WINLAB IAB Meeting May 2014 WINLAB Rutgers, The State University of New Jersey www.winlab.rutgers.edu Contact: Professor D.
Download ReportTranscript WINLAB IAB Meeting May 2014 WINLAB Rutgers, The State University of New Jersey www.winlab.rutgers.edu Contact: Professor D.
WINLAB IAB Meeting May 2014 WINLAB Rutgers, The State University of New Jersey www.winlab.rutgers.edu Contact: Professor D. Raychaudhuri, Director [email protected] 1 STATUS UPDATE 2 Status Update: General News WINLAB’s overall activity level remained stable in FY 2013-14 External research funding for FY14 ~$6M total Follow-on $5M (~$2.3M at Rutgers) “MobilityFirst” FIA-NP project Other recent grants include NSF Wideband SDR (300K), DARPA Spectrum Challenge (450K), NSF JUNO (300K), NSF Automotive Infoverse ($100K), … Faculty and visitor appointments during 2013-14 Prof. Hana Godrich, WINLAB/ECE – expertise in smart grids & control Dr. Yi Hu, Post Doctoral Associate for Future Internet research Dr. K.K. Ramakrishnan (ex-AT&T), Adjunct Prof on MobilityFirst project Prof. Larry Greenstein retiring after 12 years of service Some highlights DARPA spectrum challenge completed successfully – IEEE Spectrum Article Global-scale MobilityFirst protocol & application demos at GEC-17,18 2014 IEEE Donald J. Fink Award for IEEE Proceedings paper on future of wireless 2014 ECEDHA Innovative Program Award to WINLAB Best paper award for MobilityFirst service API at ACM MobiArch 2013 Conducted benchmark simulations on 802.11p for auto industry consortium (CAMP) WINLAB 3 Status Update: WINLAB Research Targets Mobile web services IP Routing + Cellular Mobility Content- and context-aware protocols, M2M Programmable networks, cloud services, ContentPrivacy, HCI, mobile social networks and contextaware pervasive services Storage-aware routing, global name resolution, location, vehicular nets, privacy/security, ad hoc/DTN routing, … Cooperative relay, cross-layer, beam switching, software MAC,.. Static MAC Protocols Mobility-Centric Internet Arch Flexible & Adaptive MAC Network MIMO, network coding, interference alignment, 60 Ghz, Next-Gen Gigabit PHY Single User MIMO/OFDM Spectrum sensing, NC-OFDM, Spectrum server, cognitive algorithms, Coordination protocols, .. Static Spectrum Assignment Dynamic Spectrum Assignment ~10x eficiency WINLAB Status Update: Research Topics Current research themes at WINLAB (..not an exhaustive list): Radio/PHY cluster Networking cluster Pervasive cluster Cognitive radio hardware (Seskar, Spasojevic) Collaborative PHY (Spasojevic, Petropulu) Spectrum sensing and DSA algorithms (Mandayam, Spasojevic, Seskar) White space backhaul (Mandayam, Seskar) Network coding (Mandayam, Spasojevic) Rechargeable wireless networks (Yates) Error coding techniques for mmWave 5G (Spasojevic) Sensor networks and M2M/IoT (Zhang, Trappe, Howard, Li, Martin) Network-assisted DSA and SDN implementation (Raychaudhuri, Seskar) Experimental platforms: ORBIT, SDR, GENI WiMax/LTE, OpenFlow… (Seskar, Raychaudhuri) Future Internet architecture (Raychaudhuri, Zhang, Yates, Trappe, Martin, Nagaraja) Mobile cloud computing (Zhang, Raychaudhuri, Seskar, Nagaraja) Vehicular networks – wireless and optical V2V (Gruteser) Security and privacy in wireless networks (Trappe, Gruteser, Zhang, Lindqvist) Mobile content caching delivery (Yates, Zhang, Gruteser) Context-aware mobile applications (Martin, Zhang, Gruteser, Lindqvist, Howard) Mobile social networks (Lindqvist, Gruteser) WINLAB 5 Status Update: Selected Govt Funded Projects 5/14 Govt funded projects at WINLAB ORBIT CRI: major equipment upgrade for radio grid testbed (NSF CRI, ‘10-14) GENI Spiral III projects – Open WiMAX enhancements & kit (BBN GENI ‘11-’14) MobilityFirst Future Internet Architecture – Next Phase (NSF, 2014-16) Visual MIMO (NSF, 2011-14) Bandwidth Exchange (NRL, 2011-14) Spectrum Cooperation (SAVANT) & Optical Sensing Platform (NSF EARS, 2013-16) Mobile Ad Hoc Network (ARMY, 2012-14) Spectrum Challenge (DARPA, 2013-) WISER Wideband Software Defined Radio Platform (NSF, 2013-16) Location privacy via de-identification (NSF Career 2009-14) Crowdsourcing of Physical-World Tasks with Myrmex (NSF SoCS 2012-15) Mobile Privacy Projects (NSF TWC 2012-15) Vehicle Safety Messaging (CAMP, 2013-14) Virtual Mobile Cloud Network (NSF JUNO, 2014-17) WINLAB 6 Status Update: Pending Proposals 5/14 2013-14 continued to be active year for medium/large federal proposals to NSF, NRL, DARPA, Army and other federal agencies. Several pending govt proposals including: RF Equipment for Dynamic Spectrum Access (ONR, $248K) Computing Backend for Data Analysis of Wireless & Network Data (ARO, 129K) SciWiNet: Science Wireless Network for Research Community (NSF, 100K) Dependable Context Sensing for Mobile Safety (NSF CSR, 537K) Cloud Processing Extensions to ORBIT ($2M) Security for Body Communications (NSF TWC, 250K) Status Updating Systems and Networks (NSF CIF, 476K) Redesigning Web Browsers for Privacy (NSF TWC, 250K) Robust Gesture Based Authentication (NSF TWC, 399K) Transmit Only: Cloud Enabled Green Comm (NSF NeTS, 498K) End User Behavior and Prospect Pricing in Wireless Data Networks (NSF NeTS, 500K) Software Defined Scientific Data Networks (NSF CC*IIE, 1000K) Green Architecture for Rural Backhaul using TVWS and Solar (NSF EARS, 744K) Coexistence in Spectrum: Coalitions, Machine Learning and Distributed Strategies (NSF EARS, 749K) …+ others WINLAB 7 Status Update: Major Future Research Themes Dynamic spectrum: measurement, backhaul, small cell, etc. “Cloud RAN” architecture & key technologies Software defined wireless/mobile systems (SDN + SDR) “Edge cloud” for mobile and real-time CPS services “Big Data” architecture/privacy/applications for mobile Passive sensing & localization for in-building context …other WINLAB 8 Status Update: Industry Research Topics (cont.) Some previous inputs from sponsors: Verizon: Multi-antenna dense deployment, WiFi integration, stadium deployment at Rutgers, InterDigital: ICN architecture, content network models, .. Qualcomm: vehicular networking including smartphone integration, 60 Ghz WLAN, application concepts Ericsson: 5G cellular technologies, dynamic spectrum, future Internet DoCoMo; dense small cells, 5G architecture, integrated scheduling & control of 5G macro/small cell Huawei: internet-of-things (IoT) architecture, 5G mobile network, whitespace backhaul NEC: SDN for wireless WINLAB 9 ~~ Status Update: Industry Sponsors 5/14 US Army CECOM * * * * * InPoint Semandex Zipreel * * *Research Partners WINLAB 10 ~~ WINLAB Summary: People Dipankar Raychaudhuri Yicheng Lu Melissa Gelfman Hui Xiong Roy Yates Narayan Mandayam Chris Rose Athina Petropulu Larry Greenstein Dick Frenkiel Noreen DeCarlo Janice Campanella Thu Nguyen Silvija Kokalj-Filipovic Elaine Connors K.K. Ramakrishnan Wade Trappe Predrag Spasojevic Rich Howard Richard Martin Shridatt Sugrim Ilya Chigivev Yanyong Zhang Marco Gruteser Kiran Nagaraja Kishore Ramachandran Khanh Le Ivan Seskar Janne Lindqvist ~40-PhD & MS Students as of 2011 (see www.winlab.rutgers.edu for photos) WINLAB Research Highlights 12 SAVANT: Inter-network Spectrum Coordination Leverage the fact that Internet connects almost all interfering devices! Key Advantages: • Communication over the back-end is virtually free • Precise view of the traffic is readily available • Works across technologies (WiFi/cellular/others) Underlying Network Infrastructure (Internet) Spectrum Gateway (SG) Spectrum Gateway (SG) AP WINLAB SAVANT: Architecture Architecture involves two protocol interface levels between independent wireless domains: • Lower layer for sharing aggregate radio map using technology neutral parameters • Higher layer for negotiating spectrum use policy, radio resource management (RRM) algorithms, and controller delegation WINLAB SAVANT: The Protocol Internetwork Spectrum Coordination Protocol (ISCP) S-Interface for radio map Contains network identifier, network parameters (such as location, technology type, etc.) and a summary radio map M-Interface for algorithm and policy Contains network identifier, type of action, e.g., information exchange, policy query, policy query response, controller delegation request, controller delegation response WINLAB SAVANT: Radio Map A fundamental research problem: How to best describe a radio environment? A pragmatic approach Per-Transceiver Information Frequency Domain Power Spectral Density Mask: Start-stop frequencies & corresponding Tx power level Time Domain Avg. Duty Cycle | Scheduled/Random Access Physical Characteristics X, Y, Z Coordinates | Outdoor/Indoor | Antenna Height | No. of Antennas | Antenna Directionality Receiver Characteristics Rx Sensitivity in dB | Adaptation Logic Freq. Change Algorithm | Rate Control Algorithm Aggregate Information Estimated aggregate interference power map: dBm v/s location & frequency table Estimated spectrum occupancy ranking: Ordered list of most-free frequencies WINLAB SAVANT: Algorithms Non-adaptive parameter selection (NAPS) Explicit query-response only on initialization Intended for low-cost devices with simple radios Adaptive Parameter Selection (APS) Initialization procedure same as NAPS Additional periodic spectrum usage updates provided by its radio neighbors Global Coordinated Resource Packing (GCRP) Provisions direct exchange of coordination messages between different devices in radio range Leads to iterative algorithm for globally optimal resource packing between co-existing radios WINLAB SAVANT: Quantifying The Value of Cooperation AP 4 AP 1 AP 2 AP 3 30 AP 4 20 AP 1 10 0 AP 2 AP 3 Throughput(Mbps) ORBIT experiments show local observations are not enough In all cases AP1 makes the same local observation but its throughput widely varies based on how other nodes are connected: Throughput(Mbps) AP 1 AP 2 AP 3 AP 4 30 20 10 0 AP 1 AP 2 AP 3 AP 4 AP 2 AP 3 AP 1 30 AP 2 20 10 0 AP 1 AP 2 AP 3 AP 4 AP 4 AP 3 Throughput(Mbps) AP 1 Throughput(Mbps) AP 4 30 20 10 0 AP 1 AP 2 AP 3 AP 4 WINLAB SAVANT: Cooperative Association Control Simulation study to find the potential gains from cooperation for Wi-Fi client-AP association optimization problem (i.e. given several APs in range, which one to connect to) 10 percentile client throughput 2 Mean client throughput Throughput (Mbps) 0.6 Throughput (Mbps) 0.4 0.2 0 N=2 N=3 N=4 Least Distance 1.5 1 0.5 0 N=2 N=3 N=4 Intra-network Optim. Cooperative Optim. WINLAB SAVANT: Cooperative Channel Assignment Role of cooperation for Wi-Fi channel assignment Dense Wi-Fi network simulations with realistic modeling of the normalized throughput given 3 channels Full Cooperation Mean Normalized Throughput No Cooperation 1 0.8 0.6 0.4 50 APs/sq. km. 100 APs/sq. km. 150 APs/sq. km. 200 APs/sq. km. 0.2 0 0:100 25:75 50:50 75:25 50 Percentage of starved APs 100:0 Ratio of Non-cooperative vs. Cooperative APs 50 APs/sq. km. 100 APs/sq. km. 150 APs/sq. km. 200 APs/sq. km. 40 30 20 10 0 -10 0:100 25:75 50:50 75:25 100:0 Ratio of Non-cooperative vs. Cooperative APs WINLAB SAVANT: Implementation Inter-network cooperation requires more than just a communication link Needs integration with the way control plane is implemented in wireless networks We leverage Software Defined Networking (SDN) techniques for introducing flexibility in the wireless control plane Data Plane Apps Mobility Mgmt. Wireless Access Control QoS Control Control Plane Apps Channel Assignment Tx Power Control InterNetwork Coordination Network OS with wireless abstractions Through extension of OpenFlow Match/Action Fields Through the ControlSwitch framework Wired + Wireless Network WINLAB SAVANT: ORBIT Two-Network Experiment Application: Load-dependent channel assignment Each client demands between 5-45 Mbps UDP uniformly randomly, controllers need to select best channel 2 Networks with 4 APs/4 clients each, random topology Controller A ap.py Wireless Link AP1 AP2 AP3 AP4 Cl1 Cl2 Cl3 Cl4 CSMA Topology ORBIT Nodes 1 8 controller.py Inter-network Spectrum Coordination API 3 5 6 4 7 2 Cl5 Cl6 Cl7 Cl8 AP5 AP6 AP7 AP8 controller.py ap.py Controller B WINLAB SAVANT: ORBIT Two-Network Experiment Results • Showing results for a particular topology • Initially, each network sets the channels of its own APs • At t=25 secs, Controller B provides information about AP loads to Controller A and hands over control • Controller A re-computes channels for all APs, conveys it to APs which have to change 35 8 3 30 5 6 2 7 4 1 2 3 4 5 6 7 8 Load(Mbps 40 41 9 7 12 40 9 22 ) Initial Ch. 1 1 1 11 1 11 11 1 Throughput (Mbps) 1 1 1 1 11 11 11 11 11 6 2 25 20 15 8 1 5 10 3 7 4 5 0 Final Ch. 1 2 3 4 5 6 7 8 10 20 30 40 50 Time (seconds)WINLAB 60 Rechargeable Networks: Optimal Retransmission Policies PI: Roy Yates Picture Courtesy of S. Roundy, UC Berkeley Energy replenishment rate is stochastic and environment-constrained Rechargeable battery has a long life but its energy should not be abused Message transmission carries different rewards Our goal is to maximize the average reward rate by selective transmission . WINLAB Rechargeable Networks: Continuous-Time Markov Model for Battery with Capacity N=5 Hybrid Energy Replenishment Modeled by Poisson Process PI: Roy Yates Battery replacement at rate α Energy Level high Battery recharging at rate β low 0 λ 5,1 3 2 1 λ 5,2 λ 5,3 55 4 λ 5,4 λ 5,5 Characterizing a transmission policy by a set of state-dependent (energy and capacity aware) thresholds {τ*N,i }, i=1, 2, … N Determining “optimal” thresholds by invoking the “termination rule’’ of Howard’s policy improvement algorithm WINLAB Network Coding: DE Framework for Cross-layer Resource Allocation in RNC PI: Narayan Mandayam Model RNC as a dynamical system: Packet reception rate Innovative packet probability Hyperarc capacity of Tx rate at node i Reception probability (A pkt from i can be recvd by at least 1 node in K ) MAC PHY DE framework closely models rank evolution of RNC in terms of PHY and MAC parameters Cross-layer Design problems Solving systems differential of equations Appropriate boundary conditions WINLAB 26 Network Coding Aware Power Control PI: Narayan Mandayam Without Power Control (PC), throughput is a constant With PC: the instant throughput is improved compared with no PC case With PC: throughput Converges around t=60ms Each node Tx at 1pkt/ms, min cut between. src. and dst. is 1 pkt/ms 5 1 3 4 RNC with PC achieves optimal throughput! WINLAB 27 Bandwidth Exchange: Framework for Enhancing Approach and Accomplishments Performance of CR Networks Proved the convexity/concavity of the optimization problem formulations 0 0 Compared the performance of bandwidth exchange with direct transmission scenario 1 1 2 Direct transmission 2 Bandwidth exchanged based cooperative forwarding Technical Significance Use of transmission bandwidth as a flexible resource in cooperative forwarding Allows dynamic and opportunistic assignment of non-contiguous portions of spectrum to nodes • Provided a distributed cooperative pair selection algorithm using graph theory Alternate use in transmission time slot exchange Implemented time exchange algorithm in software defined radio (USRP2) nodes of the ORBIT testbed. In Progress: Joint resource allocation, scheduling and routing in cognitive radio enabled multi-hop wireless backhaul Office of Naval Research Grant PI: Narayan Mandayam Impacts Trades off between – • Throughput gain • Power savings • Coverage extension WINLAB Testbed Results of Time Exchange PI: Narayan Mandayam 3 10m 0 9m 12m 17m 20m 5m 2 1 • Pick 4 USRP nodes of ORBIT • 1,2 & 3 are users. Node 0 is the base station. • Assign 1s orthogonal time slot to each node • ORBIT is used as the global control plane • Node 1 & 3 get selected as the optimal cooperative pair through maximum matching • Node 1 & 3 cooperate and improve goodput through proportional fair objective • Node 2 transmits with the same goodput WINLAB 29 EDMAC: An Enhanced Directional MAC Protocol for 60GHz Networks • In 60GHz networks, transmitters and receivers employ directional antennas and point their main beams toward each other to overcome high propagation losses and achieve high data rates • • Problems can be solved if nodes do not employ exponential backoff mechanism and use the same fixed contention window if they send to the same receiver. For a network with n senders and one receiver, from analysis we obtain the channel throughput S where , Ttransmit, TRTS , Tdata are constant system parameter and is the packet arrival rate of each node. The maximum channel throughput is achieved when Since we , we can find the optimal contention window W is only related with n CSMA based directional MAC (DMAC) protocols suffer from the “deafness” problem which causes unfairness and low channel utilization. PI: Roy Yates WINLAB EDMAC: An Enhanced Directional MAC Protocol for 60GHz Networks Throughput model validation using ns-2 n=2 • Performance evaluation using ns-2 Scenario 1: Two hops scenario: A UDP flow from node S to node D via relay node R. n=9 Scenario 2: Multi-hops scenario: A TCP or UDPflow from node S to node D via EDMAC: Receiver calculates W based on number of senders n and inform its multiple relay nodes sender. Sender store W in its neighbor tables which guarantees nodes send to the same receiver use same contention window size W PI: Roy Yates Example of a neighbor table WINLAB Study of Coexistence of Mobile-Fixed AP Basic Idea • Study of mobile WLAN hotspot which provides cellular-WiFi tethering services to personal devices. Objectives of the study • Heterogeneous traffic condition analysis to model coexisting fixed and mobile WLAN hotspot with saturated-unsaturated traffic conditions due to limited backhaul capacity at mobile WLAN • Proposal of Adaptive Channel Assignment (ACA) technique to mitigate interference and increase the data throughput at mobile APs. • Study of effect of mobile speed to further enhance throughput performance due to ACA Fig. Co-existence scenario of fixed and mobile AP where a user travelling with mobile WLAN from point A to B can be in the range of number of fixed WLAN. WINLAB Coexistence of Mobile Access Points Automatic Channel Assignment (ACA) algorithm: Mobile AP scans WiFi channels every few seconds and selects the channel which is least crowded in the carrier sense range Fig. Throughput at mobile AP as function of number of fixed APs Note: With application of ACA, maximum gain in throughput is 1.24 Mbps achieving up to 42% of percentage throughput gain Fig. Throughput at mobile AP (Mbps) as a function ACA scanning period when no. of fix AP = 50 and assuming channel load is determined in total time 200ms. Note: Performance is evaluated for mobility speed s = {10, 20, 40, 60} mph WINLAB Sensor-assisted anomaly detection for detecting manipulation and exploitation Network Structure for Anomaly Detection Primary (authorized) transmitter is stationary Distributed detection by a network of sensors that collaborate locally. Significance Testing Test statistic T: a measure of observed data Acceptance Region Ω: we accept the null hypothesis if T Ω Significance level : probability of false alarm When a channel is dedicated to a single authorized user we can try to distinguish between single and multiple transmissions Formulate a decision statistic that captures the characteristics of the received power in the normal case [34] [34] WINLAB PI: Wade Trappe Detecting jamming against the collected system is complicated by normal interference Normal Interference in Mobile Networks JamRX Experiments have shown that the hidden terminal problem remains in spite of MAClayer collision-avoidance (e.g. a transmitter outside of the physical carrier sensing range can still cause interference). It is equivalent to a low-power jamming attack. Other jamming attacks, such as reactive attacks, require different detection mechanism Sender-oriented detection of jamming can utilize network ACKs and signal levels to detect jamming AER-RSS signal space consists of three regions Interference-free: no hidden terminal Normal interference: caused by legitimate hidden terminals Intentional interference: malicious jamming [35] [35] ACK Missing Rate PI: Wade Trappe 1 RandTX w/P 0.8 ReactTX w/P RandTX 0.6 ReactTX 0.4 No Jam 0.2 0 0 0 0.5 0.5 ACK Block Rate 1 1 ACK Error Rate WINLAB Non-quantum photonic secret key establishment Secret key establishment is fundamental to supporting cryptographic services (confidentiality & authentication) The photonic layer can be a rich source for establishing keys between two entities Quantum Cryptography is the standard example, but this poses serious engineering challenges Can one optically establish keys without resorting to PI: Wade Trappe quantum physics? Result: Answer: Yes – Alice and Bob create correlated sources, while Eve is uncorrelated Prototype design using large-scale Mach Zehnder – Distillation and privacy amplification finalize interferometer the process, creating reliable crypto keys! WINLAB DARPA: RadioMAP Task 2, Management of RF Network and Tasking Infrastructure (MARTI) PI: Wade Trappe Proposal Team: Applied Communication Sciences WINLAB, Rutgers University Carnegie Mellon University Research Goals: A distributed system executing on participating RF devices that performs reception, transmission and local processing tasks on behalf of RadioMap applications: Without manual intervention Subject to available resources and limited impact on the primary mission of the device. Software that intelligently assigns tasks to RF devices and collects results on behalf of applications. Trading off probability of success and overhead Provides standardized mechanisms for tasking Provides standardized reporting mechanisms and formats Modularity and Layering Any RF device should be able to perform tasks for any application without customizing MARTI infrastructure RF Devices, Applications. [37] WINLAB WINLAB is part of RCS3 (DARPA-SSPARC) team, addressing S-band RADAR-Comms Interference The RCS3 SSPARC Team is developing a cognitive control system with multiple systemlevel and device-level spectrum separation techniques to address Radar-Comms interference: Scalability through separating global control from multiple local control systems, with the environment partitioned intelligently via RF propagation analysis and geospatial reasoning, Development and characterization of a range of spectrum sharing techniques leveraging closed-loop information sharing between the radar and communications systems, High-level management of the cognitive control system by a spectrum sharing priority "dial" that can be used to rebalance [38] [38] resources to meet evolving mission needs. RF Coupling Map Map Understanding Global Controller Requested Map Information (Slow) Information Sharing Subsystem WINLAB EARS: Collaborative Research: Big Bandwidth: Finding Anomalous Needles in the Spectrum Haystack PI: Wade Trappe Potential Payoffs: Spectrum is a valuable resource that, if properly used, will spur economic growth, but if used improperly could hinder economic growth. The proposed project, if successful, will have the following payoffs: Provide mechanisms by which the government can ensure spectrum is used properly by those who have negotiated access rights. Provide algorithms for mapping spectral activity across a large time, frequency and spatial domain. New algorithms and hardware will advance knowledge in sampling ultra-wide bandwidth, allowing for a comparison between state-of-the-art in sub-Nyquist and RF photonics. Impact education as the project is inherently multidisciplinary, and will lead to new curricular efforts between security, wireless and photonics. Proposal Title: EARS: Collaborative Research: Big Bandwidth: Finding Anomalous Needles in the Spectrum Haystack Proposal Numbers: 1247864 & 1247298 PI Names: Wade Trappe (Rutgers) Larry Greenstein (Rutgers) Paul Prucnal (Princeton) Research Goals: The goal behind the project is to develop a suite of tools that can facilitate the detection of improper usage of radio spectrum. To accomplish this, the project involves the following research goals Develop Algorithms and Hardware for a SingleScanner. The project explores how a single scanner: Should allocate its scanning strategy to best detect an unknown signal. Develop sub-Nyquist techniques that allow digital scanning of wide bandwidths Develop RF photonic scanning that allows for scanning of wide bandwidths Develop Algorithms for Multiple-Scanners. Multiple sensors allows for coordinated scanning. The project will examine how scanning should be allocated across sensors to detect anomalous transmissions. Frequency [39] WINLAB Pipsqueak and Owl Platform PIs: Rich Martin, Rich Howard & Yanyong Zhang Owl Platform Software Stack Scalable, distributed software architecture 10+ year ultra-low energy wireless sensor Reliable delivery in challenging radio environments Multiple sensing capabilities IoT software stack targeting data streams from wireless sensors WINLAB TO – comparing wireless sensor boards PIs: Rich Martin, Rich Howard & Yanyong Zhang Classic Transmit-Only TelosB (2004) TO-PIP(2013) Antenna Radio Micro controller Battery 41 WINLAB WINLAB and InPoint Systems PIs: Rich Martin, Rich Howard & Yanyong Zhang Foundational research enabled by access to WINLAB faculty and staff NSF I-Corps recipient Interaction with other university units SBIR/STTR permit exploration of commercial opportunities Start-up matures the technology and Rutgers receives upside First customer received delivery this week Rutgers Comparative Medicine Resources WINLAB Owl Platform @ Rutgers CMR PIs: Rich Martin, Rich Howard & Yanyong Zhang 9-month installation Improved staff workflow Reduced problem response times Supportive CMR management and faculty Product requirements developed WINLAB WiSER Cognitive Radio Platform Focus on Creativity, not Engineering Complexity : Split Baseband in two domain spaces : • Dynamic – Swappable INNOVATION CYCLE • Communication APPs (creative problem) Static - Open-sourced System-onChip (complex engineering problem) Abstract lower level design complexities from Users FSoC Features Access to lower level resources thru APIs VITA radio transport protocol for radio control Networking capable node Support up to four dynamic APPs Library of Open-sourced Communication APPs Static Framework utilization level < 15% for V5SX95, even less for newer technologies, for ex. Virtex7 . Transparent to underlying FPGA technology. Can be ported to future HW platforms and newer FPGA technologies. CRKIT = make real-time and widetuning radio a viable solution for large scale experiments. Live system runs Pis: Ivan Seskar, D. Raychaudhuri WDR from Radio Technology Solutions WINLAB WiSER Programming Model Network CRKIT HOST CRKIT development Application development Java, C# C GUI System Debugging C Comm. APP Algorithm System Test CR DSA VHDL/ Verilog Mathworks Simulink Embedded SW IP Networking HW Configuration Host CMD Parsing DHCP/ARP Lookup Tables/ RF ETH/VITA WINLAB WiSER Baseline Hardware Zynq-7000 SoC / Analog Devices Software-Defined Radio Kit ZedBoard baseboard (Zynq XC7Z020 device) Dual-core ARM® Cortex™-A9 256 KB on-chip RAM Gigabit Ethernet, 2x SD/SDIO, USB,CAN, SPI, UART,I2C 512 MB DDR3, 256 Mb QSPI Flash 85K Logic Cells, 106K FF 220 Programmable DSP Slices (18x25 MACCs) Analog Devices FMC RF Front-end Software tunable across wide frequency range (400MHz to 4GHz) with 125MHz channel bandwidth (250MSPS ADC, 1GSPS DAC) RF section bypass for baseband sampling Phase and frequency synchronization on both transmit and receive paths WINLAB WiSER Framework Architecture 1. Dual-core ARM processors • Linux support • Dual AXI bus architecture • Independent Data and Control traffic 2. Independent APP sampling rates • Support Multirate and Multi-APP systems • Decoupling of APP clock domains from overall Framework. • Permits Spectrum Sensing APP + Communication APP in same architecture 3. Applications • Reuse previously designed APPs • NC-OFDM • Spectrum Sensing 4. RF • • • 400MHz to 4GHz tuning range 125MHz Channel Bandwidth (250MSPS ADC, 1GSPS DAC) Full-duplex WINLAB SDN Wireless Research Current scope of WINLAB activities on SDN: – – – – – – – GENI campus network (OpenFlow) GENI Open Base station ORBIT SDN sandbox MobilityFirst Prototype OpenFlow extensions for WiFI, etc. SDN control plane and application to DSA Cellular/mobile network WINLAB SDN Wireless: OPEN BTS Concept GENI Project – Open WiMAX BTS Exposed all controllable parameters through API Removed all default IP routing, simplified ASN controller* All switching purely based on MAC addresses Implemented the datapath virtualization and VNTS shaping mechanism in click/openvswitch for slice isolation Ongoing work – Open LTE BTS Exposed all controllable parameters through the same REST based API Implemented the datapath with openvswitch Current development: ePC replacement with open source aggregate manager (i.e. simplification/elimination of LTE control protocols) WINLAB SDN Wireless: WiMAX Open BTS eth0 eth0.vl1 eth0.vln eth0.vl2 Traffic Scheduler/Shaper eth2 RF Aggregate Manager OpenVSwitch eth1 control data WINLAB SDN Wireless: LTE Open BTS eth2 eth0.vl1 RF/ePC Aggregate Manager eth0.vl2 X2,S1-U,S1-MME,... eth WINLAB SDN Wireless: Open BTS + network SDN Datapah Complex Adaptation/Handoff Controller Generic Resource Controller ... OPEN BS2 ... OPEN BS3 WINLAB SDN Wireless: Open API features Radio Resource Management (RRM) Set of MCS ARQ/HARQ Handoff Admission Control Frequency planning Interference Management Data Collection/Reporting… WINLAB SDN Wireless: Control Plane Concept Introducing flexibility in the wireless control plane by leveraging software defined networking techniques Inter-network cooperation translates to inter-controller interactions and setting of flow-rules Data Plane Apps Mobility Mgmt. Wireless Access Control QoS Control Control Plane Apps Channel Assignment Tx Power Control InterNetwork Coordination Network OS with wireless abstractions Through extension of OpenFlow Match/Action Fields Wired + Wireless Network Through the ControlSwitch framework WINLAB SDN Wireless: Basic Design Interpret wireless control messages as flows BS/AP uses Match/Action rules to forward incoming and outgoing control-flows Control traffic can be forwarded to/from other BSs or central controller Local SDN based controller for low latency actions Controller Insert Flow Rules BS HW Control Messages Local Controller Control Datapath Match Fields Msg Type Parameters IP Port … Action Set Channel Forward to Port 1 Report Beacons Forward to IP1,IP2 WINLAB SDN Wireless: Distributed Control Extension of traditional Enterprise Controller: Multiple copies of wireless controllers (WC) with mechanisms to cooperate, scattered throughout SDN based control plane Reduced distance between device and a controller – reduced flow setup times (reduced control latency) Shared State Wireless Controller Wireless Controller Control BS/AP Control Network (data plane) WINLAB SDN Wireless: Hierarchical Control Example: Pair of enterprises with heterogeneous decomposed controllers Authentication/Interference Tier 2 AI 1 AI 1 Handoff Tier 1 H1 H2 BS/AP Control Network (data plane) H1 H3 H2 BS/AP Control Network (data plane) WINLAB Seeded by an NSF I-Corps grant - Early stage incubation at WINLAB ZipReel Inc. Cloud video processing INPUT: High volume, pro-generated video Pis: Kishore Ramachandran, D. Raychaudhuri OUTPUT: Fast delivery of highquality, multi-format processed video “Processing” Transcoding Format conversion Object search Feature insertion Networks Delivery via CDNs Linear scaling with compute units Professional-grade quality Contact: [email protected] Zipreel cloud transcoding applications Netflix Encodes Every Movie 120 Different Ways…streams to 900 different types of devices… - gizmodo.com Broadcasters (e.g. ESPN) + Cable/Telco operators (e.g. Comcast) want to replicate tech. but buy-rather-than-build approach Critical for Mobile / Internet streaming 2-3 days to generate 120 formats for one video! - Netflix @ AWS Re-invent Content owners lose significant ad-revenues for each additional day of processing Pain point: how to process high video volumes fast? 59 WINLAB Zipreel Cloud Video Technology Before • • • After Approach: process videos on commodity clusters in software Challenge: making the cluster interconnection network scale Solution: hierarchical network topology design and/or efficient use of multicasting 60 WINLAB JUNO: Virtual Mobile Cloud Network (mVCN) (..joint project with NICT under NSF-Japan collab) Project concept: Dynamic Cloud Migration for Fast Response Real-Time Services Key Technologies: Virtual Network built on MobilityFirst GUID Foundation, Cloud Migration Strategies, etc. WINLAB JUNO: Virtual Mobile Cloud Network (mVCN) (..joint project with NICT under NSF-Japan collab) Technical Approach: MF service Addressability via GUID, anycast, virtual networks, .. WINLAB The Automotive Infoverse PI: Marco Gruteser Parking Availability Estimation Rangefinde r + GPS Wireless Service Valid Parking Spot Map MobiSys 2010 BestWINLAB Paper Development of a V2V Scalability Simulator PI: Marco Gruteser Develops a Dedicated Short Range Communications (DSRC) simulator for the Crash Avoidance Metrics Partnership Vehicle Safety Communications 3 consortium Uses field test data from hundreds of DSRC equipped vehicles to develop and calibrate simulation models Aims to accurately predict V2V communication performance in very dense, interferencelimited scenarios. WINLAB Distinguishing Users with Capacitive Communications PI: Marco Gruteser Need for better user identification and authentication techniques on post-pc devices Approach: a wearable token can transmit short codes of data through capacitive touch screens 100 2 bits 3 bits 4 bits 5 bits 80 Detection Rate (%) 60 40 20 0 4 bits/s 5 bits/s WINLAB Visual MIMO Networks Transmitter Array PI: Marco Gruteser Receiver Array Explores MIMO free space optical communications with camera receivers (e.g., download billboard adds by pointing phone camera at it) WINLAB Augmented Tags PI: Marco Gruteser WINLAB Smart Meter Privacy and Security PI: Marco Gruteser Analyzed existing meters with USRP software radio Meters broadcast every 30s Able to spoof readings and eavesdrop on hundreds of meters WINLAB Reducing Driver Distraction Phone driver distraction contributed to 995 fatalities in 2009 (NHTSA) Goal: Make phone aware of driver use and design interfaces that reduce distractions Key challenge: Distinguishing driver and passengers PI: Marco Gruteser • Idea: leverage car speakers for audio localization of cell phone WINLAB Mobile Safety Services PI: Marco Gruteser Phone driver distraction contributed to 995 fatalities in 2009 (NHTSA) Goal: Develop toolkit and platform to facilitate mobile safety service development Allow crowdsourcing of reliability data and adaptation of interventions WINLAB Elastic Pathing: Speed is Enough to Track You Ground Trut h PI: Janne Lindqvist Predict ed Pat h Lat it ude 1 Mile 2 km Longit ude WINLAB Crowdsourcing for Privacy (w/CMU) PI: Janne Lindqvist Almost nobody reads privacy policies We want to install the app Reading policies not part of main task Complexity of reading these policies (boring!!!!!) Clear cost (my time) for unclear benefit Crowdsourcing can mitigate these problems But what to crowdsource here? Our idea: expectations and misconceptions 95% users were surprised this app sent their approximate location to mobile ads providers. 95% users were surprised this app sent their phone’s unique ID to mobile ads providers. 90% users were surprised this app sent their precise location to mobile ads providers. 0% users were surprised this app can control camera flashlight. See all WINLAB Effective Location Privacy Disclosures PI: Janne Lindqvist WINLAB User-Generated Free-Form Gestures for Authentication PI: Janne Lindqvist Measuring the security and memorability of usergenerated free-form gestures No visual reference Multitouch: Single or multiple fingers To appear in MobiSys’14 WINLAB Password Security PI: Janne Lindqvist all passwords How people generate passwords and recall them on different devices? Novel ways to measure password security. 75 Password strength/bits recalled only Type of metric 50 ● ● ● ● ● ● N−gram strength Random entropy NIST entropy ● 25 0 p La t t top hone able top hone able p a T T P P L Terminal type WINLAB Local Community Crowdsourcing of Physical-World Tasks with Myrmex Approach: Opportunistically offering locationconstrained tasks to people PI: Janne Lindqvist “Mechanical turk for the real-world” Results in: Stronger communities Everybody saves WINLAB Motivations and Experiences of the On-Demand Mobile Workforce PI: Janne Lindqvist The main motivations for joining on-demand mobile workforce involve desires for monetary compensation and personal control over one’s schedule and freedom to opt into or out of tasks Situational factors such as the day of the week and weather conditions influence worker’s task selection practices. Convenient physical locations and unambiguous profile information of task requesters also influence task selection practices as well. WINLAB MobilityFirst Update FIA NP Projects Announced 5/14: WINLAB MobilityFirst Design: Architecture Features Named devices, content, and context Strong authentication, privacy 11001101011100100…0011 Public Key Based Global Identifier (GUID) Human-readable name Heterogeneous Wireless Access Service API with unicast, multi-homing, mcast, anycast, content query, etc. Routers with Integrated Storage & Computing End-Point mobility with multi-homing In-network content cache Storage-aware Intra-domain routing Edge-aware Inter-domain routing Hop-by-hop file transport MobilityFirst Protocol Design Goals: - 10B+ mobile/wireless devices Mobility as a basic service BW variation & disconnection tolerance Ad-hoc edge networks & network mobility Multihoming, multipath, multicast Content & context-aware services Strong security/trust and privacy model Connectionless Packet Switched Network with hybrid name/address routing Network Mobility & Disconnected Mode Ad-hoc p2p mode WINLAB MF Design: Protocol Stack App 1 App 2 App 3 App 4 E2E TP3 E2E TP4 Socket API Name Certification & Assignment Service NCS E2E TP1 E2E TP2 Optional Compute Layer Plug-In A Global Name Resolution Service GNRS MF Routing Control Protocol GUID Service Layer GSTAR Routing MF Inter-Domain Hop-by-Hop Block Transfer Link Layer 1 (802.11) Link Layer 2 (LTE) Narrow Waist Link Layer 3 (Ethernet) IP Switching Option Link Layer 4 (SONET) Link Layer 5 (etc.) Control Plane Data Plane WINLAB MF Design: Name-Address Separation GUIDs Separation of names (ID) from network addresses (NA) Globally unique name (GUID) for network attached objects Sue’s_mobile_2 User name, device ID, content, context, AS name, and so on Multiple domain-specific naming services Server_1234 John’s _laptop_1 Host Naming Service Media File_ABC Taxis in NB Sensor@XYZ Sensor Naming Service Content Naming Service Context Naming Service Globally Unique Flat Identifier (GUID) Global Name Resolution Service for GUID NA mappings Global Name Resolution Service Network Hybrid GUID/NA approach Both name/address headers in PDU “Fast path” when NA is available GUID resolution, late binding option Network address Net1.local_ID Net2.local_ID WINLAB MF Protocol Example: Mobility Service via Name Resolution at Device End-Points Service API capabilities: - send (GUID, options, data) Options = anycast, mcast, time, .. - get (content_GUID, options) Options = nearest, all, .. Register “John Smith22’s devices” with NCS Name Certification Services (NCS) GUID assigned GUID lookup from directory NA99 MobilityFirst Network (Data Plane) Send (GUID = 11011..011, SID=01, data) GNRS update (after link-layer association) NA32 GNRS GUID <-> NA lookup GNRS query Send (GUID = 11011..011, SID=01, NA99, NA32, data) GUID = 11011..011 Represents network object with 2 devices DATA GUID SID NAs Packet sent out by host WINLAB MF Protocol Design: Global Name Resolution Service (GNRS) Fast GNRS implementation based on DHT between routers GNRS entries (GUID <-> NA) stored at Router Addr = hash(GUID) Results in distributed in-network directory with fast access (~100 ms) Internet Scale Simulation Results Using DIMES database WINLAB MF Protocol Design: Storage-Aware Intra-Domain Routing (GSTAR) Storage aware (CNF, generalized DTN) routing exploits in-network storage to deal with varying link quality and disconnection Routing algorithm adapts seamlessly adapts from switching (good path) to store-and-forward (poor link BW/short disconnection) to DTN (longer disconnections) Storage has benefits for wired networks as well.. Temporary Storage at Router Initial Routing Path Low BW cellular link Re-routed path For delivery Mobile Device trajectory PDU Storage Router High BW WiFi link Sample CNF routing result WINLAB MF Protocol Design: Edge-Aware InterDomain Routing MF architecture uses a new “edge-aware” inter-domain routing protocol based on link-state and “pathlet” concepts Telescopic link state protocol for dissemination of NSPs Aggregation nodes (aNode) and virtual link(vLink) NSP contains aNode, vLink state including AS internal topology and aggregate edge network quality info; NSP update rate decreases with distance “Late binding” from name-to-address at routers Router has capability of rebinding <GUID=>Address> for packets in transit 1 NSP/sec 0.5 NSP/sec 0.1 NSP/sec AS640 AS993 Alternate Path AS541 vLink AS009 aNode Transit Network AS virtual topology as advertised by network AS#, aNodes, vLinks, params.. NSP packet Router decision based On edge network path Quality – late binding AS90 Edge Network Routed Path For Multi-homing Service EIR Inter Domain Routing Concept WINLAB MF Protocol Design: Hybrid GUID/NA Storage Router in MobilityFirst Hybrid name-address based routing in MobilityFirst requires a new router design with in-network storage and two lookup tables: “Virtual DHT” table for GUID-to-NA lookup as needed Conventional NA-to-port # forwarding table for “fast path” Also, enhanced routing algorithm for store/forward decisions GUID –based forwarding (slow path) GUID-Address Mapping – virtual DHT table Look up GUID-NA table when: - no NAs in pkt header - encapsulated GUID - delivery failure or expired NA entry GUID NA 11001..11 NA99,32 DATA To NA11 Router Storage DATA SID GUID= 11001…11 NA99,NA32 To NA51 Store when: - Poor short-term path quality - Delivery failure, no NA entry - GNRS query failure - etc. NA Forwarding Table – stored physically at router Dest NA Look up NA-next hop table when: - pkt header includes NAs - valid NA to next hop entry Port #, Next Hop NA99 Port 5, NA11 NA62 Port 5, NA11 Port 7, NA51 NA32 DATA Network Address Based Forwarding (fast path) WINLAB MF Protocol Example: Handling Disconnection Store-and-forward mobility service example DATA GUID NA99 rebind to NA75 Delivery failure at NA99 due to device mobility Router stores & periodically checks GNRS binding Deliver to new network NA75 when GNRS updates NA99 Disconnection interval Data Plane Device mobility NA75 DATA DATA GUID NA75 GUID SID NA99 DATA GUID SID Send data file to “John Smith22’s laptop”, SID= 11 (unicast, mobile delivery) WINLAB MF GNRS + Storage Routing Performance Result for WiFi Mobility Scenario Detailed NS3 Simulations to compare MF with TCP/IP Hotspot AP Deployment: Includes gaps and overlaps Cars move according to realistic traces & request browsing type traffic (req. size: 10KB to 5MB) Single Car: Aggregate Throughput vs.Time Empirical CDF of file transfer time 100 Total Data Received (MBits) 1 0.8 CDF 0.6 d: Average distance between APs 0.4 MF: d = 200 TCP/IP: d = 200 MF: Avg. d = 400 TCP/IP: d = 400 0.2 0 0 10 20 30 40 50 File Transfer Time (sec) 60 70 80 60 TCP/IP-30miles/hr TCP/IP-50miles/hr TCP/IP-70miles/hr MF-30miles/hr MF-50miles/hr MF-70miles/hr 40 20 0 0 50 100 Time (sec) WINLAB150 200 LTE/WiFi HetNet Results: MF vs. TCP MF provides several benefits in a heterogeneous wireless environment: Seamless mobility across network domains via dynamic GUID-NA bindings Routers automatically store packets in transit during periods of disconnection Simultaneous use of multiple networks is also possible Aggregate Throughput with Time 1000 900 Aggregate Throughput (MBytes) Throughput boost due to transmission of stored packets 800 700 TCP takes more time to re-start session (DHCP + Application reset) 600 500 400 300 200 MobilityFirst TCP/IP 100 0 0 20 40 60 Time (sec) 80 100 120 WINLAB MF Protocol Example: Dual Homing Service Multihoming service example: LTE + WiFi or LTE + 60 Ghz or LTE1+LTE2 DATA DATA Router bifurcates PDU to NA99 & NA32 (no GUID resolution needed) GUID NetAddr= NA99 NA99 Data Plane NA32 DATA DATA GUID NetAddr= NA32 GUID= SID 11001…11 NA99,NA32 DATA GUID SID Send data file to “John Smith22’s laptop”, SID= 129 (multihoming – all interfaces) WINLAB MF Multi-Homing Example Result Multipath service with data striping between LTE and WiFi Using backpressure propagation and path quality info GNRS Server Query: (GUIDY) 450 Data Response: (NA1,NA2, Policy: Stripe) GUIDY Data Chunk Chunk Chunk Ch u nk NA1 Sender: GUIDX Backpressure Ch Net Addr Path Quality NA1 4 NA2 36 un k Backpressure Ch Chunk Chunk un NA2 k Receiver: GUIDY File Transfer Completion Time (secs) GUIDY NA1 use both WiFi and LTE use only WiFi use only LTE 400 350 300 250 200 150 100 50 0 5meters/s 10meters/s 20meters/s 30meters/s Speed of Vehicle Splitting Logic Multi-homing technique can also be combined with Network Coding … Data 1000 Aggregate Throughput (Mb) GUIDY NA2 MobilityFirst Multihoming Oracle Application Using only LTE Using only Wi-Fi 800 600 400 200 0 0 10 20 30 40 50 60 70 Time (sec) WINLAB 80 90 100 MF Protocol Example: Enhanced CDN Service Using Compute Layer Feature Enhanced service example – content delivery with in-network caching & transcoding GUID=13247..99 NA31 NA43 GUID=13247..99 MF Compute Layer with Content Cache Service plug-in Filter on SID=128 Content cache at mobile Operator’s network – NA99 GUID=13247..99 NA99 GNRS query Returns list: NA99,31,22,43 NA29 GNRS Query GUID=13247..99 NA22 Content Owner’s Server Data fetch from NA99 Content file Mobile’s GUID Data fetch from NA43 Get (content_GUID, SID=128 - cache service) Get (content_GUID) Query User mobility GUID=13247..99SID=128 (enhanced service) WINLAB MobilityFirst Prototyping: Phased Strategy Phase 2 Phase 1 Content Addressi ng Stack Context Addressi ng Stack Phase 3 Host/Device Addressing Stack Encoding/Certifying Layer Global Name Resolution Service (GNRS) Storage Aware Routing Locator-X Routing (e.g., GUID-based) Context-Aware / Late-bind Routing Prototype Standalone Modules Integrated MF Protocol Stack and Services Evaluation Simulation and Emulation Smaller Scale Testbed Deployable s/w pkg., box Distributed Testbed E.g. ‘Live’ on GENI WINLAB 95 MF Host Protocol Stack ‘Socket’ API open send send_to recv recv_from close App-1 App-2 Linux PC/laptop with WiMAX & WiFi App-3 Context API Network API Context Services E2E Transport GUID Services Network Layer Security Sensors Android device with WiMAX & WiFi Routing User policies Interface Manager ‘Hop’ Link Transport Early Dev. WiFi Integrate WiMAX Device: HTC Evo 4G, Android v2.3 (rooted), NDK (C++ dev) WINLAB 96 MF Click Software Router Lightweight, scalable multicast • GNRS for maintenance of multicast memberships • Heuristic approaches to reduce network load, limit duplicated buffering, and improve aggregate delivery delays • Click prototype, with SID for multicast flows • Evaluating hail a cab application as a example multipoint delivery scenario 97 Multicast Inter-Domain (EIR) WINLAB MF Routing Prototype on ORBIT Click-based prototyping of edge-aware inter-domain routing (EIR) on Orbit nodes Implementation on 200+ nodes on the grid Routing protocol uses “aNode” concept to disseminate full topology and aggregated edge network properties Telescopic NSP (network state packet) advertisement for scalability EIR Click router RIB OSPF w. Telescoping Link state advs NSP GNRSd Binding request SID 3 SID 2 SID 1 NextHop Table EIR forwarding engine Data Packets Data Packet WINLAB OpenFlow/SDN Implementation of MF Protocol stack embedded within controller Label switching, NA or GUID-based routing (incl. GNRS lookup) Controllers interact with other controllers and network support services such as GNRS Flow rule is set up for the remaining packets in the chunk based on Hop ID (which is inserted as a VLAN tag in all packets) E.g., SRC MAC = 04:5e:3f:76:84:4a, VLAN = 101 => OUT PORT = 16 MF Protocol Stack WINLAB 99 MF Multi-Site Deployment on GENI NL R Cambridge, MA Madison, WI Ann Arbor, MI Lincoln, NE Palo Alto, CA N. Brunswick, NJ Salt Lake, UT Tokyo, Japan Los Angeles, CA I2 Atlanta, GA MobilityFirst Routing and Name Resolution Service Sites MobilityFirst Access Net Clemson, SC Long-term (nonGENI) Short-term Wide Area ProtoGENI ProtoGENI WINLAB Web Sites for More Information: WINLAB: www.winlab.rutgers.edu ORBIT: www.orbit-lab.org MobilityFirst: http: mobilityfirst.winlab.rutgers.edu WINLAB 102