Modulation Rate Adaptation in Urban and Vehicular Environments: Cross Layer Implementation and Experimental Evaluation ACM MobiCom 2008 Joseph Camp and.

Slides:



Advertisements
Similar presentations
Problems in Ad Hoc Channel Access
Advertisements

Nick Feamster CS 4251 Computer Networking II Spring 2008
Università degli Studi di Firenze 08 July 2004 COST th MCM - Budapest, Hungary 1 Cross-layer design for Multiple access techniques in wireless communications.
Dept. of computer Science and Information Management
Towards MIMO-Aware n Rate Adaptation (Ioannis Pefkianakis, Suk-Bok Lee and Songwu Lu) Towards MIMO-Aware n Rate Adaptation (Ioannis Pefkianakis,
Medium Access Issues David Holmer
EETS 8316 : WIRELESS NETWORKS Vamsi Krishna Medharametla
CMAP: Harnessing Exposed Terminals in Wireless Networks Mythili Vutukuru Joint work with Kyle Jamieson and Hari Balakrishnan.
Design and Experimental Evaluation of Multi-User Beamforming in Wireless LANs Theodoros Salonidis Technicolor ACM MobiCom 2010 Edward Knightly Rice Narendra.
2005/12/06OPLAB, Dept. of IM, NTU1 Optimizing the ARQ Performance in Downlink Packet Data Systems With Scheduling Haitao Zheng, Member, IEEE Harish Viswanathan,
– Wireless PHY and MAC Stallings Types of Infrared FHSS (frequency hopping spread spectrum) DSSS (direct sequence.
Strider : Automatic Rate Adaptation & Collision Handling Aditya Gudipati & Sachin Katti Stanford University 1.
Comp 361, Spring 20056:Basic Wireless 1 Chapter 6: Basic Wireless (last updated 02/05/05) r A quick intro to CDMA r Basic
Interactions Between the Physical Layer and Upper Layers in Wireless Networks: The devil is in the details Fouad A. Tobagi Stanford University “Broadnets.
PHY layer access misbehavior in WLAN networks Master thesis presentation Radio Communication Systems, KTH Probir Khaskel Advisor: Olav Queseth & Examiner:
CARA: Collision-Aware Rate Adaptation for IEEE WLANs Presented by Eric Wang 1.
Collision Aware Rate Adaptation (CARA) Bob Kinicki Computer Science Department Computer Science Department Advanced Computer.
MAC Layer (Mis)behaviors Christophe Augier - CSE Summer 2003.
Wireless Network Design for Distributed Control Liu and Goldsmith - Appeared at CDC 2004 Presented by Vinod Namboodiri.
Centre for Wireless Communications Opportunistic Media Access for Multirate Ad Hoc Networks B.Sadegahi, V.Kanodia, A.Sabharwal and E.Knightly Presented.
Self-Management in Chaotic Wireless Deployments A. Akella, G. Judd, S. Seshan, P. Steenkiste Presentation by: Zhichun Li.
1 Robust Rate Adaptation in Networks Starsky H.Y, Hao Yang, Songwu Lu and Vaduvur Bharghavan Presented by Meganne Atkins.
The Impact of Multihop Wireless Channel on TCP Throughput and Loss Presented by Scott McLaren Zhenghua Fu, Petros Zerfos, Haiyun Luo, Songwu Lu, Lixia.
1 Introduction to Wireless Networks Michalis Faloutsos.
A Cross Layer Approach for Power Heterogeneous Ad hoc Networks Vasudev Shah and Srikanth Krishnamurthy ICDCS 2005.
Doc.: IEEE /0861r0 SubmissionSayantan Choudhury Impact of CCA adaptation on spatial reuse in dense residential scenario Date: Authors:
MIMO and TCP: A CASE for CROSS LAYER DESIGN Soon Y. Oh, Mario Gerla Computer Science Dept. University of California, Los Angeles {soonoh,
1 How to apply Adaptation principle: case study in
Wireless Networking & Mobile Computing CS 752/852 - Spring 2012 Tamer Nadeem Dept. of Computer Science Lec #7: MAC Multi-Rate.
Page 1 January 2002 doc.: IEEE 802.RR-02/018A-d5 IEEE 802 IEEE 802 proposal relating to DFS and JPT5G proposal.
RTS/CTS-Induced Congestion in Ad Hoc Wireless LANs Saikat Ray, Jeffrey B. Carruthers, and David Starobinski Department of Electrical and Computer Engineering.
CS640: Introduction to Computer Networks Aditya Akella Lecture 22 - Wireless Networking.
Opersating Mode DCF: distributed coordination function
MAC Protocols and Security in Ad hoc and Sensor Networks
Wireless Medium Access. Multi-transmitter Interference Problem  Similar to multi-path or noise  Two transmitting stations will constructively/destructively.
A Simple and Effective Cross Layer Networking System for Mobile Ad Hoc Networks Wing Ho Yuen, Heung-no Lee and Timothy Andersen.
A Cooperative Diversity- Based Robust MAC Protocol in wireless Ad Hoc Networks Sangman Moh, Chansu Yu Chosun University, Cleveland State University Korea,
Computer and Data Communications Semester Mohd Nazri Mahmud Session 4a-12 March 2012.
ECE 256, Spring 2008 Multi-Channel MAC for Ad Hoc Networks: Handling Multi-Channel Hidden Terminals Using A Single Transceiver Jungmin So & Nitin Vaidya.
Joint PHY-MAC Designs and Smart Antennas for Wireless Ad-Hoc Networks CS Mobile and Wireless Networking (Fall 2006)
1 Core-PC: A Class of Correlative Power Control Algorithms for Single Channel Mobile Ad Hoc Networks Jun Zhang and Brahim Bensaou The Hong Kong University.
Wireless Communication Technologies 1 Outline Introduction OFDM Basics Performance sensitivity for imperfect circuit Timing and.
Recitation 8 Wireless Networks. Virtual carrier sensing First exchange control frames before transmitting data – Sender issues “Request to Send” (RTS),
Copyright: S.Krishnamurthy, UCR Power Controlled Medium Access Control in Wireless Networks – The story continues.
MAC Protocols In Sensor Networks.  MAC allows multiple users to share a common channel.  Conflict-free protocols ensure successful transmission. Channel.
Demand Based Bandwidth Assignment MAC Protocol for Wireless LANs K.Murugan, B.Dushyanth, E.Gunasekaran S.Arivuthokai, RS.Bhuvaneswaran, S.Shanmugavel.
Robust Rate Adaptation in networks Starsky H.Y. Wong, Hao Yang, Songwu Lu and Vaduvur Bharghavan UCLA WiNG Research Group and Meru Networks.
Lecture # 13 Computer Communication & Networks. Today’s Menu ↗Last Lecture Review ↗Wireless LANs ↗Introduction ↗Flavors of Wireless LANs ↗CSMA/CA Wireless.
Ubiquitous Computing Center A Rate-Adaptive MAC Protocol for Multi-hop Wireless Networks 황 태 호
Hangguan Shan, Member, IEEE, Ho Ting Cheng, Student Member, IEEE, and Weihua Zhuang, Fellow, IEEE Cross-Layer Cooperative MAC Protocol in Distributed Wireless.
Medium Access Control protocols for ad hoc wireless networks: A survey 指導教授 : 許子衡 報告者 : 黃群凱.
Doc.: IEEE /0065r0 Submission January 2005 Bahareh Sadeghi, Intel CorpSlide 1 On Hidden and Exposed Terminal Problems Notice: This document has.
Recitation 8 Wireless Networks.
Tackling Exposed Node Problem in IEEE Mac Deepanshu Shukla ( ) Guide: Dr. Sridhar Iyer.
An Energy Efficient MAC Protocol for Wireless LANs, E.-S. Jung and N.H. Vaidya, INFOCOM 2002, June 2002 吳豐州.
Universität Karlsruhe Institut für Telematik ECE 591
ECE 256, Spring 2009 __________ Multi-Channel MAC for Ad Hoc Networks: Handling Multi-Channel Hidden Terminals Using A Single Transceiver __________________.
WLAN.
OAR: An Opportunistic Auto- Rate Media Access Protocol for Ad Hoc Networks B. Sadeghi, V. Kanodia, A. Sabharwal, E. Knightly Presented by Sarwar A. Sha.
Medium Access Control in Wireless networks
MAC Layer Protocols for Wireless Networks. What is MAC? MAC stands for Media Access Control. A MAC layer protocol is the protocol that controls access.
Medium Access in Sensor Networks. Presented by: Vikram Shankar.
IEEE Rate Control Algorithms: Experimentation and Performance Evaluation in Infrastructure Mode Sourav Pal, Sumantra R. Kundu, Kalyan Basu and Sajal.
Discovering Sensor Networks: Applications in Structural Health Monitoring Summary Lecture Wireless Communications.
MAC Protocols for Sensor Networks
Trace-based Evaluation of Rate Adaptation Schemes in Vehicular Environments Kevin C. Lee WiVeC 2010, 5/17/10.
Wireless LANs Wireless proliferating rapidly.
Topics in Distributed Wireless Medium Access Control
A Rate-Adaptive MAC Protocol for Multi-Hop Wireless Networks
AccuRate: Constellation Aware Rate Estimation in Wireless Networks
Presentation transcript:

Modulation Rate Adaptation in Urban and Vehicular Environments: Cross Layer Implementation and Experimental Evaluation ACM MobiCom 2008 Joseph Camp and Ed Knightly ACM MobiCom 2008 Joseph Camp and Ed Knightly All figures taken from the paper.

Premise of the Paper  Rate adaptation techniques are used in to-days networks.  How good are these rate adaptation techniques in various settings.  Indoor and Outdoor.  They experimentally evaluate two types of rate adaptation strategies  The first is loss based (reaction to packet loss) -- used in networks.  The second is SNR based (not implemented previously) -- rate changed based on perceived SNR  Rate adaptation techniques are used in to-days networks.  How good are these rate adaptation techniques in various settings.  Indoor and Outdoor.  They experimentally evaluate two types of rate adaptation strategies  The first is loss based (reaction to packet loss) -- used in networks.  The second is SNR based (not implemented previously) -- rate changed based on perceived SNR

Explicit findings  Depending on the scenario, the different protocols behave differently.  The performance is dependent on the coherence time of the channel  Coherence time is the time for which, the channel quality remains unchanged (fade duration remains unchanged).  Some interesting observations that we will discuss.  Depending on the scenario, the different protocols behave differently.  The performance is dependent on the coherence time of the channel  Coherence time is the time for which, the channel quality remains unchanged (fade duration remains unchanged).  Some interesting observations that we will discuss.

Loss Triggered Rate Adaptation  With loss triggered adaptation,, the transmitter interprets the channel state based on time outs  Time-outs suggest failed delivery  Current.11 networks use some form of rate adaptation -- AMRR, SampleRate, Onoe.  The authors consider two loss triggered rate adaptation protocols.  With loss triggered adaptation,, the transmitter interprets the channel state based on time outs  Time-outs suggest failed delivery  Current.11 networks use some form of rate adaptation -- AMRR, SampleRate, Onoe.  The authors consider two loss triggered rate adaptation protocols.

Considered Loss Trigged Adaptation Protocols  Consecutive Packet Decision Loss-Triggered Rate Adaptation (Protocol 1)  Increase the modulation rate after a number of consecutive successful transmission (10) and decrease after a number of failures (2).  Numbers chosen based on previous studies  Two way handshake (no RTS/CTS)  Historical-Decision Loss-triggered Rate Adaptation (Protocol 2)  Window of packets to select modulation rate.  Threshold for increase and decrease based on prior work (specifics in related work)  Two way handshake (no RTS/CTS)  Consecutive Packet Decision Loss-Triggered Rate Adaptation (Protocol 1)  Increase the modulation rate after a number of consecutive successful transmission (10) and decrease after a number of failures (2).  Numbers chosen based on previous studies  Two way handshake (no RTS/CTS)  Historical-Decision Loss-triggered Rate Adaptation (Protocol 2)  Window of packets to select modulation rate.  Threshold for increase and decrease based on prior work (specifics in related work)  Two way handshake (no RTS/CTS)

Collision/Fading Differentiation  Traditional rate adaptation schemes assume that losses are due to fading -- but they could be due to interference.  A previous effort (Wong, Lu, Yang and Bharghavan, MobiCom 06) propose a way to distinguish between losses due to fading/interference.  Use RTS/CTS upon experiencing loss.  If RTS/CTS exchange is successful, loss most likely due to channel effects.  Authors implement a similar scheme in conjunction with loss triggered rate adaptation.  Traditional rate adaptation schemes assume that losses are due to fading -- but they could be due to interference.  A previous effort (Wong, Lu, Yang and Bharghavan, MobiCom 06) propose a way to distinguish between losses due to fading/interference.  Use RTS/CTS upon experiencing loss.  If RTS/CTS exchange is successful, loss most likely due to channel effects.  Authors implement a similar scheme in conjunction with loss triggered rate adaptation.

SNR triggered rate adaptation  With SNR triggered rate adaptation, the receiver measures the signal-to-noise ratio and informs the transmitter via the four way handshake.  In CTS message.  These protocols have not been implemented previously -- not available in the commodity hardware.  Authors use the WARP FPGA radios (implemented by Rice University and sold by Mango Networks) to implement SNR triggered rate adaptation.  With SNR triggered rate adaptation, the receiver measures the signal-to-noise ratio and informs the transmitter via the four way handshake.  In CTS message.  These protocols have not been implemented previously -- not available in the commodity hardware.  Authors use the WARP FPGA radios (implemented by Rice University and sold by Mango Networks) to implement SNR triggered rate adaptation.

SNR triggered schemes considered  SNR triggered Rate Adaptation (Protocol 3)  Signal quality feedback using CTS message.  Four way handshake used.  Equal Air-time Assurance (Protocol 4)  In traditional SNR based schemes, if nodes transmit at higher rates -- they occupy channel for less time.  Since offers equal transmission opportunity to all nodes, high rate links will have to share channel capacity with low rate links.  Thus, even with high rate, you may get low throughput  To overcome this, they give equal air-time to all rates  If a node is transmitting at a high rate, it gets to transmit multiple consecutive packets with a single RTS/CTS Exchange  As before, four way handshake is used.  SNR triggered Rate Adaptation (Protocol 3)  Signal quality feedback using CTS message.  Four way handshake used.  Equal Air-time Assurance (Protocol 4)  In traditional SNR based schemes, if nodes transmit at higher rates -- they occupy channel for less time.  Since offers equal transmission opportunity to all nodes, high rate links will have to share channel capacity with low rate links.  Thus, even with high rate, you may get low throughput  To overcome this, they give equal air-time to all rates  If a node is transmitting at a high rate, it gets to transmit multiple consecutive packets with a single RTS/CTS Exchange  As before, four way handshake is used.

WARP  Three main components  Xilinx Virtex-II FPGA  MAC protocols in C  PHY within the FPGA fabric  MIMO capable radios  up to four antennas  OFDM capable  BPSK, QPSK and 16 QAM are supported.  Ethernet port to report performance of the protocols.  Three main components  Xilinx Virtex-II FPGA  MAC protocols in C  PHY within the FPGA fabric  MIMO capable radios  up to four antennas  OFDM capable  BPSK, QPSK and 16 QAM are supported.  Ethernet port to report performance of the protocols.

What else has been implemented ?  Carrier Sensing  Binary Exponential back-off  NAV -- network allocation vector to facilitate virtual carrier sensing.  Time-outs  Four way handshake - RTS/CTS DATA ACK  Carrier Sensing  Binary Exponential back-off  NAV -- network allocation vector to facilitate virtual carrier sensing.  Time-outs  Four way handshake - RTS/CTS DATA ACK

In Lab Evaluations  Controlled setting -- use of a channel emulator (Spirent Communications) and a signal generator (Agilent ESG-D series)  Channel conditions specified in terms of :  Coherence time  Delay spread -- time between incidence of first multi-path ray to that of the last ray.  Interference  PHY layer capture -- ability to decode signal in presence of noise/interference.  Controlled setting -- use of a channel emulator (Spirent Communications) and a signal generator (Agilent ESG-D series)  Channel conditions specified in terms of :  Coherence time  Delay spread -- time between incidence of first multi-path ray to that of the last ray.  Interference  PHY layer capture -- ability to decode signal in presence of noise/interference. Ideal rate: Modulation rate with which the highest throughput is achieved (exhaustive search)

Impact of Coherence time  Vary coherence time on a single Rayleigh fading channel of high average quality (avg SNR = -40 dBm) 1.For long coherence times, all protocols converge to same throughput -- they can track the channel when there is slow fading. Protocol 3 suffers -- RTS/CTS overhead per high rate transmission. 2.Historical trigger based is best at small coherence times. Other protocols are poor -- for different reasons!

Performance with small coherence times  100  sec coherence time.  SNR protocols overselect  Measure SNR only during RTS -- this may decrease during packet transmission.  In essence, these assume that SNR value is valid throughout packet -- not the case!  Protocol 1 (consecutive loss triggered) underselects.  Consecutive losses common.  Chooses rates that are lower than that possible.  100  sec coherence time.  SNR protocols overselect  Measure SNR only during RTS -- this may decrease during packet transmission.  In essence, these assume that SNR value is valid throughout packet -- not the case!  Protocol 1 (consecutive loss triggered) underselects.  Consecutive losses common.  Chooses rates that are lower than that possible.

Coherence time training for SNR based adaptation  Offline measurements of performance of different modulation schemes with varying coherence time.  Depending on SNR, choose the right rate for either long or short coherence times.  Offline measurements of performance of different modulation schemes with varying coherence time.  Depending on SNR, choose the right rate for either long or short coherence times. Left figure with coherence time 80 milliseconds (long) and right figure 0.8 ms (short). When coherence time is long, with increase SNR use high rates. When coherence time is short, no benefit from using 16 QAM -- highest rate considered.

The training helped!  Allows the choice of the right rates  Performance of SNR based protocols improves.  Allows the choice of the right rates  Performance of SNR based protocols improves.

Impact of multipath fading  SNR protocols are more sensitive to coherence time in the presence of multi-path fading.  Training becomes more critical.  SNR protocols are more sensitive to coherence time in the presence of multi-path fading.  Training becomes more critical.

Impact of external interference  Slow fading channel and packet-sized noise (2 milliseconds).  The idle period between noise instants is varied.  With short idle periods, consecutive packet-decision protocol (Protocol 1) increases underselecting rate.  Historical packet decision less susceptible.  SNR protocols have lower overall throughput -- but choose the right rate based on the measured SNR (due to interference).  Slow fading channel and packet-sized noise (2 milliseconds).  The idle period between noise instants is varied.  With short idle periods, consecutive packet-decision protocol (Protocol 1) increases underselecting rate.  Historical packet decision less susceptible.  SNR protocols have lower overall throughput -- but choose the right rate based on the measured SNR (due to interference).

Evaluating Heterogeneous links  A case with hidden terminals is considered.  The goal is to see how the different rate adaptation protocols work in the different settings.  First, create links A--> B and C--> B equal and of good quality (- 45 dBm)  Then keep the quality of one of the links fixed and then vary that of the other in steps of 5 dB.  Key observation : Due to PHY layer capture, there is a mismatch in achieved throughputs.  The previous protocol that differentiates between collisions and fading (one by Wong, Lu etc.) increases the mismatch in throughputs -- why ?  Increases PHY layer capture  Increases overall throughput though!  A case with hidden terminals is considered.  The goal is to see how the different rate adaptation protocols work in the different settings.  First, create links A--> B and C--> B equal and of good quality (- 45 dBm)  Then keep the quality of one of the links fixed and then vary that of the other in steps of 5 dB.  Key observation : Due to PHY layer capture, there is a mismatch in achieved throughputs.  The previous protocol that differentiates between collisions and fading (one by Wong, Lu etc.) increases the mismatch in throughputs -- why ?  Increases PHY layer capture  Increases overall throughput though!

Measurement results on heterogeneous links Notice that mismatch increases as the difference in quality of links increases. Weaker transmitter has increased losses due to lack of RTS protection -- begins to lower rate and this leads to underselection.

Outdoor experiments  Both residential and downtown Houston Residential urban measurements: densely populated residential neighborhood with dense foliage. Downtown measurements in streets of Houston -- buildings of tens of stories high on each side.

Impact of Environment on Static nodes  First, they characterize the environment -- send UDP traffic of various packet sizes, record SNR variance to determine the coherence time.  Vehicles pass at approx 30 mph.  Coherence time milliseconds to 80 milliseconds on average, in residential and downtown areas.  However, passing cars can drive the coherence time to as low as:  15 milliseconds in the residential area  300  s in the downtown area  Why ? Moving vehicles can cause perturbations in signal quality for short periods of time.  First, they characterize the environment -- send UDP traffic of various packet sizes, record SNR variance to determine the coherence time.  Vehicles pass at approx 30 mph.  Coherence time milliseconds to 80 milliseconds on average, in residential and downtown areas.  However, passing cars can drive the coherence time to as low as:  15 milliseconds in the residential area  300  s in the downtown area  Why ? Moving vehicles can cause perturbations in signal quality for short periods of time.

Performance of different protocols: Residential Areas  Consecutive decision underselects -- presence of mobile scatterers prevents the required 10 consecutive packet successes to raise rate.  Historical decision mechanism overselects -- parameters for window appropriate for indoor -- short for outdoor.  SNR mechanisms work well -- coherence times are long enough.  Consecutive decision underselects -- presence of mobile scatterers prevents the required 10 consecutive packet successes to raise rate.  Historical decision mechanism overselects -- parameters for window appropriate for indoor -- short for outdoor.  SNR mechanisms work well -- coherence times are long enough. 60 second tests Coherence time is long -- multiple packets in duration

Performance of different protocols: Downtown  When coherence time is short -- underselection by loss triggered protocols (due to losses) and overselection by SNR triggered protocols (assume that channel state is stable for the entire packet).  Lower number of received packets compared to residential scenario -- artifact of the channel changing much more quickly.  When coherence time is short -- underselection by loss triggered protocols (due to losses) and overselection by SNR triggered protocols (assume that channel state is stable for the entire packet).  Lower number of received packets compared to residential scenario -- artifact of the channel changing much more quickly. Avg coherence time is 80 ms but as low as 300  s as cars pass.

Impact of Mobility  Goal is to evaluate the rate adaptation accuracy within the two settings  Increased fading and more dynamic channel changes with mobility.  Speeds of 20 Kph  The authors track the per packet variance in SNR to measure channel fading.  Goal is to evaluate the rate adaptation accuracy within the two settings  Increased fading and more dynamic channel changes with mobility.  Speeds of 20 Kph  The authors track the per packet variance in SNR to measure channel fading.

Mobile experiments in residential areas  Mobile node approaches a static node and passes by.  Loss triggered protocols cannot track mobile environments  SNR protocols better adapt.  Mobile node approaches a static node and passes by.  Loss triggered protocols cannot track mobile environments  SNR protocols better adapt.

Interference + Mobility  Rate decisions are affected by interference with loss triggered protocols causing them to underselect.  Rate decisions of SNR based protocols remain ok -- but lower throughput due to interference -- better protection from interference due to four way handshake.  Rate decisions are affected by interference with loss triggered protocols causing them to underselect.  Rate decisions of SNR based protocols remain ok -- but lower throughput due to interference -- better protection from interference due to four way handshake.

Heterogeneous links  Similar to that seen with in-lab experiments.  Collision differentiation with loss triggered protocols can increase throughput imbalance.  With SNR triggered protocols, good quality link achieves higher throughput -- but without equal air time, the SNR based protocol sustains equal throughputs for the longest period.  For results -- see paper.  Similar to that seen with in-lab experiments.  Collision differentiation with loss triggered protocols can increase throughput imbalance.  With SNR triggered protocols, good quality link achieves higher throughput -- but without equal air time, the SNR based protocol sustains equal throughputs for the longest period.  For results -- see paper.

To summarize  Performance of rate adaptation protocols is sensitive to environment and in particular coherence time.  Depending on coherence time, different protocols behave differently -- can either use higher rates than what can be supported (overselect) or use lower rates than what can be supported (underselect).  Indoor calibrations may not be suitable for outdoor settings.  Performance of rate adaptation protocols is sensitive to environment and in particular coherence time.  Depending on coherence time, different protocols behave differently -- can either use higher rates than what can be supported (overselect) or use lower rates than what can be supported (underselect).  Indoor calibrations may not be suitable for outdoor settings.