RobinHood Sharing the Happiness in a Wireless Jungle Tarun Bansal Wenjie Zhou Kannan Srinivasan and Prasun Sinha Department of Computer Science and Engineering Ohio State University Columbus Ohio ID: 770803
Download Presentation The PPT/PDF document "RobinHood: Sharing the Happiness in a Wi..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
RobinHood: Sharing the Happiness in a Wireless Jungle Tarun Bansal , Wenjie Zhou, Kannan Srinivasan and Prasun Sinha Department of Computer Science and Engineering Ohio State University, Columbus, Ohio
Enterprise Wireless LAN (EWLAN) 2 AP AP AP AP AP AP
Uplink Traffic 3 Traditionally, uplink traffic has received less attention in the design of algorithms/solutions for WLANs Recently, uplink traffic has been increasing at a rapid pace due to increasing popularity of mobile applications such as: Cloud Computing Online Gaming Sensor Data Upload Code Offloading VoIP, Video Chat
Existing Schemes 4 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 transmissionMU-MIMO (Multi User MIMO) in EWLAN All APs together act as a single AP with multiple antennasRequires APs to exchange samples over the backbone which is cost-prohibitive [ Gollakota et al., SIGCOMM 2009; Gowda et al., INFOCOM 2013]
AP Density in Enterprise WLANs 5 CDF of number of APs observed (Measurements conducted at Ohio State University campus) Can we leverage the underutilized backbone and the high density of APs to scale the uplink throughput?
RobinHood Highlights 6 Leverages the high density of access points Uplink throughput scales with the number of clients in the network Schedule length: Two Slots First slot: Mobile clients transmit Second slot: APs perform Blind NullingAPs only need to exchange decoded packets over the backbone
C 1 C 2 C 3 x 1 x 2 x 3 AP 1 AP 2 AP 3 AP 4 AP 5 Switch AP 6 AP 7 Example Topology (Single Collision Domain) with Omniscient TDMA Time Slot: 1 Time Slot: 2 Time Slot: 3 Three Packets received in Three Slots. Only one AP is in use. 7
C 1 C 2 C 3 x 1 x 2 x 3 AP 1 AP 2 AP 3 AP 4 AP 5 Switch AP 6 AP 7 Example Topology (Single Collision Domain) with RobinHood Time Slot: 1 h 14 x 1 + h 24 x 2 + h 34 x 3 h 15 x 1 + h 25 x 2 + h 35 x 3 ... ... Time Slot: 2 v 4 * (h 14 x 1 + h 24 x 2 + h 34 x 3 ) v 5 * (h 15 x 1 + h 25 x 2 + h 35 x 3 ) v 6 * (...) v 7 * (...) a 11 x 1 a 12 x 1 + a 22 x 2 a 13 x 1 + a 23 x 2 + a 33 x 3 8
AP 1 AP 2 AP 3 Switch Example Topology (Single Collision Domain) with RobinHood Time Slot: 2 a 11 x 1 a 12 x 1 + a 22 x 2 a 13 x 1 + a23 x2 + a 33 x 3 - a 12 x 1 = a 22 x 2 - a 13 x 1 - a 23 x 2 = a 33 x 3 9 Three Packets received in Two Slots Time Slot: Background
Number of APs Required for Blind Nulling In a network with APs, APs in RobinHood can receive N uplink packets in two slots With M APs in a single collision domain, RobinHood provides uplink throughput of compared to O(1) for omniscient TDMA. Uplink throughput in RobinHood scales with the number of clients. 10
Further Optimizations to improve SNR Which subset of APs act as transmitters and which subset as receivers? Which AP decodes which packet? C 1 C 2 C 3 AP 1 AP 2 AP 3 AP 4 AP 5 Switch AP 6 AP 7 11 RobinHood Approach: x i is decoded at the AP where it is expected to have highest SNR Transmitters Receivers x 1 x 2 x 3
Example: Estimate SNR of C1 at AP1 SNR of C 1 at AP 1 is high C 1 AP 1 AP 2 AP 3 AP4 AP5 Switch AP 6 AP 7 12 One path available with high SNR
Example: Estimate SNR of C1 at AP3 SNR of C 1 at AP 3 is low C 1 AP 1 AP 2 AP 3 AP4 AP5 Switch AP 6 AP 7 13 No path available with high SNR C 1 should be decoded by AP 1 AP 1 should act as receiver in slot 2
Trace-Driven Simulation All clients and APs are in a single collision domain Vary the number of clients (N) Number of APs is always Assume no power adaptation Other algorithms simulated Omniscient TDMAIEEE 802.11 14
Simulation Results: Throughput 15
Challenges SynchronizationMultiCollision domain Inconsistency in the number of APs Robustness Reducing the overhead of channel estimation
Summary RobinHood leverages the high density of APs to scale the uplink wireless throughput for single antenna mobile clients. 17 Thank you