Dual Wi-Fi: Dual Channel Wi-Fi for Congested WLANs with Asymmetric Traffic Loads

1. Adriana Flores, Rice University Dual Wi-Fi: Dual Channel Wi-Fi for Congested WLANs with Asymmetric Traffic Loads Authors: • Date: 2013-11-12

2. Adriana Flores, Rice University Motivation • Traffic Asymmetric • Downlink traffic >> Uplink Traffic http://netflix.com/movie

3. Adriana Flores, Rice University Motivation • Traffic Asymmetric • Downlink traffic >> Uplink Traffic • High Contention • High number of backlogged nodes competing for the same resources

4. Adriana Flores, Rice University Motivation • Traffic Asymmetric • Downlink traffic >> Uplink Traffic • High Contention • High number of backlogged nodes competing for the same resources • Hidden Terminals • Cause collisions • Spectrum Underutilization • Affects downlink

5. Adriana Flores, Rice University 802.11 in Congested WLANs with Traffic Asymmetry • Shared resources • Defer to one another transmissions • Performance dependency • Spectrum Underutilization (Coordination Time, Collisions) • E.g. Collisions by Hidden Terminals • Disproportionate contention • Uplink Data: many clients vs. Downlink Data: few APs • Same CWmin yields equal medium access probability • N backlogged Clients : • Uplink Data: N/(N+1) Downlink Data: 1/(N+1)

6. Adriana Flores, Rice University Goal • Define a random access MAC that provides configurable spectrum resources for upload vs. download traffic • Enables matching resources to demand • Enables high spectral efficiency

7. ↓ ACK AP  STA ↑ Data STA  AP Adriana Flores, Rice University ↑ ACK STA  AP ↓Data AP  STA 802.11 Channel Architecture Time Uplink Data-ACK Downlink Data-ACK Frequency Total Bandwidth

8. Adriana Flores, Rice University Dual Wi-Fi Channel Architecture ↓ ACK AP  STA ↑ACK STA  AP ↑ Data STA  AP ↓Data AP  STA Uplink Data Channel Downlink Data Channel Dual Wi-Fi Time … ↓ ACK AP  STA ↑ACK STA  AP E.g. Channel 36: 5.1 GHz Frequency E.g. Channel 165: 5.8 GHz ↑ Data STA  AP Downlink Data Channel ↓Data AP -> STA Uplink Data Channel FDD Time … Frequency

9. Adriana Flores, Rice University Features of Dual Wi-Fi Channel Architecture • Logical Division (direction of data) • Decouple medium access • Medium access directly weighted on the traffic load of that direction • Independent and asynchronous operation • Independent performance • Independent resource allocation • Flexible bandwidth division • Bi-directional traffic within channels • Support the complete MAC-layer Data-ACK handshake • In-channel control feedback • paired with transmitted data • Unlike FDD, no generic control messages use the channel

10. Adriana Flores, Rice University ↑ACK STA  AP Dual Wi-Fi Benefits ↓ ACK AP  STA ↑ACK STA  AP ↓Data AP  STA ↓ ACK AP  STA ↑ Data STA  AP ↓Data AP  STA Downlink Data Channel Uplink Data Channel ↑ Data STA  AP Time … Frequency E.g. Channel 165: 5.8 GHz E.g. Channel 36: 5.1 GHz • Match spectrum resources to traffic asymmetry • Contention asymmetry: remove uplink and downlink competition for the same spectrum resources • Reduce medium contention and collisions •  Increase spectral efficiency

11. Adriana Flores, Rice University Dual Wi-Fi MAC • Isolate downlink and uplink medium access • Dual Wi-Fi ensures APs do not contend with STAs • 802.11 CSMA basic access • Downlink Data Channel • Only same-channel APs • CW still necessary • CW size tune to # of in-channel APs • 1 AP: • Collision-Free • No Contention • Uplink Data Channel • Only STAs • Remove contention with heavy downlink traffic Smaller number of contending nodes per channel: ↓ Coordination time, collisions and retransmissions  Increased spectral efficiency

12. Adriana Flores, Rice University Dual Wi-Fi vs. EDCA variation • Identify downlink data traffic as high-priority traffic providing strict or partial priority to APs to access the medium • Advantage: • Counters traffic asymmetry with minimal protocol modifications • Disadvantages: • Issues of shared band: • Medium Access aggressiveness • Dependency in number STAs and load • Coupled Medium Access • Downlink transmissions must defer to uplink transmissions • Coupled Performance • Throughput fraction is dependent on the load • Lead to starvation • Collisions • No guaranteed resources provided to downlink data traffic

13. Adriana Flores, Rice University Dual Wi-Fi Node Architecture TCP/IP • Two radio approach • Clients and APs • Tx and Rx in each channel independently and asynchronously • Full Duplex • (Different frequencies) • Co-channel Interference • Guard Band • WiFi-NC :100 KHz Control Unit Data UL MAC Data DL MAC Data TX Data TX Data RX Data RX Data UL PHY Data DL PHY TX RF TX RF RX RF RX RF Switch Transceiver

14. Adriana Flores, Rice University Node Architecture Design Alternatives • Half-Duplex Clients • AP smart selection of downlink data transmissions • Transmit to clients which it is not currently receiving from • Single radio clients • Only Tx or Rx in a single channel at a time • Filter to select either channel • Dual radio clients • Only Tx or Rx in a single channel at a time • Only operate a single radio at a time • Avoid cross talk

15. Adriana Flores, Rice University Dual Wi-Fi Performance Gains 626% ! KEY: UL and DL Medium access isolation 14 to 32% 152% ~-30%

16. Adriana Flores, Rice University Impact of Contention Asymmetry DW: 1% Ideal 1.01 1.04 DL -25% 0.34 DL -40% 0.11

17. Adriana Flores, Rice University Conclusion • Spectrum independence between uplink and downlink MAC data traffic • Can provide performance that is proportional to imposed demand • Adaptable to any traffic asymmetry or network density • Flexible design that adapts to changes in actual usage • Applications • Efficient use of resources • Key solution to address congested scenarios • White Spaces – isolation of hidden terminals • Faster downlink data delivery – Traffic asymmetry