1. Yuan-Cheng Lai and Yen-Hung Chen Department of Information Management National Taiwan University of Science and Technology AINA 2008 Accept rate: 2008 31% 2007 <25%

2. Outline  Introduction  Background  Related Works  Proposed Bandwidth Allocation Algorithm  Phase I  Phase II  Simulation Results  Conclusions

3. Introduction  Many bandwidth allocation algorithms for supporting QoS requirements were proposed in 802.11 and DOCSIS

4. Introduction  IEEE 802.16 standard defines four kinds of service classes  Unsolicited Grant Service (UGS)  Real-Time Polling Service (rtPS)  Non-Real-Time Polling Service (nrtPS)  Best Effort Service (BE)  The standard does not recommend any particular scheme in detail

5. Introduction  Goal  To satisfy each connection’s QoS requirement  significantly increase the system gootput

6. Background

7.  IEEE 802.16 service classes n

8. Related work  Sayenko  first satisfies each connection’s minimal bandwidth requirement with considering adopted modulation  equally allocates the remaining bandwidth to each connection  CSH  first determines the DL/UL bandwidth ratio according to the ratio of the DL and UL requested bandwidth  satisfies each connection’s minimal bandwidth requirement  finally allocates the remaining bandwidth to each connection conceptually based on WFQ n

9. Proposed Bandwidth Allocation Algorithm total capacity of the wireless link newly coming connection minimal required bandwidth of connection i set of existing connections

10. Proposed Bandwidth Allocation Algorithm  Phase I : adjustment of the DL/UL bandwidth ratio  Phase II : Bandwidth allocation to each connection

11. Phase I : adjustment of the DL/UL bandwidth ratio Proposed Bandwidth Allocation Algorithm (1)modify the emergent rtPS’s R min temporarily (2) adjust the DL/UL bandwidth ratio (3) satisfy the DL/UL minimal bandwidth requirement

12. Proposed Bandwidth Allocation Algorithm Phase I required bandwidth in the BW request of connection i number of frames per second queuing delay of the Head- of-Line (HOL) packet of connection i Maximum Latency of connection i B i *FPSR min

13. Proposed Bandwidth Allocation Algorithm Phase I total requested numbers of symbols for DL transmitted data size within one symbol and one subchannel for connection i according to its transmission rate frame Uplink Downlink

14. Proposed Bandwidth Allocation Algorithm Phase I total number of symbols in one frame frame Uplink Downlink frame UplinkDownlink

15. Proposed Bandwidth Allocation Algorithm Phase I

16. frame DownlinkUplink frame DownlinkUplink DownlinkUplink

17. Proposed Bandwidth Allocation Algorithm Phase I number of symbols of DL minimal required bandwidth frame DownlinkUplink 66

18. Phase II : Bandwidth allocation to each connection Proposed Bandwidth Allocation Algorithm (1)satisfy each connection’s Rmin (2)Allocate the remaining bandwidth to the connections with better channel quality (3)allocate the remaining bandwidth to the connections with unfulfilled bandwidth

19. Proposed Bandwidth Allocation Algorithm Phase II allocated bandwidth for connection i connection’s symbol size

20. Proposed Bandwidth Allocation Algorithm Phase II unfulfilled bandwidth are used to construct the alpha weight connection’s symbol size average packet loss ratio Unfulfilled bandwidth of the connection i

21. Proposed Bandwidth Allocation Algorithm Phase II

23. Simulation Results  Parameter  OFDMA with 20MHz is used  The amount of frames in one second is assumed to be 200  the numbers of subchannels in DL and in UL are set to 60 and 70 respectively  The number of symbols is set to 48 in one frame  packet size is set to 200 bytes

24. Simulation Results  Network topology

25. Simulation Results  Basic connection settings

26. Simulation Results  Connection settings in a multi-rate environment

27. Simulation Results  Simulation results in a multi-rate environment

28. Simulation Results  Goodput of each service class

29. Simulation Results  The effects of modulation and packet loss percentage upon goodput

30. Conclusions  In order to promote the throughput and gooput, CQQ dynamically modifies DL/UL bandwidth ratio to match DL/UL traffic ratio  The simulation results show that CQQ outperforms CSH and Sayenko on system throughput and goodput in all situations, and it also provides each connection’s QoS guarantee

31. Thank you