1. Multicast Polling and Efficient VoIP Connections in IEEE 802.16 Networks Olli Alanen Telecommunication Laboratory Department of Mathematical Information Technology University of Jyvaskyla, FINLAND ACM MSWiM 2007

2. 2 Outline Introduction Contention resolution process Multicast polling Simulation Conclusions

3. 3 Introduction

4. 4 UGS rtPS ertPS nrtPS BE

5. 5 Contention resolution process 802.16-2004 and 802.16e BE nrtPS ertPS Contention resolution mechanism when the connections have some packets in the uplink buffers, and no slots are allocated for them

6. 6 Contention resolution process The worst case MAC delay no downlink part C1C1 TfTf 2‧Tf2‧Tf

7. 7 Contention resolution process The worst case MAC delay

8. 8 Contention resolution process The worst case MAC delay for the packet is: Where i equals the last possible frame where contention resolution can occur. FPS is the Frames per second value and it is statically set at the base station.

9. 9 Contention resolution process In a particular case where backoff start and backoff end are set to equal where B is the value of the backoff start and end. the number of request opportunities per frame ( N )

10. 10 Multicast polling Each SS can belong to one or zero polling groups The memberships are managed dynamically Two management messages MCA-REQ and MCA-RSP

11. 11 Multicast polling When there is insufficient bandwidth for unicast polling multicast polling Multicast polling is only allowed to be used with nrtPS ertPS BE

12. 12 Multicast polling The decision about the sizes of the polling groups is also important when setting them up the probability p for successful transmission for an SS: N : the number of connections W : the size of the backoff window

13. 13 Simulation NS-2 simulator Packing Fragmentation management signalling ARQ initial ranging 5 service classe Scheduling dynamic service flow management contention resolution process PHY layer OFDM and OFDMa PHY layers

14. 14 Simulation

15. 15 Simulation scenario 1

16. 16 Simulation scenario 1

17. 17 Simulation scenario 1 I: VoIP connections:10 BE connections:0 Unicast polling  How much bandwidth do the VoIP connections take

18. 18 Simulation scenario 1 II: BE connections:50 VoIP connections: unicast polling with 1 polling slot in ertPS

19. 19 Simulation scenario 1 III: VoIP connections: exclude ertPS polling  regular broadcast contention mechanism The backoff parameters are adjusted to minimize the packet loss, with the price of lost delay guarantees

20. 20 Simulation scenario 1 IV: Backoff parameters are changed to serve smaller MAC delays with the cost of decreased utilization

21. 21 Simulation scenario 1 V: VoIP connections: broadcast  Multicast Polling backoff parameters are now adjusted to guarantee 15ms MAC delays

22. 22 Simulation scenario 1 VI: MAC delay of 20ms is now guaranteed

23. 23 Simulation scenario 1

24. 24 Simulation scenario 2 24 VoIP users

25. 25 Simulation scenario 2 smaller group equals to lesser collisions less packet loss the size of the UL-MAP↓

26. 26 Simulation scenario 2

27. 27 Simulation scenario 3

28. 28 Simulation scenario 3

29. 29 Simulation scenario 3

30. 30 Conclusions How multicast polling groups can be used to achieve a better quality of service This paper proposition is to use separate multicast polling groups for service classes with different QoS requirements set the backoff parameters based on the delay and loss targets Different delay requirements can also be guaranteed with multicast polling

31. 31 T hank you Who am I ? I am HuHu!