BBN Throughput Scaling in Dense Enterprise WLANs with
BBN: Throughput Scaling in Dense Enterprise WLANs with Blind Beamforming and Nulling Wenjie Zhou (Co-Primary Author), Tarun Bansal (Co-Primary Author), Prasun Sinha and Kannan Srinivasan The Ohio State University
Changes in Uplink Traffic Traditionally, WLAN traffic: • downlink heavy • less attention to uplink traffic Recently, uplink traffic increased rapidly : • mobile applications Cloud Computing Code Offloading Online Gaming Vo. IP, Video Chat Sensor Data Upload 2
Can we scale the uplink throughput with the number of clients?
Network MIMO Exchange raw samples AP 1 C 1 AP 2 C 2 AP 3 C 3 Huge bandwidth consumption
Mega. MIMO 1 Does not apply to uplink : Clients do not share a backbone network [1] Rahul, H. , Kumar, S. , and Katabi, D. Mega. MIMO: Scaling Wireless Capacity with User Demand. In Proc. of ACM SIGCOMM 2012.
Interference Alignment 1 • • C 1 AP 1 C 2 AP 2 C 3 AP 3 4 packets, 3 slots Enough time slots, everyone gets half the cake Exponential slots of transmissions, not suitable for mobile clients Heavy coordination between clients [1] Cadambe, V. R. , and Jafar, S. A. Interference Alignment and the Degrees of Freedom for the K User Interference Channel. IEEE Transactions on Information Theory (2008).
Existing interference alignment and beamforming techniques are not suitable to mobile uplink traffic. How can we bring the benefits of beamforming to uplink traffic?
AP Density in Enterprise WLANs 1 CDF 0. 75 0. 5 (140, 0. 5) 0. 25 0 50 100 150 Number of Access Points (APs) 200 BBN leverages the high density of access points 8
Omniscient TDMA Single Collision Domain AP 2 AP 3 AP 4 Switch AP 1 Time Slot: 3 1 2 x 1 x 2 C 1 x 3 C 2 C 3 Three Packets received in Three Slots Only one AP is in use 9
Blind Beamforming and Nulling Single Collision Domain h(1) 11 x 1 + h(1) 21 x 2 + h(1) 31 x 3 12 x 1 + AP 1 22 x 2 + h(1) 32 x 3 Time Slot: 1 AP 2 AP 3 Switch h(1)14 x 1 + h(1)24 x 2 + h(1)34 x 3 h(1)13 x 1 + h(1)23 x 2 + h(1)33 x 3 h(1) AP 4 13 x 1 C 1 h(1)33 23 x 2 C 2 x 3 C 3 10
Blind Beamforming and Nulling Single Collision Domain Receives: a 11 x 1 + s 1 h(1)21 x 2 + s 1 h(1)31 x 3 Receives: a 12 x 1 + a 22 x 2 + a 32 x 3 AP 2 AP 3 AP 4 Switch AP 1 Time Slot: 2 Transmits: v 3 (h(1)13 x 1 + h(1)23 x 2 + h(1)33 x 3) Transmits: v 4 (h(1)14 x 1 + h(1)24 x 2 + h(1)34 x 3) 11
Blind Beamforming and Nulling Single Collision Domain Slot 1: h(1)11 x 1 + h(1)21 x 2 + h(1)31 x 3 Slot 1: h(1)12 x 1 + h(1)22 x 2 + h(1)32 x 3 Slot 2: a 11 x 1 + s 1 h(1)21 x 2 + s 1 h(1)31 x 3 Slot 2: a 12 x 1 + a 22 x 2 + a 32 x 3 (s 1 h(1)11 -a 11)x 1 AP 2 AP 3 AP 4 Switch AP 1 Three Packets received in Two Slots 12
Number of APs Required • In a network with APs, APs in BBN can receive N uplink packets in two slots • 3 clients, 4 APs • 4 clients, 7 APs • 10 clients, 46 APs 13
Throughput Improvement • Previous Example Topology – APs in BBN receive three packets in two slots: an improvement of 50% • General Topology – Uplink throughput in BBN scales with the number of clients (N/2 packets per slot). – Half of the cake as in Interference Alignment • Always two slots • No coordination between clients 14
BBN Highlights • Leverages the high density of access points • All computation and design complexity shifted to APs • APs only need to exchange decoded packets over the backbone instead of raw samples 15
Further Optimizations to Improve SNR Receivers AP 2 AP 1 Switch Transmitters x 2, x 3 x 1 AP 4 AP 3 C 1 C 2 C 3 • Which subset of APs act as transmitters and which subset as receivers? • Which AP decodes which packet? BBN Approach: xi is decoded at the APj where it is expected to have highest SNR 16
Challenge 1/4: Synchronization of APs • To perform accurate beamforming, APs need to be tightly synchronized with each other • Solution: – Source. Sync (Rahul et al. , SIGCOMM 2010): synchronizes APs within a single collision domain – Vidyut (Yenamandra et al. , SIGCOMM 2014): uses power line to synchronize APs in the same building 17
Challenge 2/4 : Multi. Collision Domain • Not all APs may be able to hear each other directly • Solution: Make smaller groups where all APs in a single group can hear each other. 18
Distributed System Group Head • Within a group, all APs can hear each other • When one group is communicating, neighboring groups remain silent 19
Challenge 3/4 : Inconsistency in the AP density • Number of APs may be less than • Solution: Appropriate MAC layer algorithm that restricts the number of participating clients 20
MAC Timeline Uplink Keep Silent – Allow neighboring groups to transmit Time Slot 1 Approve A, B and C Notification Period Downlink Uplink Time Slot 2. . . . Poll Uplink . . . Time Compute pre-coding vectors in the background 21
Challenge 4/4 : Robustness • Nulling is not always perfect. Can’t Subtract x 1 Decoding Error x 1 , x 2 , x 3 AP 2 AP 4 AP 5 x 1 x 2 C 1 C 2 Switch AP 1 x 3 C 3 22
Challenge 4/4 : Robustness • What if we have extra APs x 1 , x 2 , x 3 AP 2 AP 3 AP 4 AP 5 Switch AP 1 x 1 x 2 C 1 C 2 AP 6 AP 7 x 3 C 3 23
Experiments Intended Signal = x 1 Interference from x 2, x 3 AP 2 AP 3 AP 4 USRP N 210 Switch AP 1 x 2, x 3 x 2 x 1 C 2 x 3 C 2 24
Throughput 1. 48 X BBN provides 1. 48 x throughput compared to TDMA 25
Trace-Driven Simulation • Over multiple collision domains (divided into groups) • Field Size: 500 m X 500 m • Number of clients: 1000 • Vary the number of APs • Residual interference distribution from experiment • Other algorithms simulated – Omniscient TDMA – IEEE 802. 11 26
Throughput BBN • • • 2000 APs 4. 6 X throughput gain ~76 APs near each client 27
Fairness BBN achieves higher fairness • Beamforming increased SINR of clients that are far away 28
Summary and Future Work • BBN leverages the high density of APs to scale the uplink throughput for single antenna systems – Throughput scales linearly with the number of clients – All computational and design complexity shifted to APs • Future Work – Coexist with legacy network – Data rate selection Thank you 29
Backup Slides 30
Long Term Results
Octo. Clock-G Frequency Accuracy w/ out : 25 ppb Frequency Accuracy with GPS Lock : <1 ppb PPS Accuracy with GPS Lock : 50 ns Vidyut : approximately 225 ns
Multiple Antenna AP • Assume each AP has K antennas • For N clients, APs required • For M APs, clients
Estimate SNR of C 1 at AP 2 AP 3 AP 4 Switch No path with high SNR AP 1 C 1 SNR of C 1 at AP 2 is low 34
Estimate SNR of C 1 at AP 1 AP 2 AP 3 AP 4 Switch One path with high SNR AP 1 C 1 • C 1 should. SNR be decoded by 1 is APhigh of C 1 at AP 1 • AP 1 should act as a receiver in slot 2 35
Blind Nulling in BBN 36
MAC Layer: Phase 1 Approve IACS AP 1 : C 2 : C 3 : IACS AC 1 Packet 1 AC 2 Packet 2 AC 3 Packet 3 37
MAC Layer: Phase 2 SIFS AP 1 : AP 2: AP 3 : SIFS BIFS AC 1 AC 2 AC 3 AP 4 : v 4* Samples 4 AP 5 : v 5* Samples 5 AP 6: v 6* Samples 6 AP 7 : v 7* Samples 7 38
Experiments Setup • Performed using USRP N 210 Radio • Testbed of 4 APs and 3 clients • Modulation Scheme: OFDM with BPSK • Channel: Central Frequency 400 Mhz, Bandwith set to 500 KHz 39
Existing Schemes • Interference Alignment – Existing IA schemes perform alignment over exponential number of time slots [Cadambe et al. , IEEE Transactions on Information Theory 2007] • MU-MIMO (Multi User MIMO) – Requires transmitters to exchange each other’s data before transmission • MU-MIMO (Multi User MIMO) in EWLAN – All APs together act as a single AP with multiple antennas – Requires APs to exchange samples over the backbone which is costprohibitive [Gollakota et al. , SIGCOMM 2009; Gowda et al. , INFOCOM 2013] 40
Existing Schemes • Interference Alignment – Existing IA schemes require each transmitter to transmit exponential amount of data [Cadambe et al. , IEEE Transactions on Information Theory 2007] • MU-MIMO – All APs together act as a single AP with multiple antennas – Requires APs to exchange samples over the backbone which is cost-prohibitive [Gollakota et al. , SIGCOMM 2009] 41
Related Work (contd. ) • Interference Alignment – Existing IA schemes work over exponential number of time slots [Cadambe et al. , IEEE Transactions on Information Theory 2007] – Or, work only for downlink [Suh et al. , IEEE Transactions on Communications 2011] – Or, require multiple antennas at clients [Gollakota et al. , SIGCOMM 2009] – Or, require APs to exchange samples over backbone [Annapureddy et al. , IEEE Transactions on Information Theory 2012] 42
Related Work • Backbone Usage – Mega. MIMO (Rahul et al. , SIGCOMM 2012): Works only for downlink – Symphony (Bansal et al. , Mobi. Com 2013): Works only in multiple collision domain 43
Related Work (contd. ) • Wireless Relays – Use special relay nodes to assist high speed communication between specific transmitters and receivers – Existing algorithms do not make use of the backbone – BBN leverages the backbone to improve throughput – BBN can extend to multiple rounds to decode packets with low SNR 44
BBN Highlights • Leverages the high density of access points • Uplink throughput scales with the number of clients in the network • All computational and design complexity shifted to APs • APs only need to exchange decoded packets over the backbone 45
Example Topology: What we ideally want x 1 x 2 AP 1 AP 2 x 3 AP 3 Switch Works! But can we make the requirements less strict? AP 4 AP 5 x 1 AP 6 x 2 C 1 C 2 AP 7 x 3 C 3 46
Matching in BBN AP 1 C 1 AP 2 Find the Maximum Weight Matching. AP • • 3 Edge Weight = C Which AP decodes which packet. AP 4 2 Expected SINR of Which APC 2 transmits in the second slot. at AP 3 AP 5 C 3 AP 6 AP 7 47
Number of APs Required: Example Topology x 1 • • x 2 x 3 Two packets (x 2 and x 3) need to be nulled at AP 1 One packet (x 3) needs to be nulled at AP 2 Three transmitting APs required Guarantee non degenerate solution: Four APs required 48
AP Density in Enterprise WLANs CDF of Number of APs Observed 1 CDF 0. 75 0. 5 Can 0. 25 we leverage the high density of APs to scale the uplink throughput? 0 0 50 100 150 Number of Access Points (APs) CDF of number of APs observed (Measurements conducted at Ohio State University campus) 200 49
Enterprise Wireless LAN Internet AP AP AP 50
BBN Overview • Leverages the high density of access points • Uplink throughput scales with the number of clients in the network – Schedule length: Two Slots • First slot: Clients transmit • Second slot: APs perform blind nulling – APs only need to exchange decoded packets over the backbone 51
Contents • • • BBN Design Experiments Simulations Challenges Related Work Conclusion 52
Example Topology (Single Collision Domain) with BBN Time Slot: 1 AP 2 AP 3 Switch AP 4 AP 5 x 1 AP 6 x 2 C 1 C 2 AP 7 x 3 C 3 53
Simultaneous Nulling Goal Receive: x 1, x 2 Null: x 3 Receive: x 1 Null: x 2, x 3 AP 1 AP 2 Receive: x 1, , x 2, x 3 AP 3 Switch AP 4 AP 5 x 1 AP 6 x 2 C 1 AP 7 x 3 C 2 C 3 54
Example Topology (Single Collision Domain) with BBN a 11 x 1 a 12 x 1 + a 22 x 2 AP 1 AP 2 AP 5 . . . AP 6 v 6 * (. . . ) v 5 * (h 15 x 1 + h 25 x 2 + x 2 h 35 x 3 ) x 1 C 2 . . . Switch v 4 * (h 14 x 1 + AP 4 h 24 x 2 + h 34 x 3 ) Time Slot: 2 1 AP 3 h 15 x 1 + h 25 x 2 + h 35 x 3 h 14 x 1 + h 24 x 2 + h 34 x 3 a 13 x 1 + a 23 x 2 + a 33 x 3 AP 7 v 7 * (. . . ) x 3 C 3 55
Example Topology (Single Collision Domain) with BBN a 11 x 1 a 12 x 1 + a 22 x 2 AP 1 AP 2 a 13 x 1 + a 23 x 2 + a 33 x 3 AP 3 Switch - a 12 x 1 = a 22 x 2 Time Slot: 2 - a 13 x 1 - a 23 x 2 = a 33 x 3 Three Packets received in Two Slots 56
Blind Nulling in BBN (Contd. ) • What we want: Blindly null x 2 and x 3 at AP 1; and, blindly null x 3 at AP 2 • Assuming nulling and packet cancellation is perfect, Compute V such that 57
MAC Layer Uplink Phase I Keep Silent – Allow neighboring groups to transmit Downlink Phase III. . . . Poll Approve A, B and C Contention Period Uplink . . . Time 58
RSS Without Blind Beamforming and Nulling Signal Strength at AP 1 Before Cancellation (in d. B) -15 -8 d. B SINR -25 -35 -45 -55 Interference Intended Signal 59
RSS After Blind Beamforming Signal Strength at AP 1 After Cancellation (in d. B) -15 13 d. B SINR -25 -35 • Proper precoding increases SINR by 21 d. B • Blind Beamforming and Nulling is practical -45 -55 Interference Intended Signal 60
How to boost uplink throughput? Interference Alignment? Smartphones are mobile MIMO? Smartphones are small netwrok-MIMO? High bandwidth consumption 61
- Slides: 61