2. ISSUES IN WIRELESS MAC PROTOCOLS Mohit Virendra Peng Lin Vidhya Seran

3. OUTLINE MAC Fairness in Wireless Ad-Hoc Networks MAC Fairness in Wireless Cellular Networks Power Controlled Multiple Access Protocol for Wireless Ad-Hoc Networks

4. Why Mac Fairness? Mobile Stations share a common broadcast channel. Existing protocols cannot prevent the “Capture Effects” Hidden Terminal Problem and Exposed Terminal Problem. Tradeoff between fairness and channel utilization

5. Fairness:Various Approaches 1.DFWMAC (IEEE 802.11 std.) [1] 2.MACAW :Improvement over MACA (Multiple Access Collision Avoidance) [2] Distributed Fair Scheduling Flow-Graph Based Approach etc. Estimation Based Fair Medium Access: Improvement over earlier approaches. (based on MACAW and DFWMAC)

6. Brief Overview of MACAW Uses modified RTS-CTS-DS-DATA-ACK message exchange. Uses modified BEB algorithm(milder): Collision:Finc(x)=MIN[1.5x,BOmax] Success:Fdec(x)=MAX[x-1,BOmin] Per Stream fairness not Per Station (allocates bandwidth equally to streams and not stations) Results in 37% throughput improvement with 6% overhead addition over MACA.

7. Problems with MACAW: In above configuration when load increases to a certain degree,st3 captures channel and st2 suffers degradation in throughput Backoff Copy scheme works only when congestion is homogeneous

8. Estimation based Fair Medium Access: Notations: Ø(i) :A predefined fairshare that station i should receive W(i) :The actual throughput achieved by station i. L(i) :Station i’s offered load

9. Desirable properties Station i’s offered load to channel is less than capacity: W(i)=L(i) Station’s offered load> Capacity: each station should be able to get its fair share of the channel,i.e. prop. to Ø Thus ideally for i and j: W(i)/ Ø(i)=W(j)/ Ø(j)

10. Description We define Fairness Index (FI): FI=max{ұ i,j :max[W(i)/Ø(i),W(j)/Ø(j)] / min [W(i)/Ø(i),W(j)/Ø(j)] } Actual case: Abs(W(i)/ Ø(i)-W(j)/ Ø(j)) should be bounded by smallest value.  Our Goal: Design a dist MAC protocol that minimizes FI and achieves fairness

11. Description (contd.) Choice of Ø(i):(open research problem) Assumption here (no admission control) Ø(i) = 0.5 (regardless of neighbors) Ø(o)= 1- Ø(i)=0.5 (per station fairness) E.g. Station with two active links: Ø(i)/Ø(o) = Ø(i)/(1-Ø(i)) =2/1 Thus Ø(i) ~ 0.67 (per stream fairness)

12. Description (contd.) Back off Scheme Notations: W(ei):The estimated share of estimating station itself. W(eo):The estimated share of other stations. T(type):Time to transmit a packet of type type.

13. How Fair-Share Estimation algorithm works: Station i sees itself competing with a group of stations for channel access. Stations dynamically estimate what throughput “they” get and what throughput “others” get and adjust their contention window according to the FI. Station i estimates “others” bandwidth by looking at the packets in its vicinity FI(e)=(W(ei)/ Ø(i)) / (W(eo)/ Ø(o)) contd…..

14. The Fair Share Estimation Algorithm

15. HowFair….(contd):Adjustment of Contention Window:

16. How Fair…(contd.):Contention Window adjustment In Algorithm2, C is a constant to adjust adaptivity of the algorithm. Smaller C:more aggressively contention window adjusted. C=2, possibility of collision high in high load and large no of competing stations C close to 1 (1.01), stations busy adjusting their contention windows all the time and algorithm becomes unstable.

17. Simulation and Results (NetWk Configs)

18. Results (contd..) (a) Station throughput (b)fairness index versus station’s offered load for the 4-station scenario.

19. Results (contd..) Station throughput (a)original algorithm (b) modified algorithm

20. Results (contd..) (c) fairness index versus station offered load for the 5 station scenario

21. Results (contd..) (a) Link throughput algorithm (b) link throughput (modified algorithm,Ø=0.5 for all)

22. Results (contd..) (c)link throughput (modified algorithm,Ø=0.67 f0r station 2,3 and 4) (d) FI versus station offered load for the 5-station scenario

23. Results (contd..) (a) Station throughput, (b) fairness index versus station’s offered load for the 6-station scenario.

24. Summary A different scheme for IEEE 802.11 DFWMAC Contention window adjustment according to the estimated share. Achieves far better fairness than others though some throughput sacrificed Does not assume any knowledge of network topology,thus does not require broadcast packets to disseminate info to other stations:very simple to overlay on existing DFWMAC.

25. OUTLINE MAC Fairness in Wireless Ad-Hoc Networks MAC Fairness in Wireless Cellular Networks Power Controlled Multiple Access Protocol for Wireless Ad-Hoc Networks

26. Wireless Fairness Scheduling Why we need wireless scheduling? 1. Provide short-term fairness 2. Provide short-term throughput bounds 3. Provide delay bounds for packets 4. Decouple delay/bandwidth requirements

27. CSDPS Channel state dependent packet scheduling algorithm, proposed by P. Bhagwat One step channel prediction and no compensation Lagging flows can only make up in long run

28. IWFQ Idealized wireless fair queueing algorithm, proposed by S. Lu, V. Bhaghavan and R. Srikant Lagging flows will capture the channel whenever they perceive clean channels

29. CIF-Queueing Channel independent fair queueing Algorithm, proposed by T.S. Ng, I. Stoica and H. Zhang Leading flows relinquish their leads linearly and distribute to lagging flows proportional to their weights

30. SBFA Server-based fairness approach, proposed by P. Ramanthan and P. Agrawal Statistically reserve a fraction of the bandwidth, no compensation

31. CBQ-CSDPS Class-based queueing with channel state dependent packet scheduling Maintain lead and lag based on the actual number of bytes transmitted during a time window Lagging flows are given explicit precedence, and hence capture the channel

32. Wireless Channel Characteristics Channel capacity is dynamically time- varying, due to fading/contention Channel errors are in nature location- dependent and bursty

33. Wireless Fair Service Scheduling Targets: 1. Short-term fairness among backlogged flows with clean channels. 2. Long-term fairness among backlogged flows with bounded channel error 3. Short-term throughput bounds for flows with clean channels 4. Long-term throughput bounds for flows with bounded channel error

34. Wireless Fair Service (Cont) Definitions 1. Error free service 2. Lead & Lag Model 3. Compensation Model 4. Slot queues & packet queues 5. Channel monitoring & prediction

35. WFS Service model

36. Error-free service model A reference for how much service a flow may receive in an ideal error-free channel environment WFQ is adopted as the error-free service model

37. Lead and lag model Three types of flows: 1. Leading flows: the flows which receive excess service 2. Lagging flows: the flows which relinquish slots due to expected channel errors 3. In-sync flows: the flows which follow the idealized service model

38. Compensation model Swapping slots between leading & lagging flows In-sync flows unaffected Gradually swapping to avoid the grabbing of the channel

39. Slot queues and packet queues Separate the logic packet flow queue and the MAC slot queue Packet flow queue may adopt any packet dropping policy Slot queue follows the swapping policy

40. Channel monitoring & prediction Channel errors are highly correlated One-step prediction: The channel state for the current time slot is predicted to be the same as the monitored channel state for the previous slot

41. Comparison of Wireless Scheduling Algorithms Scenario: Flow 1 is in error till t = 100 sec, Flow 2 & 3 are always error-free

42. Comparison of Wireless Scheduling Algorithms

43. Comparison of Wireless Scheduling Algorithms(cont)

45. Summary Several wireless fair scheduling algorithms have been proposed to address the fairness issues in wireless networks with time-varying capacity The performance of such wireless scheduling algorithms depends on the precision of channel monitoring/prediction methods

46. OUTLINE MAC Fairness in Wireless Ad-Hoc Networks MAC Fairness in Wireless Cellular Networks Power Controlled Multiple Access Protocol for Wireless Ad-Hoc Networks

47. MOTIVATION One of the major issue in wireless networks is developing efficient multiple access protocols that optimizes spectral reuse and hence maximize aggregate channel utilization. Theoretical studies have shown that ideal medium access protocols using optimal power can improve aggregate channel utilization. This motivates the study for power controlled wireless medium access protocols.

48. PAST WORK ON POWER CONTROL Past work on power control has primarily dealt with cellular networks and the base station provides centralized control. Distributed power control algorithms have also been presented but still require fundamental cellular configuration. Other work focused on MAC protocols that control transmission power level to conserve power consumption.

49. PCMA PCMA differs from the related work in two significant ways: A) Focus on wireless multiple access networks where all nodes share a channel and there is no centralized control. B) Focus on power control mechanism for increasing channel efficiency rather that as a mechanism for increasing battery life

50. PCMA-contd Dominant wireless MAC is IEEE802.11 standard follows the CSMA/CA paradigm. There exists no power control MAC that fits within the collision avoidance framework. Goal is to propose a power controlled MAC that follow the same collision avoidance framework.

51. PROBLEMS AND APPROACH TO THE SOLUTION MAC have made the case that a sender receiver pair should first acquire the floor before initiating a data packet transfer. Acquiring the floor allows sender-reciver pair to avoid collision due to hidden and exposed stations in the shared channel.

52. OPERATION

53. PROBLEMS AND APPROACH TO THE SOLUTION While acquring the floor to enable collision avoidance from hidden and exposed stations,this method preculdes multiple concurrent transmissions over the region of the acquired floor. To optimize spatial channel reuse in a shared wireless channel network, a pair of communicating nodes must only acquire the minimum area of the floor that is needed for it to successfully complete a data transmission.Figure(2) Unfortunately, it turns out that for collision avoidance mechanisms to work correctly, the control and data packets must be transmitted with a fixed power

54. Motivation for power controlled in collission avoidance-based medium access

55. PCMA Goal is to change the on/off fixed power transmission model to a more flexible bounded and variable power controlled transmission model. The fundamental change:unlike current protocols that use the reception of control packets as an on/off trigger for transmission/deferral by hidden and exposed stations,this approach uses the signal strength of a received control message to bound the transmission power of these stations.

56. PRINCIPLES Tow key principles: Power conserving principle: each station must transmit at the minimum power level that is required to be successfully heard by its intended receiver. Cooperation Principle : No station that commences a new transmission must transmit loud enough to disrupt ongoing transmissions.

57. NETWORK AND CHANNEL MODEL Channel Propagation Model The amount of spatial reuse and transmission power required for a node to send a valid signal to its destination will depend on the gain between each source and destination which models the attenuation of the transmitter power over distance. Gij actual gain is measured based on sender power and the receiver power.Then overcompensate to account for the distortions introduced from fading.

58. Channel propagation model Assumptions: 1.The data and busy tone channels observe similar gains. 2.Channel reciprocity hold so that the gain between two nodes is approximately the same in both in both directions. 3.The channel gain is stationary for the duration of the control packet and data packet transmissions.

59. NOTATIONS Pt_Max and Pt_Min-maximum and minimum transmission power for the transmitter for data channel respectively. Rx_Thresh and CS_Thresh-minimum received power for receiving a valid packet and for sensing a carrier respectively. SIR_Thresh –Capture threshold,minmum signal to interference ratio for which the reciver can successfully recive a packet. Pnj is the total noise that node j observes on the data channel.

60. POWER CONSTRAINTS

61. Power Constraints Critical Issues: A)handshaking between a transmitter –receiver pair to determine the minimum transmission power that satisfies constraints 2 and 3(power conserving principle) B) For every receiver to advertise its noise tolerance so that no potential transmitter will disrupt its ongoing reception applying constraint 4 (cooperative principle)

62. PCMA PROTOCOL The on/off model proposed is a bounded power model. Two Main mechanisms to achieve this model: A request power to send(RPTS)/acceptable power to send(APTS) handshake between the data sender and receiver-used to determine minimum transmission power. The noise tolerance advertisement is used by each active receiver to advertise the maximum additional noise power it can tolerate,given its current received signal and noise power levels. The packet handshake sequence on the data channel is RPTS-APTS-DATA-ACK.

63. PROTOCOL STEPS ij l APTS Step2 RPTS Step 1 DATA Step3 Send Busy tone Step 4 Step 5 ACK Step 6 Step 7

64. PROTOCOL STEPS Step1:Node I in its IDLE state monitors the busy tone to determine its power bound Pt_bound by measuring the maximum power received on the busy tone channel over a threshold time window. Step 2:Channel Gain is computed,Gij and the receiver then requires the data to be sent at Step 3: Source receives APTS packet and transmits the DATA at Pti_des on the data channel if the bound is satisfied Step 4: Receiver starts sending busy tone pulses on busy tone channel

65. Protocol Steps contd Step 5: When A node l receives the busy tone, it calculates its transmission power bound imposed by j Step 6: When the destination receives the entire data packet without errors, it sends an ACK. Step 7: if the source receives a valid ACK it resets the max back off and returns to the IDLE state,otherwise, it increase the max back off and starts over.

66. PERFORMANCE OF PCMA Parameter settings

67. Performance of PCMA to 802.11 and IPC 100 nodes in a 1000x 1000 meter network with 100 flows each sending 2KB packets and a connectivity range of 250 meters. X axis-Flow rate,Yaxis-Utilization

68. Performance of PCMA to 802.11 and IPC Throughput for a 100x100 meter network with 100 flows each sending 2Kb packets and a connectivity range of 250 meters.

69. Performance of PCMA to 802.11 and IPC Throughput for 100x100meter network with nodes separated into clusters regions

70. Performance of PCMA to 802.11 and IPC Destination range distribution for PCMA with Pt_max=Pt+4dB X axis-Range, Y axis-Fraction of Packets sent to range

71. Performance of PCMA to 802.11 and IPC Destination range distribution for PCMA with Pt_max=Pt+8dB

72. Performance of PCMA to 802.11 and IPC Throughput for different amounts of busy tone distortion with varying compensations in a 1000x1000 meter network.

73. CONCLUSION The performance results show that PCMA can achieve more than 2 times improvement in aggregate bandwidth compared to 802.11 for highly dense networks. When users communicate locally, the protocol provides improvements in throughput and increases scalability. PCMA is a protocol design in progress. Future work may include fairness properties of PCMA, performance under mobility and evaluating in a multihop wireless networks.

74. Reference Wireless Fair Scheduling: P. Bhagwat, P. Bhattacharya, A. Krishma and S. Tripathi, “Enhancing throughput over wireless LANs using channel state dependent packet scheduling”, IEEE INFOCOM’96 T.S. Ng, I. Stoica and H. Zhang, “Packet fair queueing algorithms for wireless networks with location-dependent errors”, IEEE INFOCOM’98 M. Srivastava, C. Fragouli and V. Sivaranan, “ Controlled multimedia wireless link sharing via enhanced classbased queueing with channel-state-dependent packet scheduling”, IEEE INFOCOM’98 P. Ramanathan and P. Agrawal, “Adapting packet fair queueing algorithms to wireless networks”, ACM MOBICOM’98 S. Lu, T. Nandagopal and V. Bharghavan, “Fair scheduling in wirless packet networks”, ACM MOBICOM’98

75. Reference PCMA: Jeffrey P. Monks, Vaduvur Bharghavan, and Wen-mei Hwu, "A Power Controlled Multiple Access Protocol for Wireless Packet Networks," IEEE INFOCOM 2001, Anchorage, Alaska, April, 2001 Jeffrey P. Monks, Vaduvur Bharghavan, and Wen-mei Hwu, "Transmission Power Controlled for Multiple Access Wireless Packet Networks," Proceedings of The 25th Annual IEEE Conference on Local Computer Networks (LCN 2000), Tampa, FL, Nov., 2000 Jeffrey P. Monks, Transmission Power Control for Enhancing The Performance of Wireless Packet Data Networks, PhD Thesis, Department of Electrical and Computer Engineering, University of Illinois, Urbana, IL, March, 2001

76. Reference Ad-Hoc Mac Fairness: DFWMAC: http://www.ietf.org/html.characters/manet-characters.htmlhttp://www.ietf.org/html.characters/manet-characters.html MACAW: A Media Access Protocol forWireless LAN's (1994): Vaduvur Bharghavan, Alan Demers, Scott Shenker, Lixia Zhang, 1994 SIGCOMM Conference Fair Medium Access in 802.11 based Wireless Ad-Hoc Networks:Brahim Bensaou,Yu Wang, Chi Chung Ko,Mobihoc 2000 Achieving MAC Layer Fairness in Wireless Packet Networks:Thyagrajan Nandgopal,Tae-Eun Kim,Xin Gao,Vaduvur Bhargavan,Mobicom 2000 A New Model for Packet Scheduling in Multihop Wireless Networks: Haiyun Luo, Songwu Lu,Vaduvur Bhargavan, Mobicom 2000