A Fair Downlink Packet Scheduling Approach to Support QoS in HSDPA

Transcription

1 A Fair Downlink Packet Scheduling Approach to Support QoS in HSDPA Deepti Singhal and Naresh Jotwani The First International Conference on COMmunication Systems and NETworkS (COMSNETS) January 9, 29

2 Contents 1 Introduction 2 Problem Statement 3 Other Proposed Scheduling Algorithms 4 Proposed Solution Approach 5 Performance Measure 6 Simulation Environment 7 Simulation Results Web-Traffic Scenario FTP-Web Traffic Scenario Load Analysis 8 Summary & Conclusion 9 References 1 Thank You

3 Introduction Introduction In the present scenarios, the high-speed demands are growing rapidly specially in wireless networks. In order to satisfy fast-growing demands in downlink, the concept of High Speed Downlink Packet Access has been proposed. Packet scheduler in HSDPA controls the allocation of shared resources to users.

4 Problem Statement Problem Statement The objective of the paper is to design and evaluate a downlink-scheduling algorithm for HSDPA based UMTS networks. The resulting algorithm should ensure better channel utilization and fairness among users.

5 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Maximum Carrier to Interference Ratio Scheduling [1]: Where r i (k) is instantaneous C/I ratio j = max i r i (k) (1) Proportional Fair Scheduling [1, 2]: According to PF, user j is selected, if j = max i R i is the average received rate for user i DRC Exponent Rule [3] r i R i, i = 1, 2,..., N. (2) j = max i r i n R i, i = 1, 2,..., N. (3)

6 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Maximum Carrier to Interference Ratio Scheduling [1]: Where r i (k) is instantaneous C/I ratio j = max i r i (k) (1) Proportional Fair Scheduling [1, 2]: According to PF, user j is selected, if j = max i R i is the average received rate for user i DRC Exponent Rule [3] r i R i, i = 1, 2,..., N. (2) j = max i r i n R i, i = 1, 2,..., N. (3)

7 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Maximum Carrier to Interference Ratio Scheduling [1]: Where r i (k) is instantaneous C/I ratio j = max i r i (k) (1) Proportional Fair Scheduling [1, 2]: According to PF, user j is selected, if j = max i R i is the average received rate for user i DRC Exponent Rule [3] r i R i, i = 1, 2,..., N. (2) j = max i r i n R i, i = 1, 2,..., N. (3)

8 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Maximum Carrier to Interference Ratio Scheduling [1]: Where r i (k) is instantaneous C/I ratio j = max i r i (k) (1) Proportional Fair Scheduling [1, 2]: According to PF, user j is selected, if j = max i R i is the average received rate for user i DRC Exponent Rule [3] r i R i, i = 1, 2,..., N. (2) j = max i r i n R i, i = 1, 2,..., N. (3)

9 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Contd... Adaptive Proportional Fair Scheduling [4]: In APF, the user selection criterion is given by c i are updated with r c i i j = max i RT i, i = 1, 2,..., N, (4) R i c i = c i + c, if c i = c i c, if ( Ri ( RT i Ri RT i 1 N 1 N N j=1 N j=1 R j ) RT j < ε ) R j. (5) RT j > ε

10 Other Proposed Scheduling Algorithms Other Proposed Scheduling Algorithms Contd... Adaptive Proportional Fair Scheduling [4]: In APF, the user selection criterion is given by c i are updated with r c i i j = max i RT i, i = 1, 2,..., N, (4) R i c i = c i + c, if c i = c i c, if ( Ri ( RT i Ri RT i 1 N 1 N N j=1 N j=1 R j ) RT j < ε ) R j. (5) RT j > ε

11 Proposed Solution Approach Proposed Solution Approach Framework for developing Wireless Fair Scheduling Algorithm [5] Select user i according to Error free Model Is leading?? Yes should relinquish this slot?? Yes Select user j according to Yes Compensation Model Is i = j?? No Update the lead/lag parameterof user i & j (Lead Lag Model) Yes No bad channel?? No Schedule user i Schedule user j Main components of framework are: Error Free Service Model Compensation Model Lead and Lag Model The proposed solution approach is based on the framework discussed and it replaces Internal working of Model of this framework.

12 Proposed Solution Approach Wireless Fair - High Speed Scheduling Proportion Fair Service Model: This model selects a user with max i r i R i RT i (6) Compensation Model: The maximum limit of lead after which user will relinquishes its slot is given by max_l i = RT i TTIduration mean_channal_rate i (7) This model is also responsible for selecting user after relinquishing. Among the lagging users who perceives good channel will be scheduled.

13 Proposed Solution Approach Wireless Fair - High Speed Scheduling Proportion Fair Service Model: This model selects a user with max i r i R i RT i (6) Compensation Model: The maximum limit of lead after which user will relinquishes its slot is given by max_l i = RT i TTIduration mean_channal_rate i (7) This model is also responsible for selecting user after relinquishing. Among the lagging users who perceives good channel will be scheduled.

14 Proposed Solution Approach Wireless Fair - High Speed Scheduling Contd... Lead and Lag Model: if user i gives its slot to user j then lead-lag parameter is updated with ( ) PacketSize i l i = l i max TTIduration, (8) RT i l j = l j + PacketSize j RT j, if not lagging ( ) min., l j + PacketSize j RT j, if lagging (9)

15 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

16 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

17 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

18 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

19 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

20 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

21 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

22 Performance Measure Performance Measure Link Utilization: Measured by Mean System Throughput (MAC-hs) User Level Fairness Long-term Fairness: Measured by FI by [6] Here x i is taken as R i RT i ( N ) 2 i=1 x i N N i=1 x i 2, x i, i (1) and N is the number of users Short-term Fairness: For short-term fairness CDF of waiting time at Node B is used.

23 Simulation Environment Simulation Environment Simulation Tools Network Simulator-2 [7] and Enhanced UMTS Radio Access Network Extensions to NS2 [8] Simulation Topology EURANE Support UE 1 UE 2 UE N NODE B RNC Mb 15 ms SGSN Mb.4 ms GGSN Mb 1 ms Mb 15 ms node 1 node 2 node N Figure: Topology used in simulation

24 Simulation Environment Simulation Environment Simulation Parameters Parameters Value Cell Radius 6 m Number of codes used 15 Link Adaptation Delay 6 ms HARQ Cycle 6 TTI Node B Transmission Power 38 dbm Node B Antenna Gain 17 db Shadowing Standard Deviation 3 db Max. number of HARQ transmission 3 Number of parallel HARQ processes 6 RLC Acknowledgement Mode AM RLC Buffer Level 5 PDU s MAC Buffer Level 25 PDU s

25 Simulation Results Simulation Results Simulation Scenarios Web Traffic Scenario Average Load Distribution Scenario - Users are uniformly distributed in the cell area Worst Load Distribution Scenario - High requirement users are in bad channel condition FTP-Web Traffic Scenario Load Analysis

26 Simulation Results Web-Traffic Scenario Simulation Results - Web-Traffic Pedestrian Environments - Users are uniformly distributed in the cell area System Mean MAC-hs Throughput Achieved (kbps) APF Throughput WS-HSS Throughput APF FI WF-HSS FI Fairness Index System Mean MAC-hs Throughput Achieved / Requirements of user APF Ratio WF-HSS Ratio Algorithms ue1 ue2 ue3 ue4 ue5 ue6 ue7 ue8 ue9 ue1ue11ue12ue13ue14ue15ue16ue17ue18 Users (a) System Parameters (b) Users Analysis Figure: System & Users Analysis

27 Simulation Results Web-Traffic Scenario Simulation Results - Web-Traffic Pedestrian Environments - Users are uniformly distributed in the cell area CDF of MAC delay.5 u1 u2 u3 u4 u5 u6 u7 u8 CDF of MAC delay.5 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u9 u1.25 u11 u12.25 u11 u12 u13 u14 u13 u14 u15 u16 u15 u16 u17 u18 u17 u MAC delay (s) MAC delay (s) (a) APF (b) WF-HSS Figure: CDF of MAC-hs delay.

28 Simulation Results Web-Traffic Scenario Simulation Results - Web-Traffic Pedestrian Environments - High requirement users are in bad channel condition System Mean MAC-hs Throughput Achieved (kbps) APF Throughput WS-HSS Throughput APF FI WF-HSS FI Fairness Index System Mean MAC-hs Throughput Achieved / Requirements of user APF Ratio WF-HSS Ratio Algorithms ue1 ue2 ue3 ue4 ue5 ue6 ue7 ue8 ue9 ue1ue11ue12ue13ue14ue15ue16ue17ue18 Users (a) System Parameters (b) Users Analysis Figure: System & Users Analysis

29 Simulation Results Web-Traffic Scenario Simulation Results - Web-Traffic Pedestrian Environments - High requirement users are in bad channel condition CDF of MAC delay.5 u1 u2 u3 u4 u5 u6 u7 u8 CDF of MAC delay.5 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u9 u1.25 u11 u12.25 u11 u12 u13 u14 u13 u14 u15 u16 u15 u16 u17 u18 u17 u MAC delay (s) MAC delay (s) (a) APF (b) WF-HSS Figure: CDF of MAC-hs delay.

30 Simulation Results FTP-Web Traffic Scenario Simulation Results - FTP-Web Traffic Scenario Pedestrian Environments 2 APF Throughput WS-HSS Throughput APF FI WF-HSS FI 1.2 System Mean MAC-hs Throughput Achieved (kbps) Fairness Index Algorithms Figure: Mean System Throughput (MAC-hs) Achieved and Fairness Index.

31 Simulation Results FTP-Web Traffic Scenario Simulation Results - FTP-Web Traffic Scenario Pedestrian Environments System Mean MAC-hs Throughput Achieved / Requirements of user Web traffic APF Ratio Web traffic WF-HSS Ratio System Mean MAC-hs Throughput Achieved FTP APF throughput FTP WF-HSS throughput ue2 ue4 ue6 ue8 ue1 ue12 ue13 ue15 ue1 ue3 ue5 ue7 ue9 ue11 ue14 (a) R i RT i Users for Web traffic (b) MAC-hs Throughput Achieved for FTP users Users Figure: Users analysis

32 Simulation Results FTP-Web Traffic Scenario Simulation Results - FTP-Web Traffic Scenario Pedestrian Environments CDF of MAC delay.5 u1 u2 u3 u4 CDF of MAC delay.5 u1 u2 u3 u4 u5 u6 u5 u6.25 u7 u8 u9 u1.25 u7 u8 u9 u1 u11 u12 u11 u12 u13 u14 u13 u14 u15 u MAC delay (s) MAC delay (s) (a) APF (b) WF-HSS Figure: CDF of MAC-hs delay

33 Simulation Results Load Analysis Simulation Results - Load Analysis Pedestrian Environment 2 APF MAC-hs Throughput WF-HSS MAC-hs Throughput 1.2 APF WF-HSS 1 System Mean Throughput Achieved (kbps) Fairness Index (%) Number of Users (a) Mean System Throughput(MAChs) Number of Users (b) Fairness Index Figure: System Parameters by varying load on system End to End Throughput Vehicular Results

34 Summary & Conclusion Summary In averagely distributed load, both APF and WF-HSS algorithm provides good balance between link utilization and long-term fairness. In worst load conditions, WF-HSS provide better performance in terms of link utilization in comparison to APF algorithm. CDF of waiting time at Node B shows that WF-HSS algorithm provides short-term fairness among users for all traffic scenarios, while APF algorithm fails to do so.

35 Summary & Conclusion Summary Contd... WF-HSS scheduling provides benefits of TCP window resizing to the FTP users without affecting performance of other users, while APF algorithm doesn t do so. In comparison to APF algorithm, WF-HSS provides better link utilization without affecting fairness among users in heavy load conditions.

36 Summary & Conclusion Conclusion WF-HSS algorithm ensures long as well as short time scale fairness among users, according to their QoS requirements, while seeking to maximize link utilization.

37 Thank You References: Li-Chun Wang and Ming-Chi Chen. Comparisons of link-adaptation-based Scheduling algorithms for the WCDMA system with high-speed downlink packet access. Canadian Journal of Electrical and Computer Engineering (CJECE), Vol. 29:19 116, 24. R. Pankaj A. Jalali, R. Padovani. Data Throughput of CDMA-HDR a High Efficiency-High Data Rate Personal Communication Wireress System. IEEE, VTC, 2. J.M. Holtzman. Asymptotic Analysis of Proportional Fair Algorithm. in Proc. PIMRC, pages 33 37, 21. Ghassane Aniba and Sonia Aissa. Adaptive Proportional Fairness for Packet Scheduling in HSDPA. IEEE Communications Society Globecom, 24. Vaduvur Bharghavan Thyagarajan Nandagopal, Songwu Lu. A Unified Architecture for the Design and Evaluation of Wireless Fair-Queueing Algorithms. Wireless Networks, v.8, 22. Rajendra K Jain, Dah-Ming W Chiu, and William R Hawe. A Quantitative Measure of Fairness and Discrimination for Resource Allocation in Shared Systems Network Simulator ns-allinone 2.28 and its documentation is available on EURANE patch for clean ns 2.28 and its documentation is available on wmc.nl/eurane/.

38 Thank You Thank You

39 Thank You Backup Slides

40 Thank You Simulation Results - Web-Traffic Vehicular Environments - Users are uniformly distributed in the cell area System Mean MAC-hs Throughput Achieved (kbps) APF Throughput WS-HSS Throughput APF FI WF-HSS FI Fairness Index System Mean MAC-hs Throughput Achieved / Requirements of user APF Ratio WF-HSS Ratio Algorithms (a) System Parameters ue1 ue2 ue3 ue4 ue5 ue6 ue7 ue8 ue9 ue1ue11ue12ue13ue14ue15ue16ue17ue18 Users (b) Users Analysis Figure: System & Users Analysis

41 Thank You Simulation Results - Web-Traffic Vehicular Environments - Users are uniformly distributed in the cell area CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u15 u16 u17 u18 CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u15 u16 u17 u MAC delay (s) (a) APF MAC delay (s) (b) WF-HSS Figure: CDF of MAC-hs delay.

42 Thank You Simulation Results - Web-Traffic Vehicular Environments - High requirement users are in bad channel condition System Mean MAC-hs Throughput Achieved (kbps) APF Throughput WS-HSS Throughput APF FI WF-HSS FI Fairness Index System Mean MAC-hs Throughput Achieved / Requirements of user APF Ratio WF-HSS Ratio Algorithms (a) System Parameters ue1 ue2 ue3 ue4 ue5 ue6 ue7 ue8 ue9 ue1ue11ue12ue13ue14ue15ue16ue17ue18 Users (b) Users Analysis Figure: System & Users Analysis

43 Thank You Simulation Results - Web-Traffic Vehicular Environments - High requirement users are in bad channel condition CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u15 u16 u17 u18 CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u15 u16 u17 u MAC delay (s) (a) APF MAC delay (s) (b) WF-HSS Figure: CDF of MAC-hs delay.

44 Thank You Simulation Results - FTP-Web Traffic Scenario Vehicular Environments APF Throughput WS-HSS Throughput APF FI WF-HSS FI 1.2 System Mean MAC-hs Throughput Achieved (kbps) Fairness Index Algorithms Figure: Mean System Throughput (MAC-hs) Achieved and Fairness Index.

45 Thank You Simulation Results - FTP-Web Traffic Scenario Vehicular Environments System Mean MAC-hs Throughput Achieved / Requirements of user Web traffic APF Ratio Web traffic WF-HSS Ratio System Mean MAC-hs Throughput Achieved FTP APF throughput FTP WF-HSS throughput ue2 ue4 ue6 ue8 ue1 ue12 ue13 ue15 ue1 ue3 ue5 ue7 ue9 ue11 ue14 (a) R i RT i Users for Web traffic (b) MAC-hs Throughput Achieved for FTP users Users Figure: Users analysis

46 Thank You Simulation Results - FTP-Web Traffic Scenario Vehicular Environments CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u15 CDF of MAC delay.5.25 u1 u2 u3 u4 u5 u6 u7 u8 u9 u1 u11 u12 u13 u14 u MAC delay (s) (a) APF MAC delay (s) (b) WF-HSS Figure: CDF of MAC-hs delay

47 Thank You Simulation Results - Load Analysis Vehicular Environment 2 APF MAC-hs Throughput WF-HSS MAC-hs Throughput 1.2 APF WF-HSS 1 System Mean Throughput Achieved (kbps) Fairness Index (%) Number of Users (a) Mean System Throughput(MAC-hs) Number of Users (b) Fairness Index Figure: System Parameters by varying load on system Back to Pedestrian Results

48 Thank You Simulation Results - Load Analysis Pedestrian Environments 2 APF MAC-hs Throughput WF-HSS MAC-hs Throughput APF End-to-End Throughput WF-HSS End-to-End Throughput System Mean Throughput Achieved (kbps) Number of Users Figure: Mean System Throughput (MAC-hs) Achieved and Fairness Index. Back to Pedestrian Results