scheduling algorithms have been
spectral efficiency and reducing latency [1]. LTE uses designed for downlink. For uplink, the contiguity constraint of
Orthogonal Frequency Division Multiple Access (OFDMA) PRBs in heterogeneous traffic, with multi-traffic type users
and Single Carrier Frequency Division Multiple Access (SC- make these designs impracticable.
FDMA) as the transmission schemes for the downlink and the In [10], a scheduling scheme to ensure QoS provision to
uplink, respectively. These schemes can effectively utilize VoIP users with fairness has been designed. In [11], the target
scare radio resources by dividing the transmission bandwidth delay based scheduling algorithm for QoS provision to multi-
into subcarriers, enhance the resource allocation flexibility and class traffic has been extended from downlink to uplink by
achieve high spectral efficiency. In uplink, the low Peak-to- introducing a factor of power control to scheduling metric.
Average Power Ratio (PAPR) of SC-FDMA signals is utilized However, the negligence of the PRB contiguity constraint
to achieve low battery power consumption of User Equipment means that this algorithm may not be practically feasible for
(UE). SC-FDMA also provides bandwidth allocation wireless SC-FDMA systems. In [12,13], the contiguity
flexibility. constraint issue has been addressed and the scheduling metric
Designing a packet scheduler requires tackling of various is based on target data rate for various QoS classes. However,
conflicting requirements such as channel conditions, fairness, these algorithms support only single-bearer UEs, so the
Quality of Service (QoS) requirements etc. A major SC- resource requests can be treated independently of each other
FDMA constraint in LTE uplink is that the subcarriers and PRBs can be allocated from any portion of the bandwidth.
allocated to a single UE should to be adjacent to each other. If the multi-bearer UEs are present in the system, the resource
The UE transmit power in uplink is adapted according to requests having different QoS requirements within a single UE
500
1) Make a UE-RC table as in Figure 2. Each table element RC0 RC1 RC2
would be the RC metric value of the UE, i.e. the sum of
PRB metric values within that RC. UE0 10 6 3
2) Make all possible combinations of UE-RC allocation UE1 11 10 5
using search-tree algorithm (explained below) while
respecting the contiguity, buffer size and maximum Figure 2. A sample UE-RC table for two UEs and three RCs
allowed bandwidth constraints; determine the resulting
global metric value for each combination as in Figure
3.
3) Choose the combination with best global metric value.
4) Obtain the desired resource allocation from best
combination.
Step 2 is based on a search-tree based resource allocation
algorithm termed as unique Depth-First Search (uDFS)
algorithm with contiguity, buffer size and maximum allowed
bandwidth constraints. The uDFS checks all possible
combinations of RC allocation and discards the combinations
which do not follow the constraints; further depth of such
nodes is not explored. An example of uDFS tree is given in
Figure 3. It is assumed that at most, 2 RCs can be allocated to
a UE. Blue nodes breach contiguity constraint and red nodes
breach maximum PRBs constraint. Figure 3. uDFS for two UEs and three RCs
2) Multi-Bearer User Scheduling: If the multi-bearer users
acquire RCs, the allocated PRBs are further subdivided among The parameter settings used during the simulations are
the user bearers. Each user bearer has its own QoS given in TABLE II. In each scenario, UEs with different
requirements related to delay budget, rate budget and delay bearers are studied. In order to keep the environment loaded, 8
threshold. The user bearers require adequate resources to FTP single-bearer UEs per cell are deployed additionally to the
ensure QoS provision and avoid bearer starvation. In this UEs under study in all scenarios.
work, the bearers get weighted service, i.e. a bearer is 1) Scenario 1 (FTP): This scenario depicts a case where 9
served according to its QoS weight Wi ,k (t ) . However, the UEs are under study (along with 8 FTP UEs deployed for
bearers having reached their packet delay threshold are given loading the environment). Among the UEs under study, 3 UEs
strict priority and the available resources are allocated to them have only FTP data; 3 UEs have FTP and VoIP data; while 3
before serving other bearers. UEs have FTP and video data. The goal of this scenario is to
check whether the FTP bearer of each UE is served
IV. SIMULATION RESULTS AND ANALYSIS consistently and fairly, with or without the presence of other
The QoS performance of BQA scheduler is analyzed with bearers, i.e. VoIP and video.
several simulation scenarios. The scheduler performance is Results of this simulation are depicted in Figure 4. Results
compared with other contemporary schedulers in [15,16]. Due show that the provision of QoS to UEs with FTP data is
to the multi-bearer UE scheduling capability of the BQA consistent over all the UEs. The reason for slightly higher
scheduler; it seems probable that the QoS provision feature of response times of FTP-video UEs as compared to other UEs is
the scheduler in TDPS might prioritize a low priority bearer of the higher data rate requirement of video traffic. The
UE1 due to the presence of high priority bearer of UE1, over Cumulative Distribution Function (CDF) of the average FTP
the same type of bearer of UE2 without presence of high file upload time for all the UEs in the cell is shown in Figure 5.
priority bearer in UE2. Similarly, as compared to a high 2) Scenario 2 (VoIP): This scenario illustrates a case
priority single-bearer UE3; it might be expected that a high where 9 UEs are simulated such that 3 UEs have only VoIP
priority bearer may not be served according to its QoS in a data; 3 UEs have VoIP and FTP data; while 3 UEs have VoIP
UE4 having a low priority bearer as well. The goal of these and HTTP data. This scenario investigates whether the delay
simulations is to determine if the provision of QoS to UEs is requirements of the VoIP bearer of each UE are served
consistent, in various traffic scenarios. consistently and fairly with or without the presence of other
bearers. VoIP is a delay sensitive bearer and the presence of
other bearers within a UE should not hinder the QoS provision
to VoIP.
Results of this simulation are depicted in Figure 6. Results
show that the provision of QoS to UEs with VoIP data is not
Figure 1. Chunk size 3 of M RCs and bandwidth of N PRBs affected by presence of other bearers in a UE. The range of
501
average UE packet end-to-end delays is between 0.069 and
0.074sec and depicted as CDF in Figure 7.
Parameter Setting
Cell layout 1 eNodeB, 3 cells
System Bandwidth 5 MHz (~25 PRBs)
Frequency reuse factor 1
Cell radius 375m
UE velocity 120kmph
Max UE power 23dBm
Path loss 128.1+37.6log10(R), R in km
Log-normal shadowing, 8dB standard
Slow fading
deviation, correlation 1 Figure 4. FTP average file upload time for 9 different UEs
Fast fading Jakes-like method [17]
Mobility Model Random Way Point (RWP)
UE buffer size Infinite
Power Control FPC, = 0.6, P0 = -58dBm
Traffic environment Loaded
Max FDPS UEs 5
RC size 5
VoIP traffic model
Silence/ talk spurt length Exponential(3) sec
Encoder scheme GSM EFR
Video traffic model
Frame size 1200 bytes
Frame inter-arrival time 75ms
Figure 5. CDF of FTP average cell file upload time [sec]
HTTP traffic model
Page size 100Kbytes
Page inter-arrival time 12 sec
FTP traffic model
File size 20Mbytes
File inter-request time Uniform distribution, min 80s, max 100s
3) Scenario 3 (video): Another case is illustrated in
scenario 3 with 9 UEs being simulated. 3 UEs have only video
data to transmit; 3 UEs have video and FTP data; while 3 UEs
have video and VoIP data. This scenario aims to determine
whether the delay requirements of video bearer of each UE are
served consistently and fairly irrespective of the presence of
other bearers. The presence of other bearers should not hinder
the QoS provision to video bearer of UEs.
Figure 6. VoIP average packet end-to-end delay for 9 different UEs
Results of this simulation are depicted in Figure 8. Results
of this scenario also confirm the provision of QoS to UEs with
video data. It is shown that the video data rate is not adversely
affected by presence of other bearers within a UE. The average
UE packet delay time for all UEs is similar to each other. It
ranges between 0.034sec and 0.036sec as in Figure 9.
4) Scenario 4 (HTTP): Another case for 9 UEs is
illustrated in this scenario. 3 UEs have only HTTP data to
transmit; 3 UEs have HTTP and video data; while 3 UEs have
HTTP and VoIP data. This scenario aims to determine whether
the HTTP page response time offered by the scheduler to
HTTP bearer of each UE is consistent and fair irrespective of
the presence of other bearers. The goal is to ensure QoS
provision to HTTP bearer of all UEs. Figure 7. CDF of VoIP average cell packet end-to-end delay [sec]
502
Results of this simulation are depicted in Figure 10. The
graphs illustrate the fair provision of QoS to UEs with HTTP
data in this scenario. The presence of other bearers within a UE
does not hinder HTTP page response for the UEs. The average
UE page response time for all UEs is within the range of
0.666sec and 0.806sec as given in Figure 11.
5) Scenario 5 (FTP, VoIP and video): This scenario
inspects the performance of the scheduler by checking QoS
provision to 3 different bearers. In this case, 9 single-bearer
UEs and 6 multi-bearer UEs are simulated. The single-bearer
UEs include 3 VoIP, 3 video and 3 FTP UEs. Among the
multi-bearer UEs, 3 UEs have VoIP and FTP traffic, while 3
have video and FTP traffic for uplink. The purpose of this
Figure 8. Video average packet end-to-end delay for 9 different UEs scenario is to determine the QoS facilitation capability of the
scheduler to different traffic type UEs simultaneously in
heterogeneous traffic environment.
The simulation results are depicted in Figure 12. to Figure
14. The FTP, VoIP and video QoS performance of the
scheduler is shown to be acceptable. The low and high priority
bearers are served fairly in single and multi-bearer UEs.
V. CONCLUSION AND OUTLOOK
In this paper, the QoS performance of the BQA scheduler is
evaluated in heterogeneous traffic environment. The QoS
requirements of multi-bearer UEs are efficiently fulfilled by the
scheduler using dynamic QoS weights. The scheduler is time
and frequency domain decoupled. The resource allocation is
Figure 9. CDF of video cell packet end-to-end delay [sec] performed by exploiting the bandwidth flexibility feature of
SC-FDMA. The contiguity constraint of subcarriers is also
tackled by the scheduler design. UE buffer size plays a role in
acquiring resources. Simulation results confirm that the
scheduler performs consistently and fairly to guarantee QoS
demands of the various types of UE bearers. In future, the
scheduler design can be further enhanced by implementing the
Admission Control (AC) functionality.
Figure 10. HTTP average page response time for 9 different UEs
Figure 12. FTP average file upload time for 9 different UEs
503
Vehicular Technology Conference, 20-23 Sept. 2009, pp. 1 - 5.
[6] M. Assaad, "Frequency-Time Scheduling for Streaming Services in
OFDMA Systems," in 1st IFIP Wireless Days, 24-27 Nov. 2008, pp. 1 -
5.
[7] G. Piro, L. A. Grieco, G. Boggia, and P. Camarda, "A Two-level
Scheduling Algorithm for QoS Support in the Downlink of LTE Cellular
Networks," in European Wireless Conference, 12-15 April 2010, pp. 246
- 253.
[8] Yunzhi Qian, Canjun Ren, Suwen Tang, and Ming Chen, "Multi-Service
QoS Guaranteed Based Downlink Cross-Layer Resource Block
Allocation Algorithm in LTE Systems," in International Conference on
Wireless Communications & Signal Processing, 13-15 Nov. 2009, pp. 1 -
4.
[9] Chong Lou and Ling Qiu, "QoS-Aware Scheduling and Resource
Allocation for Video Streams in e-MBMS Towards LTE-A System," in
Figure 13. VoIP average packet end-to-end delay for 9 different UEs IEEE Vehicular Technology Conference, 5-8 Sept. 2011, pp. 1 - 5.
[10] Zhijie Wang, Yafeng Wang, Dajie Jiang, Chunchang Tian, and Dacheng
Yang, "Scheduling and Link Adaptations for VoIP in TDD-LTE Uplink,"
in 5th International Conference on Wireless Communications,
Networking and Mobile Computing, 24-26 Sept. 2009, pp. 1 - 5.
[11] Hung Pham, Xuan Nhan Vu, and Seung-Hoon Hwang, "Service Class-
aided Scheduling for LTE," in 13th International Conference on
Advanced Communication Technology, 13-16 Feb. 2011, pp. 39 - 43.
[12] M. Anas, C. Rosa, F. D. Calabrese, K. I. Pedersen, and P. E. Mogensen,
"Combined Admission Control and Scheduling for QoS Differentiation
in LTE Uplink," in IEEE 68th Vehicular Technology Conference, 21-24
Sept. 2008, pp. 1 - 5.
[13] F. Ghandour, M. Frikha, and S. Tabbane, "A Fair and Power Saving
Uplink Scheduling Scheme for 3GPP LTE Systems ," in International
Conference on the Network of the Future, 28-30 Nov. 2011, pp. 6 - 9.
[14] O. Delgado and B. Jaumard, "Scheduling and Resource Allocation for
Multiclass Services in LTE Uplink Systems," in IEEE 6th International
Figure 14. Video average packet end-to-end delay for 9 different UEs Conference on Wireless and Mobile Computing, Networking and
Communications, 11-13 Oct. 2010, pp. 355 - 360.
REFERENCES [15] S. N. K. Marwat, T. Weerawardane, Y. Zaki, C. Goerg, and A. Timm-
Giel, "Performance Evaluation of Bandwidth and QoS Aware LTE
[1] H. Holma and A. Toskala, LTE for UMTS OFDMA and SC-FDMA Uplink Scheduler," in 10th International Conference on Wired/Wireless
Based Radio Access, Illustrated ed. The Atrium, Southern Gate, Internet Communications, Santorini, Greece, 6-8 June, 2012, in press.
Chichester, West Sussex, PO19 8SQ, United Kingdom: John Wiley &
[16] S. N. K. Marwat, T. Weerawardane, Y. Zaki, C. Goerg, and A. Timm-
Sons Ltd, 2009.
Giel, "Performance of Bandwidth and QoS Aware LTE Uplink Scheduler
[2] Liljana Gavrilovska and Dejan Talevski, "Novel Scheduling Algorithms Towards Delay Sensitive Traffic," in 17. ITG Fachtagung
for LTE Downlink Transmission," in 19th Telecommunications Forum, Mobilkommunikation, Osnabrueck, Germany, 9-10 May, 2012, in press.
22-24 Nov. 2011, pp. 398 - 401.
[17] J. K. Cavers, Mobile Channel Characteristics.: Kluwer Academic
[3] Dandan Wang, R. Soni, Pichun Chen, and A. Rao, "Video Telephony Publishers, 2002.
over Downlink LTE Systems with/without QoS Provisioning," in 34th
IEEE Sarnoff Symposium, 3-4 May 2011, pp. 1 - 5.
[4] G. Monghal, K. I. Pedersen, I. Z. Kovcs, and P. E. Mogensen, "QoS
Oriented Time and Frequency Domain Packet Scheduler for the UTRAN
Long Term Evolution," in IEEE Vehicular Technology Conference,
Spring 2008, 11-14 May 2008, pp. 2532 - 2536.
[5] D. Laselva et al., "On the Impact of Realistic Control Channel
Constraints on QoS Provisioning in UTRAN LTE," in IEEE 70th
504