Virtual MISO Triggers in Wi-Fi-like Networks

Transcription

1 Virtual MISO Triggers in Wi-Fi-like Networks Oscar Bejarano Edward W. Knightly Thursday, April,

2 Signal Outage in Fading Channels Thursday, April,

3 Signal Outage in Fading Channels x Power Zero Throughput egion x Thresh S time Thursday, April,

4 Signal Outage in Fading Channels x Power Zero Throughput egion x Thresh S time x Power x Thresh S N time Thursday, April,

5 Signal Outage in Fading Channels x Power Zero Throughput egion x Thresh S time x Power x Thresh S N time How do we exploit this independence among different paths? Thursday, April,

6 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) TX X Symbol Period Symbol Period Thursday, April,

7 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) TX X S S Symbol Period Symbol Period Thursday, April,

8 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) S TX S X S S Symbol Period Symbol Period Thursday, April,

9 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) TX S S X S S Symbol Period Symbol Period Thursday, April,

10 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) TX X S S Symbol Period Symbol Period Thursday, April,

11 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) TX X S S S* -S* Symbol Period Symbol Period Thursday, April,

12 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) S* TX -S* X S S Symbol Period Symbol Period S* -S* Thursday, April,

13 Transmit Spatial Diversity Multiple transmit antennas (antenna array) separated in distance, transmit redundant signals Consider: No Channel State Infomation (CSI) at Transmitter Orthogonal Space-Time Block Codes (i.e., Alamouti Scheme) S* TX -S* X S S Symbol Period Symbol Period S* -S* Thursday, April,

14 For maximal transmit diversity gains, we require antenna separation difficult to achieve in mobile devices Thursday, April,

15 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX X Thursday, April,

16 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX TX X Thursday, April,

17 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX TX X Thursday, April,

18 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX X TX Thursday, April,

19 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX X TX Thursday, April,

20 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX X TX Thursday, April,

21 Virtual MISO (vmiso). System Model.. Distributed System.. Single-Antenna Nodes General Tx Diversity vmiso/cooperation TX X Use Alamouti Scheme TX Thursday, April,

22 vmiso Protocol Design Challenges Triggering vmiso - Establish a criteria and mechanism used to trigger vmiso transmissions, i.e., how and when? Cooperator/elay Selection - Select a neighboring user as cooperator Thursday, April, 7

23 vmiso Protocol Design Challenges Triggering vmiso - Establish a criteria and mechanism used to trigger vmiso transmissions, i.e., how and when? Cooperator/elay Selection - Select a neighboring user as cooperator Thursday, April, 8

24 Objective To develop a comprehensive understanding of vmiso in the context of WLANs: Gains for vmiso flow Implications on network graph due to increased spatial footprint Thursday, April, 9

25 Objective To develop a comprehensive understanding of vmiso in the context of WLANs: Gains for vmiso flow Implications on network graph due to increased spatial footprint In particular, to provide the understanding needed to design trigger policies that maximize throughput performance gains trigger policies when should the cooperator be used? 9 Thursday, April,

26 vmiso Triggers Thursday, April,

27 vmiso Triggers. Method employed to invoke a vmiso transmission: Thursday, April,

28 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission Thursday, April,

29 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel Thursday, April,

30 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel. Trigger policies to identify scenarios in which vmiso transmissions yield a net throughput gain (or loss) Thursday, April,

31 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel. Trigger policies to identify scenarios in which vmiso transmissions yield a net throughput gain (or loss) Thursday, April,

32 eactive S NC S C time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

33 eactive S NC S C time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

34 eactive S S Data CN C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

35 eactive S S Data CN C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

36 eactive S S Data CN C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

37 eactive S S Data CN C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

38 eactive S S Data x CN C x x x NACK time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

39 eactive S S Data x CN C x x x NACK time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

40 eactive S S Data x CN C x x x NACK time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

41 eactive S S Data x e-tx CN C x x x NACK Tx x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) If transmission fails, receiver replies with a NACK ) NACK triggers simultaneous coded retransmission (from Sender and Cooperator) eacts to Failure Thursday, April,

42 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel. Trigger policies to identify scenarios in which vmiso transmissions yield a net throughput gain (or loss) Thursday, April,

43 Proactive S NC S C time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

44 Proactive S NC S C time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

45 Proactive S S Tx NC C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

46 Proactive S S Tx NC C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

47 Proactive S S Tx NC C x x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

48 Proactive S S Tx e-tx NC C x x Tx x time ) Initial transmission - Sender to eceiver ) Neighboring node overhears transmission ) Simultaneous coded retransmission (from Sender and Cooperator) Prevents Failure Thursday, April,

49 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel We will show results only for eactive. Trigger policies to identify scenarios in which vmiso transmissions yield a net throughput gain (or loss) Thursday, April,

50 vmiso Triggers. Method employed to invoke a vmiso transmission:.. eactive (On-Demand) - eacts to a failed packet transmission.. Proactive - Prevent packet failures in an already known bad channel We will show results only for eactive. Trigger policies to identify scenarios in which vmiso transmissions yield a net throughput gain (or loss) Thursday, April,

51 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Fully Connected Identify when the cooperator should or should not be used Thursday, April, 7

52 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Fully Connected Identify when the cooperator should or should not be used Thursday, April, 8

53 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Identify when the cooperator should or should not be used Thursday, April, 8

54 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Identify when the cooperator should or should not be used Thursday, April, 8

55 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Identify when the cooperator should or should not be used Thursday, April, 8

56 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S C High throughput gains for F F S Little to no cooperator interference with F Identify when the cooperator should or should not be used Thursday, April, 8

57 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S High throughput gains for F C Little to no cooperator interference with F F S Identify when the cooperator should or should not be used Thursday, April, 9

58 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S High throughput gains for F C Little to no cooperator interference with F F S Identify when the cooperator should or should not be used Thursday, April, 9

59 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S High throughput gains for F C Little to no cooperator interference with F F S Identify when the cooperator should or should not be used Thursday, April, 9

60 Trigger Policies in WLANs Interference ange Cooperator Footprint Intended Data Tx F S Low High throughput gains for F C Little to no cooperator interference with F F S Identify when the cooperator should or should not be used Thursday, April, 9

61 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S F S Information Asymmetry Identify when the cooperator should or should not be used Thursday, April,

62 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S Collisions F S Information Asymmetry Identify when the cooperator should or should not be used Thursday, April,

63 Interference ange Cooperator Footprint Intended Data Tx Trigger Policies in WLANs F S Collisions C F S Information Asymmetry Identify when the cooperator should or should not be used Thursday, April,

64 Trigger Policies in WLANs Multi-Hop S N Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power Fewer Losses Lower vmiso Gains vmiso S C Identify when the cooperator should or should not be used Thursday, April,

65 Trigger Policies in WLANs Multi-Hop S N Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power Fewer Losses Lower vmiso Gains vmiso S C Identify when the cooperator should or should not be used Thursday, April,

66 Trigger Policies in WLANs Multi-Hop S N Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power Fewer Losses Lower vmiso Gains vmiso S C Identify when the cooperator should or should not be used Thursday, April,

67 Trigger Policies in WLANs Multi-Hop S N High Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power Fewer Losses Lower vmiso Gains vmiso S C Identify when the cooperator should or should not be used Thursday, April,

68 Trigger Policies in WLANs Multi-Hop S N High Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power More Fewer Losses Lower vmiso Gains vmiso S C Identify when the cooperator should or should not be used Thursday, April,

69 Trigger Policies in WLANs Multi-Hop S N High Low Tx ate Transmissions Over Shorter/Higher Quality Links VS Transmit Diversity or/and x Tx Power More Fewer Losses Lower vmiso Gains vmiso S C Higher Identify when the cooperator should or should not be used Thursday, April,

70 Evaluation oadmap 8 System Implementation Comprehensive vmiso Evaluation 8 Atomic Scenarios (Fundamental Small- Scale Topologies) Large-Scale Topologies (up to flows) Monday, December, Mbps vmiso Mbps Direct Tx. Mbps vmiso Mbps Direct Tx. Mbps vmiso 8Mbps Direct Tx. 8Mbps vmiso Number of Cooperative Number Packets of (x) Cooperative Pkts. (x) BPSK QPSK QAM QAM BPSK Mbps QPSK Mbps QAM Mbps QAM 8Mbps BPSK No Coop BPSK BPSK Coop Mbps Direct Tx. QPSK BPSK Mbps QPSK No Coop 8 QAM Mbps vmiso 8 QPSK Coop QAM QPSK Mbps 7 QAM No Coop Mbps Direct Tx. QAM Coop QAM 7 87 Mbps QAM Mbps No Coop vmiso QAM Coop QAM 8Mbps Mbps Direct Tx. No Direct Cooperation Tx. No Direct Cooperation Tx. Mbps Cooperation Cooperation vmiso NvMISO NvMISO 8Mbps Direct Tx Mbps vmiso 8 Perfect NACK NvMISO 8 NvMISO 8 7 Direct Tx Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Attenuation (db) F F (b) Attenuation (db) Attenuation (db) Flow Experiment Flow Experiment Flow Simulation Flow Simulation Flow -8 Experiment Flow -8 Experiment -8 Flow Simulation -8 Flow Simulation Flow Flow Flow Flow Flow F Flow Flow Flow F Experiment Simulation Experiment Simulation Attenuation (db) Attenuation (db) (a) Hidden Terminal (b) Information Asymmetry (a) (b) Throughput (Mbps) Throughput (Mbps) Throughput (Mbps) Monday, December, Throughput (kbps) Throughput (kbps) Tuesday, February 8, Throughput (Mbps) X Position (meters) X Position (meters) X Position (meters) X Position (meters) c) 7 Independent Flows 7 - Flow d) Independent 7 Flows - 8 Flow 8 Throughput Gain/Loss (%) Throughput Gain/Loss (%) Attenuation Attenuation (db) Wednesday, March 7, (db) Throughput (kbps) Throughput (kbps) Y Position (meters) Y Position (meters) 8 Number of Cooperative Pkts. (x) Number of of Cooperative Packets (x) Number of Cooperative Packets (x) X Position (meters) X Position (meters) a) Coupled Flows - Flow Flow Throughput Gain/Loss at d=7m (%) Flow Throughput Gain/Loss at d=7m (%) (%) 8 8 F F Node Index Y Position (meters) Y Position (meters) X Position (meters) X Position (meters) b) Coupled Flows - Flow Flow Throughput Gain/Loss at d=7m (%) Flow Throughput Gain/Loss at (%) d=7m (%) 8 8 F F Thursday, April,

71 8 Evaluation oadmap System Implementation Comprehensive vmiso Evaluation 8 Atomic Scenarios (Fundamental Small- Scale Topologies) Large-Scale Topologies (up to flows) Monday, December, Mbps vmiso Mbps Direct Tx. Mbps vmiso Mbps Direct Tx. Mbps vmiso 8Mbps Direct Tx. 8Mbps vmiso Number of Cooperative Number Packets of (x) Cooperative Pkts. (x) BPSK QPSK QAM QAM BPSK Mbps QPSK Mbps QAM Mbps QAM 8Mbps BPSK No Coop BPSK BPSK Coop Mbps Direct Tx. QPSK BPSK Mbps QPSK No Coop 8 QAM Mbps vmiso 8 QPSK Coop QAM QPSK Mbps 7 QAM No Coop Mbps Direct Tx. QAM Coop QAM 7 87 Mbps QAM Mbps No Coop vmiso QAM Coop QAM 8Mbps Mbps Direct Tx. No Direct Cooperation Tx. No Direct Cooperation Tx. Mbps Cooperation Cooperation vmiso NvMISO NvMISO 8Mbps Direct Tx Mbps vmiso 8 Perfect NACK NvMISO 8 NvMISO 8 7 Direct Tx Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Flow Throughput Gain/Loss at d=m (%) Attenuation (db) F F (b) Attenuation (db) Attenuation (db) Flow Experiment Flow Experiment Flow Simulation Flow Simulation Flow -8 Experiment Flow -8 Experiment -8 Flow Simulation -8 Flow Simulation Flow Flow Flow Flow Flow F Flow Flow Flow F Experiment Simulation Experiment Simulation Attenuation (db) Attenuation (db) (a) Hidden Terminal (b) Information Asymmetry (a) (b) Throughput (Mbps) Throughput (Mbps) Throughput (Mbps) Monday, December, Throughput (kbps) Throughput (kbps) Tuesday, February 8, Throughput (Mbps) X Position (meters) X Position (meters) X Position (meters) X Position (meters) c) 7 Independent Flows 7 - Flow d) Independent 7 Flows - 8 Flow 8 Throughput Gain/Loss (%) Throughput Gain/Loss (%) Attenuation Attenuation (db) Wednesday, March 7, (db) Throughput (kbps) Throughput (kbps) Y Position (meters) Y Position (meters) 8 Number of Cooperative Pkts. (x) Number of of Cooperative Packets (x) Number of Cooperative Packets (x) X Position (meters) X Position (meters) a) Coupled Flows - Flow Flow Throughput Gain/Loss at d=7m (%) Flow Throughput Gain/Loss at d=7m (%) (%) However, our evaluation explored networks of up to flows 8 8 F F Node Index Y Position (meters) Y Position (meters) X Position (meters) X Position (meters) b) Coupled Flows - Flow Flow Throughput Gain/Loss at d=7m (%) Flow Throughput Gain/Loss at (%) d=7m (%) 8 8 F F Thursday, April,

72 System Implementation Host PC SCIPT VISUALIZE CONTOLLE WAPnet Client WAPnet Client WAPnet Client WAPnet Server Combination of over-the-air experiments (small topologies) and simulation (large topologies) WAP Platform [] and WAPnet - Clean slate MAC and PHY B A EBE D C (a) m G J F (b) H I Simulations in NS- Performance Metric: Throughput (bps) Protocol Implementation: Idealized NACK-based (benchmarking) vs practical NACK-based scheme. [] Thursday, April,

73 vmiso in Atomic Scenarios Thursday, April,

74 vmiso in Atomic Scenarios S F D S F D S F Fully Connected D S F Information Asymmetry D 7 Thursday, April,

75 vmiso in Atomic Scenarios S F D S F Fully Connected D S F D S F Information Asymmetry D 7 Thursday, April,

76 S vmiso in Atomic Scenarios F D S F D S F Fully Connected D S F Information Asymmetry D 7 Thursday, April,

77 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D 7 Thursday, April,

78 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D. Throughput (Mbps)... Flow Flow Fully Connected 7 Thursday, April,

79 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D. Almost doubled throughput of flow Throughput (Mbps)... Flow Flow Fully Connected 7 Thursday, April,

80 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D Throughput (Mbps)... Almost doubled throughput of flow No negative effect of the cooperator on competing flow:. Flow Flow Fully Connected 7 Thursday, April,

81 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D Throughput (Mbps).... Flow Flow Almost doubled throughput of flow No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow Fully Connected 7 Thursday, April,

82 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D S F Information Asymmetry D. Almost doubled throughput of flow 9 8 Throughput (Mbps)... No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow Throughput (kbps) 7 Flow Flow Flow Flow Fully Connected 7 Information Asymmetry Thursday, April,

83 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso S F Fully Connected D As expected f>>f S F Information Asymmetry D. Almost doubled throughput of flow 9 8 Throughput (Mbps)... No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow Throughput (kbps) 7 Flow Flow Flow Flow Fully Connected 7 Information Asymmetry Thursday, April,

84 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso. S F Fully Connected D Almost doubled throughput of flow As expected f>>f Big increase (~%), but still unsatisfactory performance S F Information Asymmetry 9 8 D Throughput (Mbps)... No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow Throughput (kbps) 7 Flow Flow Flow Flow Thursday, April, Fully Connected 7 Information Asymmetry

85 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso Throughput (Mbps).... S F Fully Connected D Almost doubled throughput of flow No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow As expected f>>f Big increase (~%), but still unsatisfactory performance Why did it help? elay helps one link by providing further information about topology S F Information Asymmetry Throughput (kbps) D Flow Flow Flow Flow Thursday, April, Fully Connected 7 Information Asymmetry

86 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso Throughput (Mbps).... S F Fully Connected D Almost doubled throughput of flow No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow As expected f>>f Big increase (~%), but still unsatisfactory performance Why did it help? elay helps one link by providing further information about topology Why didn t it help that much? Low number of cooperative tx triggered S F Information Asymmetry Throughput (kbps) D Flow Flow Flow Flow Thursday, April, Fully Connected 7 Information Asymmetry

87 S F vmiso in Atomic Scenarios D S F D N C Direct Tx vmiso Throughput (Mbps).... S F Fully Connected Flow Flow Fully Connected D Almost doubled throughput of flow No negative effect of the cooperator on competing flow: Why? Noncooperative flow already deferring competing flow 7 As expected f>>f Big increase (~%), but still unsatisfactory performance Why did it help? elay helps one link by providing further information about topology Why didn t it help that much? Low number of cooperative tx triggered MAC behavior completely dominates the PHY behavior S F Information Asymmetry Throughput (kbps) D Flow Flow Information Asymmetry Thursday, April,

88 vmiso in Large-Scale Scenarios 8 Thursday, April,

89 vmiso in Large-Scale Scenarios. Aggregate effects observed in atomic scenarios. Complex interactions between nodes (specially between cooperator and neighboring nodes/flows) Thursday, April, 9

90 vmiso in Large-Scale Scenarios Inter-Packet Transmission Time (ms) Perfect NACK NvMISO 9 9 to Perfect Perfect flow NACK NACK networks, NvMISO random 8 position, static 8 topology Inter Packet Transmission Time (ms) Inter Packet Transmission Time (ms) 7 7 Aggregate effects due to the cooperator (increased transmission footprint) hinder gains attained by vmiso Direct Tx. Direct Tx. NvMISO Direct Tx. Tx. NvMISO Perfect NACK NvMISO NvMISO NvMISO Throughput Throughput Gain Gain (Mbps) (Mbps) Time seconds Time seconds Gains decrease from 7% ( flows) to approximately % ( flows) Throughput Gain Gain (Mbps) (Mbps) Perfect NACK NvMISO Perfect NACK NvMISO NvMISO NvMISO NvMISO NvMISO Managed Network Managed Network Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows (a) Time Time seconds seconds Time seconds Perfect Perfect Perfect NACK NACK NACK NvMISO NvMISO NvMISO NvMISO Trigger Managed Managed Threshold Network Trigger Threshold Network Policy Policy Flows Flows Flows Flows Flows (b) Thursday, April,

91 vmiso in Large-Scale Scenarios How do we diminish negative effects of vmiso? Thursday, April,

92 vmiso in Large-Scale Scenarios How do we diminish negative effects of vmiso? Thursday, April,

93 vmiso in Large-Scale Scenarios How do we diminish negative effects of vmiso? Establish a Network Wide Trigger Policy Thursday, April,

94 vmiso in Large-Scale Scenarios How do we diminish negative effects of vmiso? Establish a Network Wide Trigger Policy Thursday, April,

95 vmiso in Large-Scale Scenarios How do we diminish negative effects of vmiso? Establish a Network Wide Trigger Policy e.g., Local Approach: only flows achieving γ% gains allowed to trigger vmiso Thursday, April,

96 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso Direct Tx. Tx. NvMISO erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO ect Tx. MISO rfect NACK NvMISO Inter Packet Transmission Time (ms) Time seconds Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

97 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) Time seconds Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

98 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Time seconds Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

99 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Only flows achieving +% gains allowed to trigger vmiso (arbitrarily chosen) Time seconds Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

100 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Only flows achieving +% gains allowed to trigger vmiso (arbitrarily chosen) Time seconds Why it works? Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

101 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Only flows achieving +% gains allowed to trigger vmiso (arbitrarily chosen) Time seconds Why it works? educe aggressiveness Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

102 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Only flows achieving +% gains allowed to trigger vmiso (arbitrarily chosen) Time seconds Why it works? educe aggressiveness educe Footprint Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. (a) (b) Flows Flows Flows Flows Flows..... Thursday, April,

103 vmiso in Large-Scale Scenarios Time seconds irect Tx. vmiso erfect NACK Perfect NvMISO NACK NvMISO 9 9 Perfect NACK NvMISO Local Decisions 8 8 ect Tx. MISO Network-wide Direct Tx. trigger threshold Tx. NvMISO policy: rfect NACK NvMISO Inter Packet Transmission Time (ms) 7 7 Only flows achieving +% gains allowed to trigger vmiso (arbitrarily chosen) Time seconds Why it works? educe aggressiveness educe Footprint Throughput Flows Flows Flows Flows Flows Flows Flows Flows Flows Flows (a) Time seconds Time seconds Time seconds Perfect NACK Perfect Perfect NACK NACK NvMISO NvMISO NvMISO NvM Trigger Threshold Managed Policy Ne NvMISOPerfect NACK Perfect NACK NvMISO NvMISO Trigger Trigger Threshold NvMISO Managed Threshold Policy Network NetworP Trigger Managed Threshold Network Policy.. Flows.. Flows... Flows. Flows.. Flows.. Even a simple policy can be highly efficient in large networks (b) Flows Flows Flows Flows Flows..... Thursday, April,

104 Conclusion Objective: To develop a comprehensive understanding of vmiso in the context of WLANs that, leads to the design of trigger policies that maximize throughput performance gains Demonstrated that cooperation is able to achieve very high gains at atomic level scenarios However, the magnitude of these gains decrease at networkscale scenarios Nonetheless, simple trigger policies can have a significant positive impact on the performance of vmiso Thursday, April,

105 Thank You! Thursday, April,

106 Virtual MISO Triggers in Wi-Fi-like Networks Oscar Bejarano Edward W. Knightly Thursday, April,