Handling Interference in Wireless Networks Ma Yongsen Weekly Meeting, January 3, 2013 Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 1 / 35 

Outline 1 Introduction 2 Multiuser MIMO 3 Full-Duplex Wireless 4 Extremely High Frequency Wireless Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 2 / 35 

Introduction Interference in Wireless Networks Even after 15 years of standardization of IEEE 802.11, interference is still a tough task for wireless networking. Furthermore, interference constricts the efficiency of resource allocation. Finally, it brings great challenging to the development of emerging technologies including: 1 Multiuser MIMO 2 Full-Duplex Wireless 3 Extremely High Frequency Wireless So the problem is how to handle interference in wireless networks? Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 3 / 35 

Multiuser MIMO1 Mul)user  MIMO  Allows  Concurrent  Streams   Alice   Bob   Chris   David   Clients  contend  for  concurrent  transmissions   opportuniIes  [K.  Tan,MOBICOM’09] 1Wei-Liang Shen et al. Rate adaptation for 802.11 multiuser mimo networks. MOBICOM 2012 Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 4 / 35 

Multiuser MIMO TradiIonal  historical-­‐based  rate  adaptaIon   doesn’t  work   à  Why? A  Client  Could  Transmit  Concurrently     With  Different  Clients Bob AP Chris Alice AP Chris Alice Bob Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 5 / 35 

Multiuser MIMO How  Concurrent  Clients  Interact?   Decode  Chris  by  Zero-­‐Forcing  (ZF)   •  Project  orthogonal  to  Alice   Decode  Alice  by  Successive  Interference  CancelaIon  (SIC)     •  Remove  Chris’s  interfering  signal   AP Chris Alice Bob xa   xc   hc   ha   antenna  1   antenna  2   hc xc   ha xa   How  Does  the  AP  Decode  Streams?   ZF-­‐SIC   Alice  can  only  be  decoded  if  Chris  is  decoded  correctly   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 6 / 35 

Multiuser MIMO How  Concurrent  Clients  Interact?   AP Chris Alice Bob xa   xc   hc   ha   SNR  reducIon  a]er  projecIon  à  SNRproj  =  SNRorig  *  sin2(θ)   x'c    =  xc sinθ   antenna  1   antenna  2   hc xc   θ   ha xa   How  Does  the  AP  Decode  Streams?   ZF-­‐SIC   Decode  Chris  by  Zero-­‐Forcing  (ZF)   •  Project  orthogonal  to  Alice   Decode  Alice  by  Successive  Interference  CancelaIon  (SIC)     •  Remove  Chris’s  interfering  signal   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 7 / 35 

Multiuser MIMO 1.  Each  Client  Learns  Its  SNRorig   AP Chris Alice Bob •  Leverage  channel  reciprocity   –  Forward  channel  =  Reverse  channel   •  EsImate  the  uplink  channel  from  AP’s  beacons   –  Compute  the  direcIon  and  amplitude  (SNRorig )   antenna  1   antenna  2   hc   hd   hu   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 8 / 35 

Multiuser MIMO Say  Chris  wants  to  learn  the  angle  between  two  signals   •  Learn  his  own  channel  à  by  reciprocity   •  Learn  the  direcIon  of  ongoing  streams  à  broadcast  by  Alice   2.  Learn  Inter-­‐Client  Angle  (θ)   antenna  1   antenna  2   hc   ha   AP Chris Alice Bob θ   SNRproj  can  be  predicted  by  SNRorig  and  inter-­‐client  angle   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 9 / 35 

Multiuser MIMO Doesn’t  Work  for  More  Than  2  Antennas   AP Chris Alice Bob •  Both  Bob  and  Chris  can  overhear  Alice’s  direcIon   •  Chris  cannot  decode  Bob  in  the  presence  of  Alice’s  signal     Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 10 / 35 

Multiuser MIMO How  to  Extend  to  More  Than  2  Antennas   Chris                                                      data  1   dir                                data  2   dir   CSI   send  nulls   data  3   Bob   tnull   Alice   t   contenIon  window     &  preamble   Chris Alice Bob Ongoing  transmifers     pause  their  streams     when  later  contenIon  winners   announce  the  direcIons   AP Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 11 / 35 

Multiuser MIMO                              data  2   Bob                                                      data  1   Alice   dir   dir   CSI   send  nulls   Chris   data  3   tnull   t   •  Alice  wins  the  1st  contenIon   –  select  the  rate  based  on  SNRorig   –  broadcast  her  direcIon  and  pause  transmission  at  tnull   •  Bob  wins  the  2nd  contenIon   –  select  the  rate  based  on  SNRproj   –  broadcast  his  direcIon   •  Chris  wins  the  3rd  contenIon   –  select  the  rate  based  on  SNRproj     TurboRate’s  MAC   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 12 / 35 

Multiuser MIMO Accuracy  of  SNR  Es)ma)on   EsImated  SNR  closely  matches  the  actual  SNR -­‐10   -­‐5   0   5   10   15   20   25   30   0   10   20   30   40   50   60   70   SNR  aVer  projec)on  [dB] Loca)on  Index Actual  SNR  a]er  projecIon   EsImated  SNR  a]er  projecIon   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 13 / 35 

Multiuser MIMO Throughput  Gain  for  2-­‐Antenna  AP   0   0.2   0.4   0.6   0.8   1   0   5   10   15   20   25   30   35   40   45   CDFs   Total  throughput  in  the  2-­‐antenna  AP  scenario          MU-­‐MIMO  with  TurboRate          MU-­‐MIMO  w/o  TurboRate          exisIng  system   TuboRate  delivers  the  MU-­‐MIMO  gain     à  1.7x  over  802.11     Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 14 / 35 

Multiuser MIMO Throughput  Gain  for  3-­‐Antenna  AP 0   0.2   0.4   0.6   0.8   1   0   10   20   30   40   50   60   70   CDFs   Total  throughput  in  the  3-­‐antenna  AP  scenario          MU-­‐MIMO  with  TurboRate          MU-­‐MIMO  w/o  TurboRate          exisIng  system   TurboRate  improves  the  throughput     by  2.3x  over  802.11   Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 15 / 35 

Full-Duplex Wireless2 Current wireless radios • Time Division Duplexing 5 • Frequency Division Duplexing TX RX RX TX TX/RX TX/RX TX/RX TX/RX Node 1 Node 2 Node 1 Node 2 Timeslot 1 Timeslot 2 Frequency 1 Frequency 2 Node 1 Node 2 2Mayank Jain et al. Practical, real-time, full duplex wireless. MOBICOM 2011 Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 16 / 35 

Full-Duplex Wireless Single channel full-duplex ➔ Very strong self-interference: ~70dB for 802.11 8 TX RX RX TX Node 1 Node 2 Self- Interference Main idea: cancel self-interference Combine RF and digital techniques for cancellation Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 17 / 35 

Full-Duplex Wireless Cancellation using Phase Offset 15 Self- Interference Cancellation Signal ∑ Self- Interference Cancellation Signal ∑ Frequency dependent, narrowband Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 18 / 35 

Full-Duplex Wireless Cancellation using Signal Inversion 17 Self- Interference Cancellation Signal ∑ Self- Interference Cancellation Signal ∑ Frequency and bandwidth independent Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 19 / 35 

Full-Duplex Wireless Time 21 TX RX TX RF Frontend Attenuator and Delay Line Xt +Xt/2 -Xt/2 ∑ RX RF Frontend Xt +Xt/2 -Xt/2 BALUN Signal Inversion Cancellation Over the air attenuation and delay Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 20 / 35 

Full-Duplex Wireless Time 24 ~50dB cancellation at 20MHz bandwidth with balun vs ~38dB with phase offset cancellation. Significant improvement in wideband cancellation Lower is better Higher is better Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 21 / 35 

Full-Duplex Wireless • Need to match self-interference power and delay • Can’t use digital samples: Saturated ADC Adaptive RF Cancellation 30 Use RSSI as an indicator of self-interference TX RX Attenuation & Delay Wireless Receiver Wireless Transmitter RF Cancellation TX Signal Path RX Signal Path RF Reference Σ Balun RSSI Control Feedback Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 22 / 35 

Full-Duplex Wireless 46 Digital Interference Cancellation TX RX Attenuation & Delay RF ➔ Baseband ADC Baseband ➔ RF DAC Encoder Decoder Digital Interference Reference RF Cancellation TX Signal Path RX Signal Path RF Reference Σ FIR filter - RSSI Control Feedback Channel Estimate Balun Bringing It All Together Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 23 / 35 

Full-Duplex Wireless 47 Performance • ~73 dB cancellation • WLAN full-duplex: Yes, with reasonable antenna separation • Not enough for cellular full-duplex Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 24 / 35 

Full-Duplex Wireless 71 Access Point networks Multi-hop Networks Secure Networks[1,2] Cellular networks Cell Basestation Relay Full-duplex Networking [1] Gollakota et al. “They Can Hear Your Heartbeats: Non-Invasive Security for Implantable Medical Devices.”, in Sigcomm 2011. [2] Lee et al. “Secured Bilateral Rendezvous using Self-interference Cancellation in Wireless Networks”, in IFIP 2011. Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 25 / 35 

Extremely High Frequency Wireless3 Existing Work: Connecting Neighboring Racks • 60GHz flyways[1] address local traffic hotspots by connecting neighboring racks wirelessly 6 [1] Halperin, D., et al. Augmenting data center networks with multi-gigabit wireless links. In Proc. of SIGCOMM (2011) 3Xia Zhou, et al. Mirror mirror on the ceiling: flexible wireless links for data centers. SIGCOMM 2012. Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 26 / 35 

Extremely High Frequency Wireless Our Goal: Any-to-any Communication • Traffic hotspots can appear between any rack pair  Connect any rack pair wirelessly 7 Hard to do using existing 60GHz beamforming! Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 27 / 35 

Extremely High Frequency Wireless Challenge #1: Link Blockage • 60GHz transmissions are blocked by small obstacles (anything larger than 2.5mm!) • Confirmed by our testbed measurements – Signal strength dropped by 10-30dB – Up to 15-90% throughput loss • Must use multi-hop forwarding – Antenna rotation delay – Reduce throughput by at least half 8 C A B Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 28 / 35 

Extremely High Frequency Wireless Challenge #2: Radio Interference • Beam interferes with racks in its direction – Exacerbated by dense rack deployment – Signal leakage makes it worse • Verified via testbed measurements – A single link causes 15-20dB drop in signal quality for 15 nearby links • Links interfere with each other – Very few links can run concurrently – Put a hard limit on aggregate bandwidth 9 TX RX Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 29 / 35 

Extremely High Frequency Wireless 3D Beamforming Key Benefits ✔ No more link blockage ✔ Much smaller interference Connect racks by reflecting signal off the ceiling! A B C 2D 3D RX 12 Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 30 / 35 

Extremely High Frequency Wireless Benchmark #1: Link Connectivity Q1: Does reflection cause any energy loss? 15 Even cheap metal plate provides perfect reflection! Propagation path -80 -70 -60 -50 4 8 16 30 Received signal strength (dBm) Propagation path length (m) Direct path Reflected path Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 31 / 35 

Extremely High Frequency Wireless Benchmark #1: Link Connectivity 16 0 2 4 6 8 0 10 20 30 40 50 Data rate (Gbps) Link distance (m) 2D w/o blockage 3D (h=2m) 3D (h=3m) 2D Link distance Q2: How does longer propagation path impact data rate? Link distance Negligible data rate loss 3D Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 32 / 35 

Extremely High Frequency Wireless Benchmark #2: Interference Footprint • A transmitter (0,0) communicates with a receiver (2,0) • Measure the received energy at multiple locations 17 2D Energy Map 3D Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 33 / 35 

Extremely High Frequency Wireless 0 3 6 9 12 15 0 2 4 6 8 10 Signal degradation (dB) Alignment error (degree) Link distance = 3m Link distance = 10m Benchmark #3: Robustness to Alignment Errors • How does alignment accuracy impact signal strength? • Fine grain experiment – Measure received signal when antennas perfectly tuned – Measure signal strength while introducing artificial alignment errors at 1o increments 18 Reflector 1o Today’s rotators: 0.006o- 0.09o accuracy Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 34 / 35 

Handling Interference in Wireless Networks THANKS! Ma Yongsen (Shanghai Jiao Tong University) Handling Interference in Wireless Networks Weekly Meeting 35 / 35