A crash course in Mesh Networking Victor Bahl Microsoft Research bahl@microsoft.com http://research.microsoft.com/~bahl/ © 2007 Victor Bahl. All Rights Reserved. Notice These lecture notes are for educating. Feel free to incorporate these slides in your presentations but please cite the source on each borrowed slide and as a courtesy to the author please inform him of such use. Do not post a copy of these slides, or slides derived from these on a web site without the author’s written permission. The contents of this deck may change without notice © 2007 Victor Bahl. All Rights Reserved. Foreword • Mobile ad hoc networking and mesh networking is a thriving area of research. The number of solutions & results are simply too large to cover in a short tutorial. • This is not a tutorial on (1) wireless communications (2) MAC protocols, and (3) multi-hop routing protocols. • This is crash course on what I know about building mesh networks. Exhaustive & deep treatment of all existing results is not provided. • These notes are an attempt to describe the main problems & the general idea behind some of the promising solutions. At the end of this tutorial you should have a reasonably good understanding of the state-of-art in mesh networking. • I would appreciate your pointing out any mistakes you may find. © 2007 Victor Bahl. All Rights Reserved. Acknowledgements Some of the work presented in this tutorial is a result of the research conducted in my group at MSR. I want to particularly acknowledge my colleagues Jitu Padhye, Ranveer Chandra, Richard Draves, and Sharad Agarwal in MSR. I consulted with my colleagues in academia as well. Special thanks to Edward Knightly (Rice University), Brian Levine (University of Massachusetts), Suman Banerjee (University of Wisconsin), Bhaskaran Raman (IIT Kanpur), and Yair Amir (Johns Hopkins University) for providing me the details of their testbeds. Finally, thanks to our interns Pradeep Kyasanur (UIUC), Krishna Ramachandran (UCSB) Saumitra M. Das (Purdue) whoes research results are incorporated in this tutorial. © 2007 Victor Bahl. All Rights Reserved. Course Roadmap 1. Mesh Networking & Applications 2. Basics of Radio Frequency Communications 3. Multi-hop Wireless Networking – – Historical background Challenges: Mesh networking with 802.11 4. Handling the Challenges – Capacity Enhancement & Calculation MMAC, SSCH, BFS-CA, HMCP, MUP, Network Coding, conflict graphs, …. – Routing Protocols & Link Quality Metrics RFC 2501, RFC 3626, RFC 3684, RFC 3561, RFC 4728, ETX, LQSR, EXoR, HWMP, ETT, WCETT – Security & Network Management 5. Mesh Deployments – Discoveries & Innovations – MSR’s Mesh, MIT’s RoofNet, IIT’s DGP, Rice’s TFA, UMASS’s DieselNet, UCSB’s Mesh, Madcity’s Mesh, JHU’s SMesh 6. Mesh Networking Standards – IEEE 802.11s (IETF standards covered previously) 7. References © 2007 Victor Bahl. All Rights Reserved. Mesh Networking & Applications Wireless Mesh Networking Definition A wireless mesh network is a peer-to-peer multi-hop wireless network in which participant nodes connect with redundant interconnections and cooperate with one another to route packets. – Unlike Mobile Ad hoc NETworks (MANETs) where routings node are mobile, in mesh networks routing nodes are stationary. – Mesh nodes may form the network's backbone. Other non-routing mobile nodes ("clients") may connect to the mesh nodes and use the backbone to communicate with one another over large distances and with nodes on the Internet © 2007 Victor Bahl. All Rights Reserved. Characteristics of a Mesh Network Mesh Network – – – – – Classic Hub & Spoke Network Grows “organically” Does not require infrastructure support Is fault tolerant Requires distributed management Offers higher capacity (via spatial diversity & power management), but • Too many nodes shared bandwidth may suffer due to interference • Too few nodes route maintenance difficult disconnections possible – Identity and security management is a challenge © 2007 Victor Bahl. All Rights Reserved. The Mesh Networking World Mesh Node Mesh Node Web Content Mesh Node Corporate Data Communication Mesh Node Games Mesh Node Shopping Mesh Node Media Internet Broadband Neighbourhood Traditional Last Mile Territory © 2007 Victor Bahl. All Rights Reserved. Home Mesh Scenario 1: Broadband Internet Access Internet • Last Mile Equipment capital cost The scale of touching / maintaining so many endpoints The physics of running cable large distances over unfriendly terrain Political, social and territorial implications Wireless mitigates these issues but introduces others – – – – • Middle Mile Cost of middle and last miles make physical wired infrastructure not an option in rural areas and many countries – – – – • Backbone Range Bandwidth Spectrum availability Cost & maintenance issues of new hardware / standards Mesh networking makes wireless workable – – Range & bandwidth addressed by shorter links Cost & maintenance addressed by building on commodity standards © 2007 Victor Bahl. All Rights Reserved. Scenario 2: A Community Mesh Network Organic – Participants own the equipment and the network © 2007 Victor Bahl. All Rights Reserved. Community Mesh Network Applications • • • • • • Shared broadband Internet access Neighborhood watch (e.g. video surveillance) Shared media content (e.g. neighborhood DVR) Medical & emergency response Neighborhood eBay (garage sales, swaps) Billboards (babysitter/service recommendations, lost cat, newsletter) Bits produced locally, gets used locally Social interaction • Distributed backup Internet use increased social contact, public participation and size of social network. (social capital - access to people, information and resources) Prof. Keith N. Hampton (author of “Netville Neighborhood Study”) URL: http://www.asanet.org/media/neville.html © 2007 Victor Bahl. All Rights Reserved. Scenario 3: Home Mesh Services Entertainment Entertainment E-Business, Services Pre-Recorded Content Personal Media Broadband • • • Extend Access Point (AP) coverage Better spectrum (re)use greater capacity Automatic discovery, plug-and-play networked home devices: – AV equipment (Cameras, TV, DVD, DVR, satellite/cable) – Phones (Cellular and POTS) – Traditionally disassociated smart devices (PDAs, AutoPC) – Home infrastructure items (Light switches, HVAC controls) © 2007 Victor Bahl. All Rights Reserved. Scenario 4: Blanket City-wide Wireless Coverage Philadelphia picks Earthlink for City Wireless, TechNew World, October 5, 2005 San Francisco Keeps Pushing City Wide WiFi, CNET News.com, August 17, 2005 “San Francisco Mayor Gavin Newsom wants to make Wi-Fi coverage in the city as ubiquitous as the fog that blankets its neighborhoods.” Wi-Fi Hits the Hinterlands, BusinessWeek Online, July 5, 2004 “Who needs DSL or cable? New “mesh” technology is turning entire small towns into broadband hot spots”, Rio Rancho N.M., population 60,000, 500 routers covering 103 miles2 NYC wireless network will be unprecedented, Computerworld, June 18, 2004 “New York City plans to build a public safety wireless network of unprecedented scale and scope, with a capacity to provide tens of thousands of mobile users” Rural Areas need Internet too! Newsweek, June 7, 2004 Issue “EZ Wireless built the country's largest regional wireless broadband network, a 600-square-mile Wi-Fi blanket, and activated it this February”, Hermiston, Oregon, population 13,200, 35 routers with 75 antennas covering 600 miles2 Mesh Casts Its Net, Unstrung, January 23, 2004 “Providing 57 miles2 of wireless coverage for public safety personnel in Garland Texas” PCCW takes Wireless Broadband to London, The Register, September 2, 2005 “Prices for the service in UK start from £10 / month for 256 Kbps to £18 /month for 1 Mbps” Anacapa and Firetide Bring Free Wireless Internet to La Semaine Italienne in Paris, France , Business Wire, 24 May, 2005 Bell Canada and Nortel Networks launch Project Chapleau, designed to evaluate broadband in rural Canada, Optical Networks Daily, 18 July 2005 © 2007 Victor Bahl. All Rights Reserved. Scenario 5: All-Wireless Office • Older buildings • No wires • For small offices (~100 PCs) • No switches • No APs – Rapid deployment – Low cost – Short-term offices • Not a replacement for wire © 2007 Victor Bahl. All Rights Reserved. Scenario 6: Spontaneous “Mesh” Definition A temporary ad-hoc multihop wireless network for exchanging voice, video or data, for collaboration in a locally distributed environment, when no permanent infrastructure or central control is present. Usually between portable wireless devices. 1. Peer Calling & Party Lines – P2P calling within local groups – conferences, events, school campus,… 2. Public Safety Fire and rescue teams need ad-hoc communication at incident sites 3. Real Time Advisory Drivers need traffic information and advisories generated in real time © 2007 Victor Bahl. All Rights Reserved. Grass Roots Mesh Deployments Academia – The Roofnet Project (MIT, USA) - http://pdos.csail.mit.edu/roofnet/doku.php 802.11 mesh network for broadband IA in cities – The CITRIS TIER Project (UC Berkeley, USA) - http://tier.cs.berkeley.edu/ Technology and Infrastructure for emerging regions – The Digital Gangetic Plains Project (IIT Kanpur, India) - http://www.iitk.ac.in/mladgp 802.11-based low-cost Networking for rural India – The TFA Project (Rice University, USA) - http://taps.rice.edu/index.html Technology for All Project – …. Community Mesh Networks – Community Network Movement - http://www.scn.org/commnet/ – Seattle Wireless - http://www.seattlewireless.net/ – Champaign-Urbana Community Wireless Network - http://www.cuwireless.net/ – Kingsbridge Link, U.K. - http://www.kingsbridgelink.co.uk/ – …. © 2007 Victor Bahl. All Rights Reserved. Industry Breakdown Infrastructure Based Infrastructure-less Internet UNIVERSITY Internet 101 Bus Stop 206 Poletop Radio Gas Station (Internet TAP) Mesh Router 7 EXIT 90 Mesh Router 5 Mesh Router 2 Mesh Router 3 MeshStreet Zone Any SkyPilot, Flarion, Motorola (Canopy) Invisible Networks, RoamAD, Vivato, Arraycomm, Malibu Networks, BeamReach Networks, NextNet Wireless, Navini Networks, etc. Mesh Router 1 Mesh End Device End Device (Guest to Router 1) Meshnetworks Inc.,Radiant Networks, Invisible Networks, FHP, Green Packet Inc., LocustWorld, etc. Architecture effects design decisions on Capacity management, fairness, addressing & routing, mobility management, energy management, service levels, integration with the Internet, etc. © 2007 Victor Bahl. All Rights Reserved. Industry Deployment Scenarios http://www.unstrung.com/insider/ March 2005, Source: Unstrung Insider © 2007 Victor Bahl. All Rights Reserved. What about WiMAX? IEEE 802.16d for developing/rural use (.16e targets mobile scenarios) – Still needs market momentum around hardware optimisation: size, power, efficiency and most important—cost WiMAX as a last-mile solution? – In low-density areas, WiMAX requires high-power towers or lots of towers: (=> cost goes up) – In NLOS environments, range impacts bandwidth through reduced modulation – WiMAX CPE expensive in next 3-5 years (~ $150-250) WiMAX feeding a mesh can be a good solution – Mesh extends WiMAX tower reach – Mesh simplifies the financials by greatly reducing equipment cost – Mesh is robust and deal with network vagaries © 2007 Victor Bahl. All Rights Reserved. WiMAX + Mesh WiFi Meshes can add value to WiMAX in several ways: – Reduce CPE costs – Extend range of WiMAX tower without compromising speed – Replace high-price WiMAX towers with cheaper mesh nodes WiMAX Only WiMAX with Mesh 16 QAM 8PSK QPSK FSK A © 2007 Victor Bahl. All Rights Reserved. 16 QAM 16 QAM 16 QAM 16 QAM A Radio Frequency Communications Basics © 2007 Victor Bahl. All Rights Reserved. RF Communications Radio Frequency (RF) waves are effected by – Distance between the transmitter and receiver Inverse power law – Reflection (e.g. ground reflection) – Diffraction (e.g. from building) – Scattering (e.g. from trees) Path Loss Multipath Fading – Links may not be bi-directional A can hear B, but B can’t hear A (e.g. because of receiver sensitivity) – Radio waves may be blocked (absorbed) by objects e.g by buildings, humans, rain, walls, glass windows Shadowing Degree of attenuation generally depends on frequency © 2007 Victor Bahl. All Rights Reserved. Radio Frequency Wave Propagation 101 A P 101 Access Point 1 House House House House Building 1 A P Access Point 1 House Tree Tree House Tree Tree Tree Tree Condos Tree Tree Tree Tree Condos Hospital Tree Gas station Shadowing Multipath Fading - caused by multiple reflections - caused by physical obstructions © 2007 Victor Bahl. All Rights Reserved. Multipath Fading What causes multipath fading? – – – – Transmitter radiates power in many directions Receiver collects power from many directions Signals are reflected by various objects Many different paths exists between transmitter & receiver Results in – Delay spreads Signals along different paths arrive at different times One “symbol” bit may overlap with another – Time varying signal amplitudes at receiver Types of Fading – Fading can be fast or slow (rapid fluctuation of signal) – Fading can be flat or frequency-selective – All four combinations possible © 2007 Victor Bahl. All Rights Reserved Multipath is not all bad • The multipath phenomena can be exploited to increase data throughput and range by reducing bit error rates • From information theory, channel capacity increase is proportional to the number of transmit and receive antennas – MIMO (Multiple-Input Multiple-Output) • MIMO techniques are the key reason for higher data rates in IEEE 802.11n and IEEE 802.16 standards • Signals arriving via different paths are picked up at different strengthlevel by the different antennas and combined to form a cleaner & stronger signal – The correlation between antennas must be de-couple by DSP algorithms Statistical Models for Propagation Effects – Rayleigh Fading Each reflected wave may have a different phase Signals arriving out of phase may cancel each other Models RF propagation in heavily built-up urban environments – Rician Fading Essentially like Rayleigh, except that there is one dominant (line of sight) wave Models RF propagation inside buildings – Doppler Shifts Movement creates shifts in frequency Different path lengths have different shifts Introduces random frequency modulation at receiver Challenge in using these models is what parameters should you use © 2007 Victor Bahl. All Rights Reserved. Effect of Frequency At higher frequency – signals are absorbed more rapidly by water in the air attenuate faster – signals are blocked by objects do not refract; leave a complete shadow behind obstacles lower frequency signals refract (bend) around obstacles Friis Transmission Equation Free-space path loss Lfs 32.44 20 log d 20 log 10f GT GR Where, Lfs is the loss in dB; f is the frequency in MHz; GT & GR are the transmitter & reciever antenna gain in dBi; and d is the distance in Km at which the loss is calculated © 2007 Victor Bahl. All Rights Reserved. Range Calculations Link budget calculations for line of sight communication with free space loss PR PT Lfs LT LR where PR & PT are received & transmitted powers in dBm; Lfs is path loss; LT & LR are signal loss at the transmitter & receiver in dB • When the receiver sensitivity is known, link budget calculations can provide an estimate of the range. • The formula captures the fact that range is effected by frequency, distance, transmission power, antenna gains, and losses at the transmitter and receiver. • For non-line of sight communications (e.g. indoor). The formula has to be modified [out of scope for this tutorial] © 2007 Victor Bahl. All Rights Reserved. Should know “The Decibel” • The dB is a dimensional-less logarithmic unit used to describe a ratio. • In RF communications dB used to describe the ratio between two measurements of power. Mathematically, the difference in decibels between two powers = 10 log10 (P2/P1) • dBm (or dBmW) is the absolute unit of measured power referenced to 1 mW – 3 dBm is 2 (= 103/10) mW – 0.01 mW is -20 (= 10log10(0.01)) dBm • The receiver sensitivity of typical 802.11 wireless cards ranges between -60 dBm and -80 dBm, i.e. they can decode RF signals received at power levels between 0.000001 mW and 0.0000001 mW © 2007 Victor Bahl. All Rights Reserved. Summing up observations about the RF channel • Radio channel exhibits unpredictable behavior – Signal strength fluctuates rapidly and widely – Error rates are high – Fading varies with time and causes attenuation • Understanding it’s characteristics leads to better system design (e.g. MIMO, Routing protocol design,…) • Statistical models should be used in simulations • Interference is hardest to deal with and can limit performance severely • Transmission power control & frequency reuse are important techniques for capacity enhancement © 2007 Victor Bahl. All Rights Reserved. Channel Access Protocols: Wired vs. Wireless Wired networking protocols such as Ethernet perform poorly when used in wireless communication – Why? Because of media dependent differences Should know: – Hidden terminal problem – Exposed terminal problem – Collision detection problem – Interference problem – Deaf-time problem Turn-around time between Rx/Tx can have significant effect on the design of the MAC layer © 2007 Victor Bahl. All Rights Reserved. The Hidden Terminal Problem Consider the following scenario – – – – Tobagi-Comm-1975 B is within range of A and C; A and C are out of range of each other i.e. A and C are hidden from each other A sends a packet to B C sends a packet to B The packets collide at B results in reduction of throughput CSMA doesn’t work – C can’t know that it has to wait since it can’t hear A B A C Solution: RTS/CTS - with intended transmission duration Karn-CNC-1997 © 2007 Victor Bahl. All Rights Reserved. The Exposed Terminal Problem Consider the following: – C is within range of B’s transmitter – B sends a packet to A – C wants to send a packet to D but doesn’t, since C is exposed to B (Note: transmitter B must be protected because it expects an ACK from A) CSMA doesn’t work – C thinks it should wait since it can hear B – C waits needlessly results in reduction of throughput A C B Can transmit, but doesn’t D © 2007 Victor Bahl. All Rights Reserved. The Collision Detection Problem No collision detection in wireless communications In wireless can’t listen while you send – Generally hardware is not flexible enough – All you hear is your own signal Your own signal at your antenna is much stronger than anyone else’s signal n do – The power law Pr Pt dr Consequently, – wireless can’t do collision detect like Ethernet © 2007 Victor Bahl. All Rights Reserved. Things to know about Spectrum Use (US View) • WLANs operate in the 2400 – 2483.5 MHz (2.4 GHz), 5150 – 5250 MHz (lower U-NII), 5250 – 5350 MHz (mid U-NII), and 5725 – 5825 MHz (upper U-NII) bands under FCC’s Part 15 “unlicensed” rules • Part 15 operation is subject to no harmful interference caused to; and interference accepted from, licensed services, other unlicensed devices, ISM equipment, and incidental radiators • IEEE Standards operating in these bands include: – – – • Other Part 15 devices and Part 18 ISM devices that also use these bands: – – • 802.11{b,g} in the 2.4 GHz band; 802.11a in the U-NII band 802.15 WPAN (Bluetooth) in the 2.4 GHz band 802.16 WirelessHUMAN in the mid and upper U-NII bands Field disturbance sensors, cordless telephones, low power devices, and microwave ovens Non-802.11 Part 15 devices: cordless telephones, A/V repeaters, security cameras, baby monitors, & digital data links Licensed services that operate in these bands include: – Amateur radio in the 2.4 GHz and upper U-NII bands; fixed microwave in the 2.4 GHz band; and satellite in the lower U-NII band © 2007 Victor Bahl. All Rights Reserved. 2.4 GHz Band Allocation (“The junk” band) 802.11b&g WLAN 802.15 – Bluetooth WPAN 15.209 – Generic Unlicensed 15.245 – Field Disturbance Sensors 15.247 – Spread Spectrum 15.249 – Low Power Part 18 – ISM Part 90 – Private Land Mobile Part 97 – Amateur Part 101 – Fixed Microwave Government 2400 MHz 2450 MHz © 2007 Victor Bahl. All Rights Reserved. 5.3 GHz Band Allocation (US View) 5.3 GHz Band 802.11a – WLAN 802.16a – WirelessHUMAN 15.209 – Generic Unlicensed 15.4XX – U-NII Part 25 – Satellite Part 87 – Aviation Government 5150 MHz 5200 MHz 5250 MHz 5300 MHz © 2007 Victor Bahl. All Rights Reserved. 5350 MHz 5.8 GHz Band Allocation (US View) 802.11a – WLAN 802.16a – WirelessHUMAN 15.209 – Generic Unlicensed 15.245 – Field Disturbance Sensors 15.247 – Spread Spectrum 15.249 – Low Power 15.4XX – U-NII Part 18 – ISM Part 97 – Amateur Government 5725 MHz 5775 MHz © 2007 Victor Bahl. All Rights Reserved. 5825 MHz The Interference Problem Following rules and regulations but…. 802.11 in presence of BT Phone on TCP download from a 802.11 AP Panasonic 2.4GHz Spread Spectrum Phone 5 m and 1 wall from receiver Performance worsens when there are large number of short-range radios in the vicinity Badly written rules: Colliding standards © 2007 Victor Bahl. All Rights Reserved. Things to know about IEEE 802.11{b,g} 22 MHz channels, K = 1 … 11 – Channels are 5 MHz apart. Center frequencies: 2412 + 5 (K-1) MHz – Channels 1, 6, 11 are non-overlapping (“orthogonal”) Transmit power limit is 1 Watt (30 dBm) .11g data rates Data Rate (Mbps) Modulation Sensitivity (dBm) 1, 2 DSSS -80 5.5 CCK -80 Sensitivity (dBm) 6 BPSK -82 9 BPSK -81 .11b data rates Data Rate (Mbps) Modulation 1, 2 DSSS -80 11 CCK -76 5.5 CCK -80 12 QPSK -79 11 CCK -76 18 QPSK -77 24 4-QAM -74 36 4-QAM -70 48 16-QAM -66 54 16-QAM -65 © 2007 Victor Bahl. All Rights Reserved. Things to know about IEEE 802.11a Channelization & Data Rates Channel (20 MHz width) Center Freq (MHz) 40 mW (2.5 mW/MHz) 36 40 44 48 5180 5200 5220 5240 200 mW (12.5 mW/MHz) 52 56 60 64 5260 5280 5300 5320 149 153 157 161 5745 5765 5785 5805 Power Limits* U-NII Lower Band (5150 – 5350 MHz) U-NII Mid Band (5250 – 5350 MHz) U-NII Upper Band (5725 – 5825 MHz) 800 mW (50 mW/MHz) Data Rate (Mbps) Modulation Sensitivity (dBm) 6 BPSK -82 9 BPSK -81 12 QPSK -79 18 QPSK -77 24 16-QAM -74 36 16-QAM -70 48 64-QAM -66 54 64-QAM -65 Range: Path loss at 5765 MHz is 20 x log10[5765/2437] = 7.5 dB greater than at 2437 MHz independent of distance (Remember Friis’s formula!) © 2007 Victor Bahl. All Rights Reserved. Summarizing: Wireless Communications is Hard Unlicensed spectrum is not pristine Current spectrum etiquette rules are insufficient Available spectrum is limited => Capacity is limited Shannon’s Law: can pack only n bits / m Hz (Hz is not getting bigger) Transmit power is strictly regulated Limits the operational range Channel is unpredictable Wide & rapid fluctuation of signal strengths, high error rates, and attenuation depends heavily on environment (fading, multipath, absorption,…) Wireless is a broadcast channel Snooping is easy Battery technology is not following Moore’s law Limits usage, battery power doubles once every 35 years! Node mobility stresses protocols © 2007 Victor Bahl. All Rights Reserved. Powers-Proc-1995 Solution Space is Large • • • • • • Microelectronics Modulation schemes Error correction Signal processing Smart antennas Energy conservation • • • • • • Channel access protocols Routing protocols Transport protocols Operating system support Application adaptation Human interface A simple tweak in any one of these results in a paper © 2007 Victor Bahl. All Rights Reserved. A word about the Integrity & Credibility of Mesh Simulations Andel-Computer-2006 Is it repeatable? – Document all setting; provide freely available code, models & app. traffic Is it statistically valid? – Use the Central Limit Theorem to determine the no. of independent runs; Address sources of randomness (e.g. pseudo number generator); delete transient values or eliminate by preloading routing tables, queues etc. Is the radio model reasonable? – Use two-ray and shadowing models rather than free-space model Did you use the correct traffic type? – Traffic generation should be based on target applications, CBR may be incorrect Did you perform sensitivity analysis? – Use analysis of variation (ANOVA) technique to understand significance of a particular parameter setting Did you validate against real implementation? – When possible, validate simulation runs on real testbed, even if at a smaller scale Did you try to compare? – Use simulations to understand general behavior, not to compare protocols © 2007 Victor Bahl. All Rights Reserved. Multi-Hop Wireless Networking Historical Perspective Packet Radio Network (PRNET), 1972-1982 • Band: 1718.4-1840 MHz; Power: 5 W; Range: 10 km; Speed:100-400 Kbps, Addressing: Flat; Routing: Distance Vector; Scale: 50+ Survivable Adaptive Networks (SURAN), 1983-1992 • Band: 1718.4-1840 MHz; Power: 5 W; Range: 10 Km; Speed: 100-400 kbps, Addressing: Hierarchical; Routing: Distance Vector; Scale: 1000+ (Low cost packet radio) Global Mobile Information Systems (GLOMO), 1995-2000 • e.g. NTDR, Band: 225-450 MHz; Power: 20 W; Range: 11-20 Km; Speed: 300 kbps, Addressing: Flat; Routing: Link-state / 2-level clusters; Scale: 400+ IETF Mobile Ad Hoc Networks (MANET) Working Group, 1997 – • RFC 2501 (Eval), RFC 3561 (AODV), RFC 3626 (OLSR), RFC 3684 (TBRPF), Drafts – DSR, DYMO, Multicast, OLSRv2 PRNET Van IEEE 802.11s Working Group, 2004 - © 2007 Victor Bahl. All Rights Reserved. Challenge: Mesh Networking with IEEE 802.11 When does a viable mesh form? Problem Formulation How many nodes have to sign up before a viable mesh forms? Answer depends on – – – – Definition of “viable” Wireless range Neighborhood topology Probability of participation Example Scenario: Community Mesh Network Viable mesh: group of at least 25 homes that form a connected graph Topology: A North Seattle Neighborhood. 8214 houses, 4Km x 4Km Wireless range: 50, 100, 200 and 1000 meters Houses decide to join at random, independent of each other. We consider 0.1% to 10% participation rates. © 2007 Victor Bahl. All Rights Reserved. Mesh Formation Results • 5-10% subscription rate needed for suburban topologies with current wireless ranges (> 100 m) • Once a mesh forms, it is usually well-connected – i.e. number of outliers are few (most nodes have > 2 neighbors) • Need to investigate other joining models • Business model considerations are important Increasing range is key for good mesh connectivity © 2007 Victor Bahl. All Rights Reserved. But…The Range Problem 5 GHz: • Bandwidth is good, • Published 802.11a ranges (Yellow circles) decent • Measured range (red circle) poor • Range is not sufficient to bootstrap mesh until installed % is quite high The Antenna Problem Need software steerable antennas to reduce interference & increase range Challenges Advantages – Directional high gain antenna makes long-range meshes viable. – Topology control minimizes interference – High gain, low sidelobe design may enable indoor placement – Exacerbates hidden terminal problem Solution requires control channel and omni antenna – Must change FCC reqs for ERP Currently set up to preclude neighborhood mesh – Cost increment over omni Separate band/radio for control High gain antenna is more expensive – Software steerable © 2007 Victor Bahl. All Rights Reserved. Antenna Evaluation – Sample Study Study Objective Can directional antennas reduce interference, increase spatial reuse, and increase capacity in indoor meshes Antenna characteristics – 6.5// diameter; 3// height; 1.85 lbs – Continuously steerable via software on a PC – Gain: 5.5 - 6.0 dBi; Frequency 2.4-2.497 GHz; – Azimuth Beamwidth: 50o ; Elevation Beamwidth: 90o – Beam position transition time 1 msec © 2007 Victor Bahl. All Rights Reserved. Indoor Performance of Directional Antenna Interior effectiveness severely limited by multi-path © 2007 Victor Bahl. All Rights Reserved. Indoor Performance of Directional Antenna Interior effectiveness severely limited by multi-path © 2007 Victor Bahl. All Rights Reserved. Indoor Performance of Directional Antenna Interior effectiveness severely limited by multi-path © 2007 Victor Bahl. All Rights Reserved. Indoor Performance of Directional Antenna Interior effectiveness severely limited by multi-path © 2007 Victor Bahl. All Rights Reserved. Indoor Performance of Directional Antenna Interior effectiveness severely limited by multi-path © 2007 Victor Bahl. All Rights Reserved. Implications of Antenna Study Study Objective: Can directional antennas reduce interference, increase spatial reuse, and increase capacity in indoor meshes Results: – Wide azimuth beamwidth (50o) antenna do not work when placed indoors reflections and multipath compromise directional gain – MIMO is better for indoor and dense apartment complexes – Antenna placement is critical – We have no 5 GHz antennas to experiment with (FCC rules?) – We don’t have good antenna solutions yet © 2007 Victor Bahl. All Rights Reserved. The MAC Problem – Packets in Flight Example RTS RTS RTS RTS 1 2 3 RTS 4 5 6 7 8 9 10 CTS CTS Backoff window doubles 2 packets in flight! Only 4 out of 11 nodes are active…. © 2007 Victor Bahl. All Rights Reserved. 11 Throughput: Internet Gateway Example Internet RTS RTS RTS CTS Backoff window doubles © 2007 Victor Bahl. All Rights Reserved. Backoff window doubles The Scheduling Problem Conflict graph A B A A B B D D A C B D C E D E 1 Choice 1 Choice 2 D yes A B 0 1 1 D C D A B A D D yes B A C A B E E B D D B C A B A E If future traffic is not known, which one do you schedule first? © 2007 Victor Bahl. All Rights Reserved. The Fairness Problem A 1 B C 2 D Jinyang-MobiCom-2001 Information Asymmetry – A & C do not have the same information C knows about flow 1 (knows how to contend) A does not know about flow 2 – Flow 2 always succeeds, Flow 1 suffers When RTS/CTS is used – A’s packets are not acknowledged by B » A times out & doubles it’s contention window When RTS/CTS is not used – A’s packet collide at B, but Flow 2 is succesful » A times out & double it’s contention window – Downstream links suffer © 2007 Victor Bahl. All Rights Reserved. Gambiroza-MoiCom-2004 The Fairness Problem (2) ITAP Camp-DC-2005 • Location closest to gateway gets the more packets • Nodes farthest from the gateway get very little bandwidth and can get starved • Possible solution: Rate control on each node with fairness in mind – Need topology & traffic information to calculate fair amount – Global vs. distributed solution © 2007 Victor Bahl. All Rights Reserved. The Fairness Problem (3) MAC attempts to provide fairness at packet level not flow level Capture phenomena Winner of competing flows has a higher chance of winning contention again Different levels of interference at different links (different neighborhood) Highly interfered flows can be drowned Nandagopal-MobiCom-2000 Qiu-MSRTR-2003 A C B F1 F2 O P D F3 Q E F F4 R G F5 S T U Flow1 Flow2 Flow3 Flow4 Flow5 2.5 Mbps 0.23 Mbps 2.09 Mbps 0.17 Mbps 2.55 Mbps Active area of research - MACAW, WFQ, DFS, Balanced MAC, EBF-MAC, PFCR, …. © 2007 Victor Bahl. All Rights Reserved. The Path Length Problem Experimental Setup Impact of path length on throughput • 23 node testbed 10000 9000 One IEEE 802.11a radio per node (NetGear card) • Randomly selected 100 senderreceiver pairs (out of 23x22 = 506) 8000 Throughput (Kbps) • 7000 6000 5000 4000 3000 2000 1000 0 • 3-minute TCP transfer, only one connection at a time 0 1 2 3 4 5 6 Byte-Averaged Path Length (Hops) If a connection takes multiple paths over lifetime, lengths are byte-averaged Total 506 points. © 2007 Victor Bahl. All Rights Reserved. Robert Morris’s Rooftnet MSR Mesh Summit 2004 Presentation The Collision Problem Actual Roofnet b/w is often much lower Multi-hop collisions cut b/w by about 2x Expected multi-hop b/w based on single-hop b/w The Node Density Problem Round trip delay versus node density Average RTT avg_rtt = 0.1*curr_sample + 0.9*avg_rtt One sample every 0.5 seconds 0.2 0.18 0.16 Average RTT 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0 20 40 60 80 100 120 140 160 180 Time A new 100Kbps CBR connection starts every 10 seconds, between a new pair of nodes. All nodes hear each other. © 2007 Victor Bahl. All Rights Reserved. The Power Control Problem Tight power control reduces interference and increases throughput A D B C – A & B do not detect RTS/CTS exchange between C & D – B does not detect data transmission from D to C – B’s transmission to A results in packet collision at C © 2007 Victor Bahl. All Rights Reserved. The Power Control Problem (2) Tight power control reduces interference & increases overall throughput A D B C But it also disconnects the network. So what’s the “right” power control algorithm? A D B C © 2007 Victor Bahl. All Rights Reserved. The Capacity of Mesh Nodes What is the maximum achievable capacity of a mesh network with N nodes? Gupta-IEEEIT-2000 Optimal Case Average Case – Nodes are optimally located, destinations are optimally located – Randomly located nodes and destinations – Traffic pattern are random – Traffic patterns are fixed – Optimally spatio-temporal scheduling, routes, ranges for each transmission – As n each node obtains 1 bits/sec O – Each node chooses same range – Each n node obtains 1 bits/sec O n © 2007 Victor Bahl. All Rights Reserved. n log n The Capacity Calculation Problem Gupta and Kumar 2000 – Theorem for stationary ad hoc nodes in the worst case traffic scenario Determines asymptotic, pessimistic bounds on performance – Every node in the mesh is active (either transmitting or receiving) Does not answer: – What is the capacity of a mesh which is using multiple channels, directional antennas, tight power control? © 2007 Victor Bahl. All Rights Reserved. What is the Real Capacity of a Chain? …but the radio’s interferance range is > radios communication range Source 1 Destination 2 3 4 5 6 With Ideal MAC, Chain Utilization = 1/3 With interferences, Chain Utilization = 1/4! Jinyang-MobiCom-2001 ….but this is achievable only with optimum scheduling and optimum offered load!, …with random scheduling and random load, utilization ~ 1/7 ! © 2007 Victor Bahl. All Rights Reserved. The Path Selection Metric Problem Several link quality metrics to select from – Hop count – – – – – Round trip time Packet pair Expected data transmission count incl. retransmission Weighted cumulative expected transmission time Signal strength stability – – – – Energy related Link error rate Air Time … Which to select? We still don’t have a interference-aware metric! We still don’t know how to measure interference….. © 2007 Victor Bahl. All Rights Reserved. Baseline comparison of Metrics Single Radio Mesh Experimental Setup Median path length: HOP: 2, ETX: 3.01, RTT: 3.43, PktPair: 3.46 • 23 node testbed 1600 One IEEE 802.11a radio per node (NetGear card) • Randomly selected 100 sender-receiver pairs (out of 23x22 = 506) • 3-minute TCP transfer, only one connection at a time Draves-MobiCom-2004 1400 Median Throughput (Kbps) • 1200 1000 800 600 400 200 0 HOP ETX RTT PktPair ETX performs the best © 2007 Victor Bahl. All Rights Reserved. Baseline Comparison of Metrics Two Radio Mesh Draves-SIGCOMM-2004 Experimental Setup Median path length: HOP: 2, ETX: 2.4, WCETT: 3 • 23 node testbed Median Throughput of 100 transfers • 3-minute TCP transfer • Two scenarios: – Baseline (Single radio): 802.11a NetGear cards – Two radios 802.11a NetGear cards 802.11g Proxim cards 3500 2989.5 3000 Throughput (Kbps) • Randomly selected 100 sender-receiver pairs (out of 23x22 = 506) Single Radio Two Radios 2500 2000 1601 1379 1500 1508 1155 844 1000 500 0 WCETT ETX Shortest Path WCETT utilizes 2nd radio better than ETX or shortest path © 2007 Victor Bahl. All Rights Reserved. But with different traffic pattern…. 226 Trace Capture • 1 workstations connected via Ethernet • Traces captured during 1-month period 227 Erickson-MobiSys-2006 225 219 EL32 Trace Replayed • Testbed of 22 mesh computers in office environment • 2 IEEE 802.11a/b/g cards per computer 215 217 220 218 216 214 211 210 209 UP DN 208 207 204 205 ~ 76 m 206 203 202 ~ 32 m Similar performance © 2007 Victor Bahl. All Rights Reserved. 201 The Multicast Problem A multicast group is defined with a unique group identifier. Nodes may leave or join the group anytime In wired networks physical network topology is static In ad hoc multi-hop wireless networks physical topology can change often – Need to Integrate with unicast routing protocols Many proposals [Tree-based, Mesh-based, Location-based] which one to use? - ABAM (On-Demand Associatively-Based Multicast) - FGMP (Forwarding Group Multicast Protocol) - ADMIR (Adaptive Demand-Driven Multicast Routing) - LAM (Lightweight Adaptive Multicast) - AMRIS (Ad hoc Multicast Routing utilizing Increased id-numberS) - MAODV (Multicast AODV) - DCMP (Dynamic Core Based Multicast Routing) - MCEDAR (Multicast CEDAR) - AMRoute (Adhoc Multicast Routing) - MZR (Multicast Zone Routing) - CAMO (Core-Assisted Mesh Protocol) - ODMRP (On-Deman Multicast Routing Protocol) - CBM (Content Based Multicast) - SPBM (Scalable Position-Based Multicast) - DDM (Differential Destination Multicast) - SRMP (Source Routing-based Multicast Protocol) - DSR-MB (Simple Protocol for Multicast and Broadcast using DSR) - … -… © 2007 Victor Bahl. All Rights Reserved. The Interference Detection Problem When two systems operate on overlapping frequencies, there exists a potential for harmful interference between them – Performance degradation on both systems Conflict graph is determined by the “Interference Graph” To determine the Interference Graph, require – Knowledge of packet transmission from nodes that are not “visible” – Knowledge of physical location of nodes within the network – Knowledge of whether or not multiple transmissions increase ot decrease interference? Interference Graph can change – as rapidly as the environment – when a node leaves or join the network © 2007 Victor Bahl. All Rights Reserved. The Transport Layer Problem • Majority of the Internet traffic is TCP • Packet losses & delays in wireless can occur due to – Environmental fluctuations resulting link failures – Stochastic link performance due to rapidly changing error rates – CSMA/CA assumes loss is due to congestion and back-off’s • TCP assumes packet losses are due to congestion – Times out when no ACK is received – Invokes slow start, when instead the best response would be to retransmit lost packets quickly • RTT calculation can change as rapidly as the environment (link) changes Can we solve this problem without changing the end-to-end semantics? © 2007 Victor Bahl. All Rights Reserved. The Security Problem Two type of attackers: – External malicious node (no crypto keys) – Compromised node (attacker captures legitimate node and reads out all cryptographic information) Attacks – Selfish behavior, do not forward other node’s packets – Denial of Service (DoS) Jamming Resource consumption attack Routing disruption (e.g. Wormhole attack) Inject malicious routing information Hu-MobiCom-2002 Ongoing Research – Possible solutions: SEAD, Ariadne, SRP, CONFIDANT, … Bucheggar-MobiHoc-2002 © 2007 Victor Bahl. All Rights Reserved. The Spectrum Etiquette Problem Local behavior affects Global Performance! Doesn’t care Node D Node E 100 meters Node A Node B 200 meters Node C 200 meters 120 Packets get dropped! Normalized Percentage 100 80 60 40 20 0 Base One TCP © 2007 Victor Bahl. All Rights Reserved. 10% Drop rate Consequently we….. Must Increase Range and Capacity Single radio meshes built on 802.11 technologies are not good enough. We must extend the range of radios; we must understand the achievable capacity in an ideal wireless mesh and we must build technology to approach this capacity? Must Improve Routing Performance Routing protocols based on shortest-hop are sub-optimal. We must build a routing protocol that adapts quickly to topology changes, incorporates wireless interference and link quality. Must Provide Security and Fairness Is it possible to ensure fairness and privacy for end-users and security for the network? We must ensure that no mesh nodes starves and that the mesh guards itself against malicious users. Must Provide Self Management An “organic network” should be both self-organizing and self managing? To what extent can we remove the human out of the loop? . Must Develop a Resilient Framework for Applications In a environmentally hostile environment, we must provide a framework for applications to work robustly. © 2007 Victor Bahl. All Rights Reserved. Handling the Challenges © 2007 Victor Bahl. All Rights Reserved. Strategies for increasing Capacity Strategy 1: Use all available channels – Avoid spectrum waste Strategy 2: Improve modulation, reception, and coding – Today ~ 2.5 bits/Hz (.11g), Soon ~ 4.5 bits / Hz (.11n) – Network coding Strategy 3: Improve spatial reuse by reducing interference – Fine grain transmit power control – (Steerable) directional antennas and directional MACs Strategy 4: Navigate around harmful interference – Interference aware least cost routing © 2007 Victor Bahl. All Rights Reserved. Will not cover Strategy 1: Multi-Channel Communications Goal Assign n non-interfering channels to n pair of nodes such that n packet transmissions can occur simultaneously. Knobs Single Channel Single Radio Multiple Radio Multiple Channels Today ☺ X ☺ © 2007 Victor Bahl. All Rights Reserved. Single Radio – Multiple Channels (SR-MC) Distributed: Use a modified RTS/CTS sequence to negotiate channels – Problem How does the sender know which channel the receiver is listening on? – Solutions Receive on all channels simultaneously – Simplest solution but too costly - will not consider here Use a dedicated rendezvous channel Use a synchronized hopping protocol Provide multiple rendezvous opportunities Centralized: Compute channel assignments using global knowledge Scope of Coverage We will cover schemes that work on commodity radios only © 2007 Victor Bahl. All Rights Reserved. Packets-in-Flight Example Revisited Negotiating Channel with RTS / CTS RTS (C1,C3,C7) C2 1 RTS (C3,C5,C7,C11) C1 2 3 C2 4 5 6 C11 7 CTS (C1, C7) 8 C1 9 CTS (C11) 10 nodes are active, 5 packets in flight, 150% improvement! © 2007 Victor Bahl. All Rights Reserved. 10 11 Note: Hidden Terminal – Multi-Channel Case Let C1 be the rendezvous channel γ can hear traffic on C1 only, doesn’t hear the CTS from β consequently doesn’t know anything about traffic on C6 (δ is too far to hear anything from β) α γ β C1 RTS C1 C6 C6 CTS Data C6 C6 on Data (C 6) on C 6 C6 Data on C6 Time (C 6) CTS C1 C11 C1 RTS δ Collision Possible solution: Use multiple radios © 2007 Victor Bahl. All Rights Reserved. So-MobiHoc-2004 Implementation Option for SR-MC Buffer packets, switch between channels Chandra-INFOCOM-2004 Channel switching speed: Today - 5 milliseconds Possible - 80 microseconds Application Layer User-level Kernel-level TCP/IP, Network Stack 802.11 Device Driver Switching logic Packets for C1 Packets for C11 Firmware 802.11 hardware Packets for C6 Multi-Channel Medium Access Control (MMAC) Idea: Periodically rendezvous on a fixed channel to decide the next channel • Divide time into beacon intervals • Divide a beacon interval into two phase So-MobiHoc-2004 – Negotiation Phase: All nodes switch to a pre-defined common channel and negotiate the channel to use – Transfer Phase: Once a channel is selected, the source & receiver switch to this channel and data transfer occurs during this phase Issues • Requires tight clock synchronization • Packets to multiple destinations can incur high delays • Congestion on the common channel • Common channel goes bad, everything goes bad • Not able to handle broadcasts © 2007 Victor Bahl. All Rights Reserved. Slotted Seeded Channel Hopping (SSCH) • • Divide time into slots At each slot hop to a different channel – • • Bahl-MobiCom-2004 Nodes hop across channels to distribute traffic Senders and receivers probabilistically meet & exchange schedules Senders loosely synchronize hopping schedule to receivers Characteristics Distributed: every node makes independent choices Optimistic: exploits common case that nodes know each others’ channel hopping schedules Traffic-driven: nodes repeatedly overlap when they have packets to exchange © 2007 Victor Bahl. All Rights Reserved. SSCH Rendezvous Divide time into slots: switch channels at beginning of a slot New Channel = (Old Channel + seed) mod (Number of Channels) seed is from 1 to (Number of Channels - 1) (1 + 2) mod 3 = 0 Seed = 2 Seed = 1 1 0 2 1 0 2 1 0 0 1 2 0 1 2 0 1 3 channels E.g. for 802.11b Ch 1 maps to 0 Ch 6 maps to 1 Ch 11 maps to 2 (0 + 1) mod 3 = 1 • Enables bandwidth utilization across all channels • Does not need control channel rendezvous © 2007 Victor Bahl. All Rights Reserved. SSCH Syncing Seeds • Each node broadcasts (channel, seed) once every slot • If B has to send packets to A, it adjusts its (channel, seed) Seed 2 2 2 2 2 2 2 2 2 2 1 0 2 1 0 2 1 0 3 channels B wants to start a flow with A Seed 1 2 0 2 1 0 2 1 0 1 1 2 2 2 2 2 2 2 Follow A: Change next (channel, seed) to (2, 2) Stale (channel, seed) info simply results in delayed syncing © 2007 Victor Bahl. All Rights Reserved. Using all Available Channels with SSCH In current IEEE 802.11 meshes 1 3 2 5 4 6 Only one of 3 pairs is active @ any given time With proper use of channels Ch 1 1 2 1 4 5 4 Ch 6 3 4 5 2 1 6 Ch 11 5 6 3 6 3 2 10 msecs 10 msecs 10 msecs © 2007 Victor Bahl. All Rights Reserved. … SSCH Performance 100 nodes, IEEE 802.11a, 13 channels, every flow is multihop Total System Throughput Avg. per node Throughput 4 80 Throughput (Mbps) 60 SSCH 50 40 30 SSCH 3 2.5 2 SSCH 1.5 IEEE 802.11 20 IEEE 802.11a 10 Throughput (Mbps) 3.5 70 1 SSCH 0.5 0 10 20 30 Number of Flows 40 50 M802.11 IEEE 802.11a 0 10 20 30 40 50 Number of Flows Significant capacity improvement when traffic load is on multiple separate flows © 2007 Victor Bahl. All Rights Reserved. How many Channels can we really use? Banerjee-SIGMETRICS-2006 2.4 GHz ISM band use in 802.11b Ch 1 Ch 6 Ch 11 • IEEE 802.11{b,g} partitions the allocated 83.5 MHz spectrum into 11 channels – Only channels 1, 6 and 11 are mutually non-overlapping • But…using only these channels may waste spectrum © 2007 Victor Bahl. All Rights Reserved. How many Channels can we really use? Ch 1 Ch 4 Ch 6 • Can we use more channels by using partially-overlapped channels? • Caution: may increase interference and cause more harm than good • Need an appropriate model to capture interference-effects and make correct choices Overlapped Channels do work! Link A, Channel 1 Distance (X-axis) Link B, Channel Y UDP Throughput (Mbps) 6 ChSep = 5 ChSep = 2 ChSep = 1 5 4 ChSep = 0 3 0 10 20 30 40 50 Distance (meters) LEGEND • Minimum distance between links for different choices of channel Non-overlapping channels, A = 1, B = 6 5 separation Partially Overlapped Channels, A = 1, B = 3 2 Partially Overlapped Channels, A = 1, B = 2 1 • For every channel separation there is a minimal distance 0 Same channel, A = 1, B = 1 • Model-based algorithmic approach for channel assignment in wireless Channel Separation mesh networks © 2007 Victor Bahl. All Rights Reserved. 60 Notes on Single Radio – Multiple Channels • Single radio solutions can be applied to multi-radios nodes since in most cases the number of channels is greater than the number of radios in the node. • Compared to multi-radio solutions, single radio solutions are power efficient – but power is not the primary concern in most mesh networks • Single radio solutions are less costly than multi-radio solutions but radios are fairly inexpensive • Switching speeds and mute-deaf-time is a problem in single radio solutions but switching speeds are being reduced dramatically • When distance between nodes is large, need not restrict operation to nonoverlapping channels only …so now lets look at multi-radio solutions © 2007 Victor Bahl. All Rights Reserved. Single Node Multiple Radio - Interference Study Question: Do two radios operating on non-overlapping channel interfere? Experimental setup: HOP 1 A TCP B 6” separation between B & C radios C TCP HOP 2 © 2007 Victor Bahl. All Rights Reserved. D 802.11a/g Interference Results 40.00 35 Hop 2 Netgear: A to B Hop TCP Throughput (Mb/Sec) 30 Throughput (Mb/sec) Netgear: C to D Hop 25 20 15 10 5 35.00 Hop 1 30.00 25.00 20.00 15.00 10.00 5.00 0.00 0 64,64 60,64 Channels 56,64 Hop1 = A, Hop2 = G 52,64 Same channel or channel separation of 4 causes 46% - 49% reduction in overall throughput Hop1 = G, Hop2 = A 802.11a link causes a 22% reduction in overall throughput, and a 63% reduction in throughput on the 802.11g link. Surprise: 802.11g does not affect 802.11a Implications • Interference even when radios are placed 6” apart is significant • Significant RF hardware shielding work is needed © 2007 Victor Bahl. All Rights Reserved. Single Node Multiple Radios Channel 1 Channel 11 t1 Source Mesh Box Destination Let’s assume we can build “mesh-boxes” with enough separation / shielding between radios that performance does not suffer. Then interesting problem to consider (1) How should we assign channels to each interface? Don’t want to cause network partitions (2) Which interface should we send the packet on? State-of-art metrics (hop count, ETX, SRTT, packet-pair) are not suitable for multiple radio / node. As they do not leverage channel, range, data rate diversity. © 2007 Victor Bahl. All Rights Reserved. Multiple Radios - Multiple Channels Options to consider Static Assignment One channel / radio for all time Suboptimal use of spectrum Some routes may be suboptimal Dynamic Assignment (all SR-MC strategies apply) Channels assigned to match traffic patterns and/or to reduce interference Interference patterns can change, network may get disconnected Hybrid Assignment One channel to one radio for all time, for all other radios, channels are assigned dynamically to match traffic patterns and/or reduce interference © 2007 Victor Bahl. All Rights Reserved. Static Assignment (1) 2 radios / node Draves-MobiCom-2004 All nodes use common set of channels A 11 1 B 11 1 11 1 11 1 D 11 C 1 E 11 F © 2007 Victor Bahl. All Rights Reserved. Suboptimal use of spectrum Static Assignment (2) 2 radios / node, 4 channels Raniwala-Infocom-2005 Different nodes use different channels A 52 56 52 B 52 60 C 60 Some routes may be suboptimal (e.g. B->F) D 60 E 64 F © 2007 Victor Bahl. All Rights Reserved. Dynamic Assignment N radios / node; M channels; N < M Interfaces can switch channels as needed A B C Coordination may be needed before each transmission D • MMAC • SSCH • BFS-CA E F See section on single radio – multiple channels © 2007 Victor Bahl. All Rights Reserved. Dynamic Assignment Breadth First Search – Channel Assignment (BFS-CA) Ramachandran-Infocom-2006 Goals: – External interference can severely degrade mesh performance Measure & avoid external interference – Internal interference between mesh links should be avoided Assign orthogonal channels to any two interfering mesh links Nodes periodically estimate surrounding interference levels – A radio per-node monitors 802.11 data and control traffic – Channels ranked from least interfered to most interfered – Ranking sent to centralized channel assignment server Distributed channel sensing and channel assignment can break network connectivity – A radio per-router is tuned to a common channel to ensure connected mesh © 2007 Victor Bahl. All Rights Reserved. BFS-CA Interference-Aware Channel Assignment • Multi-radio conflict graph models interference between mesh links • Breadth first search algorithm selects channels for mesh radios • Significant performance improvement over static channel assignment in the presence of varying interference levels Issues – Interference patterns can change rapidly – Dependence on existance of traffic patterns to determine interferance Incorrect channel assignment possible © 2007 Victor Bahl. All Rights Reserved. Hybrid Assignment Hybrid Multichannel Control Protocol (HMCP) • Each node has two interfaces (1 fixed, 1 switch-able) – Connectivity is maintained + all channels used Kyasanur-WCM-2006 • Every node picks a channel as it’s fixed channel – Different nodes use different fixed channels • Sender tunes it’s “switchable” interface to receiver’s fixed channel to send packets – Once a “connection” is made, there may not be a reason to switch channels again for that particular flow. © 2007 Victor Bahl. All Rights Reserved. HMCP Channel Selection Challenge: Nodes in a neighborhood should use different fixed channels Fixed Channel Selection – On startup pick a random fixed channel – Periodically send a “hello” pkt. containing fixed channel & 1-hop neighbors info. on all channels (using the switchable interface) – Maintain a NeighborTable containing fixed channels being used by neighbors – Select the channel with fewest nodes as a candidate Use 2-hop neighbor information – Change fixed channel to candidate channel probabilistically to avoid oscillations Issues – High overhead for broadcast packets © 2007 Victor Bahl. All Rights Reserved. Which Interface should we send the packet on? Note: Must use off-the-shelf wireless NICs A Simple Approach: Multi-Radio Unification Protocol (MUP) – For every transmission select the interface with the “best” channel & transmit on it Pros: - Locally optimizes use of available spectrum - Does not require changes to routing protocols or application-level software - Interoperates with legacy hardware - Does not require global topology information Cons: – For one-hop ad hoc – works great. For meshes need metrics that combine link selection metrics into a path selection metric (will see) © 2007 Victor Bahl. All Rights Reserved. MultiRadio Unification Protocol (MUP) Illustration of Channel Switching Goal Allow nodes with multiple radios to locally optimize use of available spectrum and hence increase capacity Ch. 0 0 MUP Enabled 1 Operation • Set the network interface cards on different frequency channels Ch. 1 • Periodically monitor channel / Link quality on each interface • Select the interface with the “best” channel and transmit packets Does not require global topology information Adya-BroadNets-2004 © 2007 Victor Bahl. All Rights Reserved. MUP in a Neighborhood 252 houses in a Seattle neighborhood (Green Lake Area) Mesh formation among 35 randomly selected houses Web surfer 40-50% reduction in delay compared to a one-radio network ITAP Routes via RFC 3561 (AODV) © 2007 Victor Bahl. All Rights Reserved. Why MUP is not enough? MUP is a link metric not a path metric. Routing protocols that use MUP do not – Leverage channel diversity A two hop path with hops on different channels is better than a path with both hops are on the same channel. – Leverage range and data rate diversity A path with two 6 Mbps hops is better than a path with a single 1 Mbps hop. MUP will take the 1Mbps path. …..but a path with four 6 Mbps hops is worse than a path with a single 2 Mbps hop. MUP and metrics like ETX may take the four-hop path, depending on delay & loss rate. Note: Striping protocols are not enough – – – Packet-level striping results is packet reordering, and hence poor TCP throughput Flow-level striping requires a routing algorithm! MUP may work, but only if radios have identical range. Bottom Line: Need a routing protocol / metric that takes bandwidth, loss rate, and channel diversity into account. © 2007 Victor Bahl. All Rights Reserved. Strategy 2: Higher Capacity by Improving Coding = 4 transmissions in the air A C A R Throughput (Mbps) UDP (A→C, C→A) 11.02 TCP (A→C, C→A) 10.91 TCP (C →A) 10.55 © 2007 Victor Bahl. All Rights Reserved. C Network Coding Improves Capacity C A = 3 transmissions in the air broadcast XOR C++ A Throughput (in Mbps) A R w/o Network Coding Network Coding UDP (A→C, C→A) 11.02 18.13 (+64%) TCP (A→C, C→A) 10.91 13.97 (+28%) TCP (C →A) 10.55 12.11 (+15%) © 2007 Victor Bahl. All Rights Reserved. C Calculating Mesh Capacity • Gupta & Kumar focused on determining asymptotic, pessimistic bounds on performance. – O(1/sqrt(N)) • Want to answer the questions about achievable capacity for specific topologies with specific traffic pattern and node specific configurations Example: 4 houses talk to the central ITAP. What is the maximum possible throughput? Asymptotic analysis is not useful in this case © 2007 Victor Bahl. All Rights Reserved. Calculating Mesh Capacity (2) Analytical Approach Jain-MobiCom-2003 Create a connectivity graph – Each vertex represents a wireless node – Draw a directed edge from vertex A to vertex B if B is within range of A – We can incorporate more sophisticated connectivity models – Capacity of each edge is determined by wireless technology in use Create a conflict graph that shows which wireless links interfere with each other – Each edge in the connectivity graph represented by a vertex – Draw an edge between two vertices if the links conflict with each other © 2007 Victor Bahl. All Rights Reserved. How does it work? • Create a linear program that solves the basic MAXFLOW problem on this connectivity graph • Cliques in conflict graph – At most one of the links can be operating at any given instant – Sum of utilization of links belonging to a clique is <= 1 – MAXFLOW LP can be augmented with Clique constraints to get a better upper bound • Independent sets in conflict graph – All links belonging to an independent set can be active at the same time. – MAXFLOW LP can be augmented with constraints derived from independent sets to get a lower bound – Lower bound is always feasible. • If upper and lower bounds converge, we have optimal answer © 2007 Victor Bahl. All Rights Reserved. Sample Result Mesh Example: 4 houses talk to the central ITAP. What is the maximum possible throughput? 1 0.9 Upper Bound 0.8 Lower Bound Throughput 0.7 0.6 0.5 0.4 Optimal Throughput 0.3 0.2 0.1 0 0 50 100 150 200 Effort Houses talk to immediate neighbors, all links are capacity 1, 802.11-like MAC, Multipath routing. Better results with more computation time © 2007 Victor Bahl. All Rights Reserved. The “What if” Analysis What if we double the range? 1 0.9 Upper Bound 0.8 Lower Bound 0.9 Upper Bound 0.8 Lower Bound 0.7 0.7 0.6 0.6 Throughput Throughput 1 0.5 0.4 0.5 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 20 40 60 80 100 120 140 160 0 20 40 Effort Original scenario 60 80 100 Effort Doubling the range provides no benefit in this scenario! © 2007 Victor Bahl. All Rights Reserved. 120 Conflict Graph Analytical Framework Advantages – Scenario based numbers instead of asymptotic bounds – Framework is flexible. It can accommodate: Multiple radios on multiple channels Directional antennas Single path or Multi-path routing Different ranges, data rates Different wireless interference models Different topologies – We can also include fairness criteria, pricing. – Great for “what if” analysis Limitations – Requires time to generate accurate results. Not suitable for making on-the-fly routing decisions. – Problem of finding optimal throughput is NP-Hard. Best we can do is to find upper and lower bounds and bounds may not always converge – Requires global knowledge of network topology © 2007 Victor Bahl. All Rights Reserved. Problem: How to Determine the Conflict Graph We know that – Interference degrades performance – Conflict graph captures link interference – Conflict graph is necessary for capacity estimation, better routing, channel assignment & power management Question: How should we determine the conflict graph? – How should we study the structure and the stability of the interference graph? – How can we understand the relationship between interference graph and the communication topology of the network? © 2007 Victor Bahl. All Rights Reserved. An attempt at Determining Interference • Establish a base maximum sending rate for each node • Select two or more nodes to simultaneously broadcast packets as fast as possible • Study the sending rate and the reception rate at every node – Calculate Broadcast Interference Ratio (BIR) Approximates Link Interference Ratio (LIR) – O(n2) experiments suffice to estimate O(n4) LIR values • Vary the transmit power, PHY data rate etc. – Repeat the experiment • BIR needs to be calculated regularly as interference patterns change over time. © 2007 Victor Bahl. All Rights Reserved. Determining Interference Pairwise Interference Measure A’s receive rate @ B = M Measure A’s receive rate @ B = M/ Calculate BIR: (M/ + N/) / (M + N) =1 no interference A B < 1 interference = 0 severe interference C D Measure C’s receive rate @ D = N Measure C’s receive rate @ D = N/ © 2007 Victor Bahl. All Rights Reserved. Padhye-IMC-2005 Interference – IEEE 802.11a (Indoors) 620 6000 600 5000 580 4000 560 3000 540 2000 520 1000 500 0 480 460 12 Link Quality 440 420 10 400 450 500 550 600 650 Carrier Sense Interference 700 750 8 Observation: The difference between the interference and the communication range is small for 802.11a 6 4 We may be able to get a reasonable approximation for the interference graph from the communication graph using simple heuristics 2 0 0 10 20 30 40 © 2007 Victor Bahl. All Rights Reserved. 50 60 70 80 90 Interference – IEEE 802.11b (Indoors) 650 1000 900 800 700 600 500 400 300 200 100 0 600 550 500 450 2 Link Quality Carrier Sense Interference 1.8 400 1.6 200 300 400 500 600 700 800 1.4 1.2 Observation 1 The relationship between the interference and the communication range is difficult to characterize in 802.11b. 0.8 0.6 0.4 0.2 0 0 20 40 © 2007 Victor Bahl. All Rights Reserved. 60 80 100 120 140 …about routing in mesh networks The Routing Problem Why not simply use traditional routing protocols (RIP/OSPF/etc)? • Network topologies are dynamic due to router mobility & environmental fluctuations – Dynamic topology may prevent routing protocol convergence • Many links are redundant (routing updates can be large) • Periodic updates may waste bandwidth & batteries • Computed routes may not work due to unidirectional links • Wireless makes routing protocols easy to attack • Link quality, spectrum utilization and interferences are uniquely important for path selection © 2007 Victor Bahl. All Rights Reserved. Desirable Qualitative Properties • • • • • • • Distributed operation Loop-freedom Corson-RFC2501-1999 Demand-based operation Proactive operation Attack resistant & Secure “Sleep” period operation (friendly to power management) Unidirectional link support / asymmetric link support Implementation – Layer 3 - traditional “network layer” / IP layer Interoperable internetworking capability and consistency over a heterogeneous networking infrastructure. – Layer 2.5 Agnostic of IPv4 or IPv6 issues and can incorporate link quality measures more easily Capable of handling multiple wireless & wired networking technologies © 2007 Victor Bahl. All Rights Reserved. Routing and Addressing Many choices, which is the best one? – Flat addressing - Each node runs the routing protocol; node’s address is independent of its location (e.g. PRNET, TORA, DSR, AODV,..) – Clustering – Only cluster heads run routing protocol; addressing is flat and independent of node’s location (NTDR, CEDAR,..) – Hierarchical – Only cluster heads run routing protocol, a node’s form subnets, each node acquire’s address of its subnet (SURAN). Many protocols to consider – – – – Proactive based on traditional distance vector (DSDV, OLSR, TBRPF) Reactive or on-demand (DSR, AODV) Hybrid: proactive + reactive (ZRP) Geographical (LAR) Multiple path routing – Many choices: MSDR, AOMDV, AODV-BR, APR, SMR, ROAM, …. © 2007 Victor Bahl. All Rights Reserved. Proposals: Unicast Ad Hoc Multi-hop Routing Protocols • • • • • • • • • • • • • • • • • • • • • • • • • • • • • ABR (Associativity-Based Routing Protocol) AODV (Ad Hoc On Demand Distance Vector) ARA (Ant-based Routing Algorithm) BSR (Backup Source Routing) CBRP (Cluster Based Routing Protocol) CEDAR (Core Extraction Distributed Ad hoc Routing) CHAMP (CacHing And MultiPath routing Protocol) CSGR (Cluster Gateway Switch Routing) DART (Dynamic Address Routing) DBF (Distributed Bellman-Ford) DDR (Distributed Dynamic Routing) DNVR (Dynamic Nix-Vector Routing) DSDV (Dynamic Destination-Seq. Dist. Vector) DSR (Dynamic Source Routing) DSRFLOW (Flow State in the DSR) DYMO (Dynamic Manet On-Demand) FORP (Flow Oriented Routing Protocol) FSR (Fisheye State Routing) GB (Gafni-Bertsekas) GLS(Grid) (Geographic Location Service) GPSAL (GPS Ant-Like) GSR (Global State Routing) Guesswork HARP (Hybrid Ad hoc Routing Protocol) HSLS (Hazy Sighted Link State) HSR (Hierarchical State Routing) HSR (Host Specific Routing) IARP (Intrazone Routing Protocol) IERP (Interzone Routing Protocol) • • • • • • • • • • • • • • • • • • • • • • • • LANMAR (LANdMARk Routing Protocol) LAR (Location-Aided Routing) LBR (Link life Based Routing) LCA (Linked Cluster Architecture) LMR (Lightweight Mobile Routing) LQSR (Link Quality Source Routing) LUNAR (Lightweight Underlay Network Ad hoc Routing) MMRP (Mobile Mesh Routing Protocol) MOR (Multipoint On-demand Routing) MPRDV (Multi Point Relay Distance Vector) OLSR (Optimized Link State Routing) OORP (OrderOne Routing Protocol) DREAM (Distance Routing Effect Algorithm for Mobility) PLBR (Preferred Link Based Routing) RDMAR (Relative-Distance Micro-discover Ad hoc Routing) Scar (DSR and ETX based) SSR (Signal Stability Routing) STAR (Source Tree Adaptive Routing) TBRPF (Topology dissemination Based on ReversePath Forwarding) TORA (Temporally-Ordered Routing Algorithm) WRP (Wireless Routing Protocol) ZHLS (Zone-Based Hierarchical Link State) ZRP (Zone Routing Protocol) …. © 2007 Victor Bahl. All Rights Reserved. Popular Taxonomy Multihop Routing Protocols Proactive Reactive Hybrid Hierarchical © 2007 Victor Bahl. All Rights Reserved. Geographical Power Aware Mapping Protocols to Taxonomy Multihop Routing Protocols Proactive Reactive CSGR DBF DSDV Guesswork HSLS HSR IARP LCA MMRP OLSR STAR TBRPF WRP Hybrid ARA ABR AODV BSR CHAMP DSR DSRFLOW DNVR DYMO FORP GB IERP LBR LMR LQSR LUNAR MOR MPRDV RDMAR SrcRR SSR TORA PLBR ZRP Hierarchical Geographical CBRP CEDAR DART DDR FSR GSR HARP HSR HSR LANMAR OORP Power Aware DREAM GLS(Grid) LAR GPRS GPSL ZHLS © 2007 Victor Bahl. All Rights Reserved. ISAIAH PARO EADSR PAMAS Common Metrics for Comparing Routing Protocols • Route Acquisition Time – Time required to establish route(s) Broch-MobiCom-1998 • Routing Overhead – Total number of routing packets transmitted (for discovery & maintenance) for a fixed amount of transfer over multiple hops with random node mobility. • Path Optimality / End-to-End Throughput – TCP & UDP data throughput and delay Previously path optimality was measured in terms of the difference between the number of hops a packet took to reach it’s destination and the length of the shortest path that physically existed through the network when the packet originated. However, the hop count metric has been overshadowed by other link metrics when it came to end-to-end throughput • Packet Delivery Ratio – Ratio between the number of packets originated by the application layer and the number of packets received by the sink at the final destination © 2007 Victor Bahl. All Rights Reserved. Context for Comparing Metrics • Network size Corson-RFC2501-1999 Measured in terms of the number of nodes • Network connectivity Measured in terms of avg. node degree (i.e. the avg. number of neighbors) • Topological rate of change Speed with which a network's topology changes • Link capacity Effective link speed in bps, after accounting for losses due to multiple access, coding, framing, etc. • Fraction of unidirectional links Transmission ranges of radios may be different • Traffic patterns Long-lived versus bursty non-uniform • Mobility Described in terms of dwell time, movement direction, speed etc. • Fraction and frequency of sleeping nodes © 2007 Victor Bahl. All Rights Reserved. Most Studied Proactive (periodic) – Each node maintains route to each other network node (Global state) – – – – – Routes are determined independent of traffic All topology changes propagated to all nodes Periodic routing advertisements (neighbor discovery is beacon based) Generally longer route convergence time Examples: Distance vector and link state (DSDV, OLSR, TBRPF) Reactive (on-demand) – Actions driven by data packet requiring delivery – Source builds route only when needed by “flooding” (Route Discovery) – Maintain only active routes (Route Maintenance) – Pro: Typically less overhead, better scaling properties – Cons: Route acquisition latency – Examples: DSR, AODV © 2007 Victor Bahl. All Rights Reserved. Conventional Wisdom • Proactive protocols perform best in networks with low to moderate mobility, few nodes and many data sessions – E.g. OLSR (RFC 3626), TBRPF (RFC 3684) • Reactive protocols perform best in resource-limited, dynamic networks where nodes are mobile. Tradeoff routing overhead for start-up delay – E.g. AODV (RFC 3561), DSR (IETF Draft) © 2007 Victor Bahl. All Rights Reserved. Courtesy: David B. Johnson RFC 4728: Dynamic Source Routing (DSR) Reactive / On demand – When node S wants to send data to node D, but does not have a route to it IETF Experimental RFC 4728 Properties: – – – – – Johnson-Draft-2004 All aspects operate entirely on-demand Ignores all topology changes not affecting you Overhead scales automatically with movement No overhead when stationary and found routes Supports unidirectional links & asymmetric routes Johnson-RFC4728-2007 Divides routing problem into two parts: – Route Discovery: Find route only when don’t have one and need one – Route Maintenance: Fix a broken route only while you’re actually using it © 2007 Victor Bahl. All Rights Reserved. DSR Basic Route Discovery Overview To discover a route to some address: – Broadcast a ROUTE REQUEST with a unique id – When receiving a ROUTE REQUEST: If destination IP address is yourself Return the recorded route to the initiator in a ROUTE REPLY Else if recently seen this id Drop the REQUEST Else Append your own address and rebroadcast the ROUTE REQUEST Notes: – Nodes overhearing source routes learn topology – Nodes can answer ROUTE REQUEST using cached routes Speeds up discovery and reduces propagation of ROUTE REQUEST – Can store multiple routes to same node © 2007 Victor Bahl. All Rights Reserved. DSR Basic Route Maintenance Overview After transmitting a packet to the next hop: – Listen for link-level per-hop acknowledgement, or – Listen for that node sending packet to next hop, or – Set a bit in the packet to request explicit next-hop acknowledgement When a problem with forwarding is detected: – Send a ROUTE ERROR to original sender – Sender removes the broken link from its cache © 2007 Victor Bahl. All Rights Reserved. DSR Issues • Scalability: Packet header size grows with route length – Can be an issue specially when data packets are small (e.g. VoIP) • Route Caching: Each node caches a route it learns – Pros: Can speed up route discovery & reduce route request propagation – Cons: Stale routes can adversely affect performance TCP applications suffer the most • ROUTE REQUEST flooding Ni-MobiCom-1999 – Broadcast storm problem Collisions between (redundant) route requests by neighbors – Solutions: (1) re-broadcast Route Request with probability p (2) re-broadcast by using classical collision avoidance techniques, e.g. introducing random delays (3) … © 2007 Victor Bahl. All Rights Reserved. RFC 3561 Adhoc On-Demand Distance Vector • Table driven Reactive / On demand Perkins-RFC3561-2003 • IETF Experimental RFC 3561 • Like DSR, but maintains routing tables at the nodes – data packets do not have to carry routes • Sequence numbers are used for route freshness and loop prevention • Protocol divided in Route Discovery and Route Maintenance – Maintains active routes only • Popular routing protocol, many public implementations available – Check out: http://moment.cs.ucsb.edu/AODV/ © 2007 Victor Bahl. All Rights Reserved. AODV Basic Route Discovery Overview To discover a route to some address D: – Broadcast a ROUTE REQUEST (RREQ) – On receiving a ROUTE REQUEST If destination IP address is yourself – Create a ROUTE REPLY (RREP) – Unicast RREP to node from which RREQ was received Else, If have a route to D, and the seq # for route to D is >= D’s seq # in RREQ – Create a RREP – Unicast RREP to node from which RREQ was received Else – Make a reverse entry for Source IP address – Rebroadcast RREQ © 2007 Victor Bahl. All Rights Reserved. AODV Basic Route Discovery Overview (2) To discover a route to some address D (cont.): – On receiving a ROUTE REPLY If Source IP address is yourself – Make forward route entry to D <Destination IP address, next hop IP Address, HopCount to Destination> Else – Make a forward route entry to D – Unicast RREP to node from which RREQ was received © 2007 Victor Bahl. All Rights Reserved. AODV Basic Route Maintenance Overview When a intermediate node detects a link problem: – It invalidate the route to destination – Creates a ROUTE ERROR (RERR) message Lists all destinations that are unreachable Sends to upstream neighbors On Receiving ROUTE ERROR – If Source IP is yourself if node sending RERR is it’s next hop to destinations – Delete route(s) to destination(s) – Restart Route Discovery if needed – Else If node sending RERR is it’s next hop to destination – Deletes route(s) to destination – Forwards RERR to source address © 2007 Victor Bahl. All Rights Reserved. AODV Optimizations Expanding Ring Search – Prevents flooding of network during route discovery – Control Time to Live (TTL) of RREQ to search incrementally larger areas of network Less overhead when successful Longer delay if route not found immediately Local Repair – Repair breaks in active routes locally instead of notifying source – Use small TTL because destination probably hasn’t moved far – If first repair attempt is unsuccessful, send RERR to source Repairs links with less overhead, delay and packet loss Longer delay and greater packet loss when unsuccessful © 2007 Victor Bahl. All Rights Reserved. AODV Issues • AODV assumes symmetric (bi-directional) links – The intermediate nodes set up reverse paths towards source • AODV does not support multiple routes between source / destination pair • Routes can expire even when topology does not change © 2007 Victor Bahl. All Rights Reserved. Link / Path Selection Metrics Min. hop count results in lower-quality links © 2007 Victor Bahl. All Rights Reserved. Path Selection Metrics • Link Metric: Assign a weight to each link – Prefer high bandwidth, low-loss links – RTT, Packet Pair, ETX • Path Metric: Combine metrics of links on path – Prefer short, channel-diverse paths – WCETT © 2007 Victor Bahl. All Rights Reserved. Expected Transmission Count (ETX) Couto-MobiCom-2003 • Each node periodically broadcasts a probe • The probe carries information about probes received from neighbors • Each node can calculate loss rate on forward (Pf) and reverse (Pr) link to each neighbor • Selects the path with least total ETX ETX 1 (1 Pf) * (1 Pr) Advantages – Explicitly takes loss rate into account – Implicitly takes interference between successive hops into account – Low overhead Disadvantages – PHY-layer loss rate of broadcast probe packets is not the same as PHY-layer loss rate of data packets Broadcast probe packets are smaller Broadcast packets are sent at lower data rate – Does not take data rate or link load into account © 2007 Victor Bahl. All Rights Reserved. Expected Transmission Time (ETT) Link loss rate = p – Expected number of transmissions (ETX) = 1/(1-p) Mean Packet size = S, Link bandwidth = B – Each transmission lasts for S/B Lower ETT implies better link S ETT * ETX B © 2007 Victor Bahl. All Rights Reserved. Expected Transmission Time (ETT) Taken the MAC delays into account – Min backoff window CWmin ETT ETxmit ETbackoff where, ETxmit S B(1 p) i 7 f(p) 1 2 i 0 (i 1) ETbackoff p CWmin f(p) 2(1 p) i © 2007 Victor Bahl. All Rights Reserved. Path Selection Metric in Multi-Radio Mesh • Link metrics such as shortest path, Packet Pair, RTT, ETX & ETT do not leverage channel, range and data rate diversity • In multiradio meshes, need a path metric that does. • This metric should • avoid unnecessarily long paths – bad for TCP performance – bad for global resources • know that all hops in the path on the same channel interfere • Path throughput is dominated by the maximum of these sums © 2007 Victor Bahl. All Rights Reserved. WCETT = Combining link ETTs Given a n hop path, where each hop can be on any one of k channels, and two tuning parameters, a and b: Path throughput is dominated by the max of the sum of ETTs of path links on the same channel Sum of ETTs of all links on the path - Favors short paths a * ETT b * max WCETT n i 1 i 1 j k X j ab where X j Draves-MobiCom-2004 ETT hop i is on channel j i Sum of ETTs of all links on the path that are on the same channel Select the path with min WCETT © 2007 Victor Bahl. All Rights Reserved. Incorporating Link Quality Metrics into Routing Protocols… Extremely Opportunistic Routing (ExOR) Probabilistic Broadcast Biswas-SIGCOMM-2005 – Decides who forwards after reception Desirable: node closest to destination should forward Challenge: avoid duplicate transmissions How to determine who is closest – Nodes periodically flood ETX “link state” measurements Path ETX is weighted shortest path (Dijkstra’s algorithm) – Closest node is determined by sorting via ETX metric To avoid duplicate transmissions – Packets includes a priority list of forwarder in ExOR header – One node sends at a time, ordered from higher priority to lower priority Layer 2.5 design © 2007 Victor Bahl. All Rights Reserved. Link Quality Source Routing (LQSR) • Derived from Dynamic Source Routing Protocol Draves-SIGCOMM-2004 • Optimizations for sporadic node mobility at the virtual link layer • Incorporates 5 different link selection metrics: – Hop count, RTT, Packet Pair, ETX, WCETT • A layer 2.5 protocol – For the protocols above looks a like virtual adapter on virtual link – For the drivers below binds to physical adapters that provide next-hop connectivity – Inserts a new L2.5 header • Why Layer 2.5? – Works over heterogeneous links (e.g. wireless, powerline) – Transparent to higher layer protocols. • works equally well with IPv4 and IPv6 – ARP etc. continue to work without any changes © 2007 Victor Bahl. All Rights Reserved. Hybrid Wireless Mesh Protocol (HWMP) • Proposal for default routing protocol in IEEE 802.11s meshes • On-demand with option for pro-active routing IEEE-Mesh-2006 • Support multiple path selection metrics – vendor defined (by default, Airtime) • On-demand version – RM-AODV Essentially RFC 3561 RM is for Radio Metric (or link metric) • Pro-active version – Distance-vector routing tree from a root portal (e.g. an AP as root) Avoid unnecessary discovery flooding © 2007 Victor Bahl. All Rights Reserved. Path Length and Throughput Eriksson-MobiSys-2006 Which metric is best? (“Wireless Office Study”) WCETT Experimental Setup ETX HOP 3.5 • • 23 node testbed Randomly selected 100 senderreceiver pairs (out of 23x22 = 506) 3-minute TCP transfer (transmit as many bytes as possible in 2 minutes, followed by 1 minute of silence) For 1 or 2 hop the choice of metric doesn’t matter 2.5 2 1.5 1 0.5 0 A C D E F Testbed Configuration WCETT ETX HOP 4000 Throughput (Kbps) • Hop Length 3 3500 3000 2500 2000 1500 1000 500 0 A C D E Testbed Configuration © 2007 Victor Bahl. All Rights Reserved. F Comparison of Metrics Eriksson-MobiSys-2006 Wireless Office Scenario 23 node indoor testbed. Two radios (both 802.11a) per node. 11 active clients, 4 servers. Heavy Office Traffic 1 hour, 308 sessions, 587.5 MB total Light Office Traffic 1 hour, 415 sessions, 19.72 MB total 10000 1000 474 100 10 89 120 179 82 11 4 6 4 5 3 8 3 6 ETX HOP PKTPAIR 1000 RTT 590 862 943 31 30 3 3 ETX HOP 100 27 10 4 2 1 WCETT Additional Delay (ms) Additional Delay (ms) 10000 1 WCETT PKTPAIR Relatively light traffic means performance is okay for all metrics. WCETT does better under heavy load (worst case delay) © 2007 Victor Bahl. All Rights Reserved. RTT Summarizing • Many routing protocols to choose from • Protocols that take link quality into account show most promise • A link quality metric that incorporates interference is still needed • Adaptive protocols that change behavior in different environments might be best © 2007 Victor Bahl. All Rights Reserved. Securing & Managing Meshes Network management is a process of controlling a complex data network so as to maximize its efficiency and productivity Network Management a Practical Perspective Addison Wesley 1993 Security & Privacy Requirements • Only “authorized” nodes should be allowed to route through mesh – Packet level authentication • Authentic route setup (e.g. Ariadne, Sead) • Robust & adaptive forwarding in face of malicious or congested nodes or links • Detect malicious nodes who promised to route during the route setup, but actually don’t forward • Limit the resource consumption of greedy nodes – Rate control/policing • Preserve the anonymity of communicating parties © 2007 Victor Bahl. All Rights Reserved. MSR’s Mesh Security Architecture For Wireless Office and Community Networks Scenario Philosophy Internet Mesh is open to all, however, people who contribute resources to the mesh get priority. Internet Gateway Goals Certificate • Guard against faulty or hacked mesh nodes • Defend against disruption (e.g., denial-of-service attacks) by malicious end devices • Protect mesh traffic privacy • Laptop Certificate Snooper Server Neighborhood File Server Protect access to network resources Compromised Mesh Box Assumptions • End Device Difficult to hack into a “Mesh Box” (similar to cable modems) End Device Malicious End Device © 2007 Victor Bahl. All Rights Reserved. Regulations handle malicious RF Interference attacks Certificate Basic Security Framework Certification Encryption & Anonymity • • Traffic between mesh nodes is encrypted to foil eavesdroppers • Misbehaving mesh nodes have their certificates “blackballed” • Traffic from uncertified end devices is prevented from disrupting certified traffic • Mesh nodes have built-in publickey certificates for authenticating to each other Authenticated mesh nodes issue certificates to end devices • Mesh nodes accept certificates issued by any mesh node within range • Access to resources is controlled based on policy and end device certificate © 2007 Victor Bahl. All Rights Reserved. Troubleshooting Wireless Meshes The least well understood area of research Difficult because: Network is hostile Unpredictable physical medium, Dynamic topology, Overhead limitations, Link vulnerabilities, ….. Root cause fault identification is hard Determining normal behavior is difficult – Just knowing link statistics is insufficient – Typical signature based techniques don’t work – Handling multiple faults » Complicated interactions between faults and traffic » Among faults themselves © 2007 Victor Bahl. All Rights Reserved. Mesh Management Design Reactive and Pro-active – Investigate reported performance problems Time-series analysis to detect deviation from normal behavior – Localize and Isolate trouble spots Collect and analyze traffic reports from mesh nodes – Determine possible causes for the trouble spots Interference, or hardware problems, or network congestion, or malicious nodes …. Respond to troubled spots – – – – Re-route traffic Rate limit Change topology via power control & directional antenna control Flag environmental changes & problems © 2007 Victor Bahl. All Rights Reserved. Management Workflow Mesh Configuration & Setup (scope out network) Initialiazation Gather & Distribute Data Step 1 Clean & Analyze Data Determine Physical Topology Step 2 Model Network Behavior “What-if” Analysis Detect Anomaly Improve Routing/ Capacity Locate Hot Spots & Inform Step 3 Step 4 Diagnose Problem Suspect software/ hardware Poor local connectivity Inform/Fix Reconfigure Topology Suspect attack Congestion Rate Limit Perform Security Analysis © 2007 Victor Bahl. All Rights Reserved. Step 5 Some Approaches Data distribution protocols for network management ANMP Chen-JSAC-1999 Guerrilla Shen-Milcom-2002 Network monitoring to detect routing and MAC misbehavior Watchdog & Pathrater MACMis Kyasanur-TMC-2005 Filtering and graph dependency analysis Marti-MobiCom-2000 Fault detection, isolation & root cause analysis Badonnei-JNM-2005 Qiu-CCR-2006 On-line modeling based analysis © 2007 Victor Bahl. All Rights Reserved. Qiu-MSRTR-2003 Step 1: Monitor & Distribute Data Monitoring: What should we collect? – Link Info: Noise level, signal strength, loss rate to direct neighbor (packet retransmission count) – Connectivity Info: Network topology / connectivity Info (Neighbor Table) – Traffic Info: Load to direct neighbor – … Distribution: Minimize (overhead) bandwidth consumption – Dynamic scoping Each node takes a local view of the network The coverage of the local view adapts to traffic patterns – Adaptive monitoring Minimize measurement overhead in normal case Change update period Push and pull – Delta compression – Multicast © 2007 Victor Bahl. All Rights Reserved. Step 2: Clean & Analyze Data Data may not be pristine. Why? – Liars, malicious users – Missing data – Measurement errors Clean the Data – Detect Liars Assumption: most nodes are honest Approach: – Neighborhood Watch – Find the smallest number of lying nodes to explain inconsistency in traffic reports – Smoothing & Interpolation © 2007 Victor Bahl. All Rights Reserved. Step 2: Clean & Analyze: The Watchdog B S C S sends a packet to B; If B doesn’t forward packet to C; S assumes B is malicious Nodes may not forward packets Marti-MobiCom-2000 – Because they are malicious – Because they are selfish (e.g. to conserve energy, or get greater bandwidth) Watchdog – When a node drops more than a Threshold number of packets, Watchdog can inform the sender of the misbehaving node Limitations – Ambiguous collisions, partial dropping, collution, environmental fluctuations, false misbehaviour reporting, etc. © 2007 Victor Bahl. All Rights Reserved. Resiliency against Liars/Lossy Links Qiu-CCR-2006 Problem • Identify nodes that report incorrect information (liars) Detect lossy links Assume • • Nodes monitor neighboring traffic, build traffic reports and periodically share info. Most nodes provide reliable information 1 0.8 0.6 0.4 0.2 0 NL=1 NL=2 Challenge NL=8 Watchdogs Find the smallest number of lying nodes to explain inconsistency in traffic reports Use the consistent information to estimate link loss rates NL=10 NL=15 NL=20 false positive Detect lossy links Fraction of lossy links identified Approach • NL=5 coverage Wireless links are error prone and unstable • • Qiu-MSRTR-2003 Detect liars Fraction of lying nodes identified • Simulation Results 1 0.8 0.6 0.4 0.2 0 NL=1 NL=2 NL=5 NL=8 NL=10 NL=15 NL=20 coverage © 2007 Victor Bahl. All Rights Reserved. false positive Guarding against Attacks - Corrupted Routers • Monitor router behavior when throughput problems are detected • The original router queries each downstream router in turn, obtaining verifiable traffic throughput information Internet Internet Gateway Certificate Laptop Certificate Snooper • • A “bad” link, where traffic isn’t getting through, can be identified Server Neighborhood File Server Which end of the bad link is misbehaving can be investigated outof-band, or the link can simply be avoided End Device Compromised Mesh Box End Device Malicious © 2007 Victor Bahl. All Rights Reserved. End Device Certificate Qiu-CCR-2006 Step 3 & 4: Model Network Behavior & Perform Root Cause Analysis Faults Directory Root Cause Inject Candidate Faults Collect Data Raw Data Agent Module • SNMP MIBs • Performance Counters • Routing Table • Native WiFi Clean Data Topology Link Loads Signal Strength Simulate Network Delay Diagnosis Module © 2007 Victor Bahl. All Rights Reserved. Estimated Link Layer Performance Compare Measured Link Layer Performance Troubleshooting Framework Challenges [in Online Simulation based Diagnostics]: – – Accurately reproduce the behavior of the network inside a simulator Build a fault diagnosis technique using the simulator as a diagnosis tool Advantages – – Flexible & customizable for a large class of networks Captures complicated interactions – – within the network between the network & environment, and among multiple faults Extensible in its ability to detect new faults Allows what-if analysis © 2007 Victor Bahl. All Rights Reserved. Handling the Challenges Reproducing network behavior • • • • Identify the set of traces to collect Rule out erroneous data from the trace Drive the simulator with the cleaned traces Self-correct performance with feedback circuitry Building fault diagnosis • Use performance results from trace-driven simulation to establish normal behavior • Deviation from normal behavior indicates a potential fault • Identify root cause by efficiently searching over fault space to re-produce symptoms © 2007 Victor Bahl. All Rights Reserved. Step 5: Mitigation Responding to troubled spots – – – – Re-route traffic Rate-limit Change topology via power control & directional antenna control Flag environmental changes & problems Malfunctioning hardware – Launch DoS attacks against the possible attacker – etc. © 2007 Victor Bahl. All Rights Reserved. Sample Performance 25 node random topology Number of faults 4 6 8 10 12 14 Coverage 1 1 0.75 0.7 0.92 0.86 False Positive 0 0 0 0 0.25 0.29 Faults detected: - Random packet dropping - MAC misbehavior - External noise © 2007 Victor Bahl. All Rights Reserved. Mesh Deployments Discoveries & Innovations CMU’s Monarch Testbed • • • • • • Motivation: Study multi-hop routing protocols & their interaction with MobileIP Scale: 6 car-mounted nodes, 2 fixed Internet gateways Access: 900 MHz WaveLAN CSMA/CA (pre-802.11) Routing: DSR, MobileIP Applications: VANETs with car speeds of 0-50 KPH; rtk-GPS (< 1 cm) When: 1998-1999 © 2007 Victor Bahl. All Rights Reserved. CMU’s Monarch Innovations • MANETs outside the military! – – • Integration of MANET with Internet – • Node could roam from WLAN, to cellular IP service (CDPD), to DSR MANET cloud MANET/Internet gateway served as Foreign Agent MIP/DSR integration made MNs full members of the MANET Evaluation of predictive hand-off – • All nodes on the Internet via gateway Integration with MobileIPv4 – – – • Supported a mix of data voice data with PTT (walkie-talkie) quality, GPS updates, situational awareness data Telemetry: serial-line replacement, bulk file transfer Topology changes every 4 s, each link lasted at most 220 s Looked at both GPS/topology/predicted-trajectory and RSSI Identified routing problem caused by wireless propagation that is transiently better than expected © 2007 Victor Bahl. All Rights Reserved. MSR’s Mesh Network Project http://research.microsoft.com/mesh • Motivation: Alternative broadband technology for Internet access • Scenarios: Initial focus: NeighborNet (“community network that grows organically”) Longer term: WirelessOfficeNet, WirelessVillageNet, WirelessCityNet • Testbeds: MSR Redmond Offices, Redmond Apartment Complex, MSR Cambridge Offices - DVB distribution system • Design Objectives – – – – In-expensive (use .11-like radio - same PHY, possibly different MAC) Self managing - no truck-toll, take the human operator out of the equation High capacity – multi Mbps in the backbone, .11{a,b,g} on access links Coverage – typical suburban neighborhoods • Success Metric: Point of irrefutably, catalyst for wide-spread broadband Internet availability • When: 2003-2005 © 2007 Victor Bahl. All Rights Reserved. MSR’s Mesh Architecture End Device – Connects to a Mesh Router – Standards Compliant Network Interface Two Tier Architecture Internet ITAP 101 Bus Stop 206 Gas Station (Internet TAP) Mesh Router 7 EXIT 90 Mesh Router 5 Mesh Router / MeshBox – Routes traffic within the mesh and to the neighborhood Internet Gateway – Serves as access point for end devices Mesh Router 2 Mesh Router Mesh Router 3 MeshStreet Zone Any Mesh Router 1 Mesh End Device End Device (Guest to Router 1) End Device Neighborhood Internet Gateway – Gateway between the mesh nodes and the Internet Key: Multiple radios, cognitive software MSR’s Multi-Radio Mesh Box CPU 800 MHz VIA C3 Eden 133 MHz FSB DISPLAY Trident Blade 3D 2 PCI Slot Riser NORTH BRIDGE VIA Apollo PLE133 Video & RAM RAM 512 MB PC133 SDRAM PCI Bus LOW-FREQUENCY M ESH SOUTH BRIDGE VIA VT8231 Audio, ATA, USB, Ethernet, Serial, Parallel, PS/ 2 HIGH-FREQUENCY M ESH 802.11a D-Link DWL-AG520 802.11a Linksys WMP55AG 900 M Hz tbd 900 M Hz tbd NETWORK VT6103 AUDIO VT1612A TV OUT VT1621 USB HDD 512 MB Flash Card EPIA 800 M AINBOARD © 2007 Victor Bahl. All Rights Reserved. MSR’s All-Wireless Office (Testbeds 1 & 2) Details 201 Testbed 1: 23 to 30 nodes; Testbed 2: 45 to 60 nodes Inexpensive desktops (HP d530 SF) Two radios in each node – – – 210 220 205 203 226 204 NetGear WAG or WAB Proxim OriNOCO Cards can operate in a, b or g mode. 227 221 207 Verification of the mesh software stack – – – – Routing protocol behavior Fault diagnosis and mesh management algorithms Security and privacy architecture Range and robustness @ 5 GHz with different 802.11a hardware Study of Wireless Office Scenario Stress Testing Approx. 61 m Purpose 206 208 225 211 223 209 214 215 217 219 Various methods of loading testbed: – – – 224 Harpoon traffic generator (UWisconsin) Peer Metric traffic generator (Internal tool) Ad-hoc use by researchers © 2007 Victor Bahl. All Rights Reserved. 218 216 Approx. 32 m MSR’s Community Network (Testbed 3) Apartments in Redmond, WA Road B UNIT FF B UNIT GG B 31 B UNIT BB B A a FF102 FF203 BB103 a B b A ControlApt GG302 b GG202 Parking Lot 32 B BB302 ControlApt GG302 Bellaire Apts B A B BB201 Carport A A A HH301 Microsoft Campus B A A = MeshBox UNIT CC Road Control Apt GG302 Mesh Box 20 Feet UNIT HH Mesh Hall (Kitchen) © 2007 Victor Bahl. All Rights Reserved. B MSR’s Cambridge UK Trial (Testbed 4) 10 node 802.11a mesh Worked with ehome team to create a media sharing demo in collaboration with ZCast DVB trial MSR-Cambridge - 1st Floor, Mesh box Locations UK3Gtwy UK6 UK-MCE20 is the Kiosk with posters. = Mesh Box location = Mesh Box location © 2007 Victor Bahl. All Rights Reserved. UK1 UK8 UK MCE20 UK2 MSR’s Mesh Discoveries • Certain applications are not as attractive as were initially thought – Neighborhood distributed backup met with great suspicion • Percentage of node participation doesn’t have to be high for viable meshes to be formed – Low frequency meshes & high gain antennas can compensate for lower densities • Multi-radio meshes are necessary for reasonable capacity – A new link & interface selection metric is required for path selection • Performance largely dictated by Internet gateway placement – P2P applications work well regardless • Qiu-ICNP-2004 Hooks for management of the network have to be part of the design requirement – Monitoring, automatic alerts & root-cause analysis Qiu-CCR-2006 © 2007 Victor Bahl. All Rights Reserved. MSR’s All-Wireless Office Discoveries Eriksson-MobiSys-2006 Hardware – 802.11 hardware can result in 2.5x difference in performance Hardware from different vendors performs very differently Channel selection – Selection of 802.11 band can lead to 2x difference Automatically sniffing & selecting channels where there are fewer competing wireless networks helps Topology – Server placement can have up to 3x difference Experiment performed with central, distant, and random server placement Routing metrics – – Hop count works as well as WCETT under light loads With lots of cross traffic routing metrics matter 802.11 MAC – Enabling RTS and CTS has minimal benefit No benefit of spatial reuse, hidden node avoidance Most configurations – median delay <30ms © 2007 Victor Bahl. All Rights Reserved. MSR’s MCL Innovations • Software steerable directional antenna Kuga-NRSM-2004 – 8x8 dipole array, 20o beam-width, coplaner waveguide phase shifter, new form factor for indoor wall mounting, low loss feed network • Multi-radio meshboxes – Two radios on different channels for simultaneous rcv/send on the backbone, a third radio for standard’s based client connections, a forth radio for control & management channel • Dual Frequency Meshes – Lower (< 1 GHz) frequency mesh when node density is low, higher frequency mesh (2.4 GHz, 5 GHz) mesh when density and capacity reuqirement is higher • Multi-radio link selection metric – Weighted Cumulative ETT (WCETT) – Reduces to ETT for single radio case. Recognizes self-interference. • Link Quality Source Routing (LQSR) Draves-SIGCOMM-2004 Draves-MobiCom-2004 – Layer 2.5 design - hardware and upper protocol agnostic (supports both IPv4 and IPv6) – Includes multiple metrics incl. Hop-count, PacketPair, RTT, ETT, WCETT © 2007 Victor Bahl. All Rights Reserved. MSR’s MCL Innovations • Conflict-graph framework for calculating mesh capacity Jain-MobiCom-2003 – Allows “what-if” analysis when hardware / software node configurations and network density is changed • Single radio-channel hopping protocol for better spectrum utilization and improved capacity (SSCH) Bahl-MobiCom-2004 • A split-MAC control channel based protocol (C2M) Kyasanpur-Broadnets-2005 – Channel contention with advanced reservations for packet chains on lowbandwidth control channel; data transmission on the high bandwidth data channel • Academic toolkit – Source code and tools for 440 Universities world-wide using it – http://research.microsoft.com/mesh/kit/ © 2007 Victor Bahl. All Rights Reserved. MIT’s Roofnet http://pdos.lcs.mit.edu/roofnet/ • • • • • • Motivation: Unplanned community mesh network Location: Cambridge, Massachusetts, USA Scale: ~ 20 mesh nodes operational at any time, 3 sq. km coverage, 100+ users Access: IEEE 802.11b network Routing: ExOR with ETT Applications: Broadband Internet Access Briket-MobiCom-2005 © 2007 Victor Bahl. All Rights Reserved. MIT’s Roofnet Discoveries Aguayo-SIGCOMM-2004 • For organic networks, with no network planning lossy radio links are common – sometimes need to handle as much as 70% loss • Signal to Noise Ratio (SNR) does not predict delivery probability – Bursty noise can corrupt packets without affecting SNR But most links aren’t bursty • In single-radio meshes, multi-hop collisions cut bandwidth by greater than 2x • Minimum hop-count routing metric uses low-quality links. – More hops with better quality links generally perform better • Routing protocols should – incorporate link quality metrics (such as ETX, WCETT) – Opportunistic transmissions (MAC, ExOR) – Multicast distribution © 2007 Victor Bahl. All Rights Reserved. MIT’s Roofnet Discoveries • Higher node density improves connectivity and throughput – Even though the hop count increases • Even with 9 hops, were able to achieve reasonable throughput – Avg. TCP throughput: 152 Kbps, latency: 121 msec • N mesh Internet gateway nodes give better throughput than N (hotspots”) APs connected to the Internet – In mesh often use a sequence of short high-quality links rather than a single long low-quality link. © 2007 Victor Bahl. All Rights Reserved. MIT’s Roofnet Innovations • Self-installation kit – 8dBi Antenna, PC, .11b card, 50 ft. low (3 dB / 100 ft) loss cable, easy to install Linux based software • Link quality metric - Expected Transmission Count (ETX) – Predicts the total amount of time it would take to send a data packet along a route. Couto-MobiCom-2003 • Link quality aware source routing protocol (SrcrRR) – Tries to find the highest-throughput route between any pair of nodes. • Rate Selection Algorithm (SampleRate) – Selects bit-rate that will provide the highest throughput based on delivery probabilities measured at the different bit-rates. • Opportunistic routing (ExOR) Biswas-SIGCOMM-2005 © 2007 Victor Bahl. All Rights Reserved. Roofnet Kit Courtesy: Bhaskaran Raman IIT’s Digital Gangetic Plains (DGP) Project http://www.cse.iitk.ac.in/users/braman/dgp.html • • • • • • Motivation: WiFi for rural India Bhagwat-HotNets-2003 Location: Kanpur and vicinity, India Scale: 13 backbone mesh nodes, longest link ~ 39 Km, with multi-hop ~ 80 Km Access Speed: > 1 Mbps, Backhaul links: 11 Mbps Routing: Bridging Application: Phone service (Ashwini Project: video-based services: distance-education, tele-health, & agricultural advice) Mar ‘03 Mar ‘04 Point-to-Point 802.11 link Rasoolabad Sep ‘02 Safipur 802.11 for last-hop access within a village 5 Km 17.3 Km 22.5 Km Mandhana 12 Km Jun ‘02 Nov ‘02 Bithoor 0.9 Km 5.1 Km 12 Km Mohanpur Sep ‘04 3.5 Km Apr ‘02 IITK Rajajipuram/ Lucknow 39 Km Dec ‘02 Sarauhan Jun ‘03 Banthar 2004 - present End to end distance ~80 Km 23 Km 22 Km Dec ‘03 2.3 Km Lodhar Dhaura 17.5 Km 37 Km MS3 Jun ‘02 Aug ‘04 Sawayajpur Land-line access point (close to high-population density area) © 2007 Victor Bahl. All Rights Reserved. River Ganges IIT’s DGP Discoveries • Links seemed to sensitive to multi-path Chebrolu-MobiCom-2006 – Off-the-shelf hardware designed for indoor delay spreads (~100 ns max) doesn’t to too well outdoors ~ 1 microsec delay spread • System cost reduction is more important than spectral efficiency in rural settings • Careful network planning is necessary for predictable performance in long distance links – Line of sight necessary for long-distance links • 802.11 CSMA/CA is unsuitable for long-distance communication. – Timing parameters (need to account for RF propagation speeds) – Arbitrary contention resolution needs to be replaced by simple hop-by-hop scheduling • Interference causes drastic reduction in performance • Power conservation necessary. Turn off equipment when link idle – Wake-on-WLAN © 2007 Victor Bahl. All Rights Reserved. IIT’s DGP Project Innovations For long distance connectivity WiFi Radio – Built towers since line of sight is necessary Towers were expensive, function of height (45 m ~ $5K) – Antenna selection & transmit power made a big difference WiFi Radi o Antennae at Mandhana 802.11 (CSMA/CA) Issues – Single channel operation - only one link operational per node Designed a simple scenario-based scheduling – Tweaked MAC parameters to handle long distance links 802.11b/g Z X Raman-MobiCom-2005 RF switch or splitter Sensor Mote 12 V Battery Energy preservation via wake-on-WLAN – A second low-power radio Antenna IEEE 802.15.4 Sensor mote Clear channel when received energy is below threshold © 2007 Victor Bahl. All Rights Reserved. Soekris Power Switching Circuit Courtesy: Edward Knightly Rice’s Technology for All (TFA) Project http://www.tfa-wireless.ece.rice.edu/ • • • • • Motivation: “Empower low income communities through technology” Location: Houston’s East End Scale: 15 nodes deployed, 2 Km2 coverage, 700+ users Access Speed: > 1 Mbps, Backhaul links > 3 Mbps Application: Education and work-at-home Two-Tier Architecture • • • Limited gateways wired to Internet Backhaul tier - Mesh Access tier – Client to mesh node Coverage map with location of mesh nodes © 2007 Victor Bahl. All Rights Reserved. Rice’s TFA Discoveries • Don’t believe the manufacturer’s specifications Camp-MobiSys-2006 – Radio & antenna specifications severely optimistic • Measure before deploying. Deployment cost may be reduced if you have – Accurate knowledge of the RF Propagation environment Measured path loss = 3.3 (specified 2 to 5) – Knowledge of throughput-signal-strength function manufacturer values over-estimate link range by 3 times yielding disconnected network • When measuring always take concurrent traffic patterns into account • Only small number of measurements required for good prediction • Identification of the “Parking-Lot Fairness” Problem • Placement of Internet Gateway has a significant effect on mesh performance © 2007 Victor Bahl. All Rights Reserved. Rice’s TFA Innovations Programmable, single-radio mesh node with storage Client node Specifications: • 200 mW 802.11b • LocustWorld Mesh SW • VIA C3 1Ghz • 5 GB Hard Drives • 4 GB Flash to run Linux • HostAP driver • 15 dBi Omni-directional Antenna Specifications: • 200 mW 802.11b • 3 dBi Omni at 2 m • Ethernet Port Similar functionality to DSL or cable modem Camp-DC-2005 • Layer 2 rate-limiting algorithm for handling the parking lot fairness problem – Spatial Bias Nodes closer to the wired gateway obtain higher throughput Starvation occurs in nodes farthest from the gateway – Overhead of RTS/CTS outweighs slight gains for starved nodes © 2007 Victor Bahl. All Rights Reserved. Courtesy: Mark Corner and Brian Neil Levine UMASS’s DieselNet http://prisms.cs.umass.edu/diesel/ • • • • • Motivation: Disruption-tolerant mesh networking Location: Amherst, Massachusetts, USA Scale: 40 buses, 14 Mesh APs, 4 Throwboxes, 150 miles2 Access Speed: 802.11a/b, GPRS, MaxStream, XTend, CC1000 Applications: Mesh/DTN synergy, vehicular networking Amherst downtown and UMass © 2007 Victor Bahl. All Rights Reserved. DieselNet: Bus Equipment • 40 equipped buses, routes spanning 150 sq. mi. • Town center (4 sq.mi.) is hub of network. • Each bus: – HaCom Open Brick computer (577MHz CPU, 256MB RAM) – 802.11b USB dongle – Linksys/Netgear 802.11b AP – GPS Receiver – 40GB drive – XTend radio – GPRS – Custom software © 2007 Victor Bahl. All Rights Reserved. Dieselnet Innovations: ThrowBox • • • • TelosB Mote (sensor) 900 MHz XTend radio 8 Mhz microcontroller Stargate - 802.11b CF card - 400Mhz PXA255 Xscale - 64 MB ram, 32MB store • Java 1.3 • DieselNet code • AA rechargeable batteries / solar power © 2007 Victor Bahl. All Rights Reserved. DieselNet’s Throwbox Use DieselNet’s Discoveries • Real deployments must build in a strong software update and data logging mechanism. • Routing: Information about the network allows routing protocols to manage resource constraints correctly. (Use as much bandwidth as necessary during a short-lived DTN contact opportunity to exchange meta-data.) • DTNs are a failure mode for meshes; Meshes are a performance benefit for DTNs. • Two fundamental “Throwboxes” challenges: – Placement: Throwbox deployment that incorporates knowledge about contact opportunities performs better than deployment that ignores this knowledge. – Power management: long-range radio coupled with Mote used for discovery and mobility prediction; normal-range radio and Stargate used for transfer and routing. © 2007 Victor Bahl. All Rights Reserved. MadCity’s Madison Broadband Downtown Mesh • • • • • Motivation: Commercial endeavor, local ISP, $20/month to cover the entire city Location: Madison, Wisconsin, USA Scale: 200+ mesh nodes (not all completely functional), coverage 9 miles2 Access Speed: 802.11a speed in the backbone, 802.11b access link Applications: Broadband Internet Access © 2007 Victor Bahl. All Rights Reserved. MadCity’s Mesh Network • Carefully planned network – Grid-like topology. Every mesh node has at least one other mesh node win it’s line of sights • Two-tier network – 5.8 GHz (for the mesh backbone), 2.4 GHz (for client access) – Client connect to mesh as if connecting to infra-structure network AP • Mesh nodes run Cisco’s propriety software – Tree based protocol with AODV like features • Injection Points – A 18 GHz (6-mile) licensed link between a tower and the NOC Tower has a 3 ft. diameter antenna and located high above the city – The tower points to 3 mesh nodes (Internet Gateways) © 2007 Victor Bahl. All Rights Reserved. MadCity’s Discoveries From UW WiNGs Research Group • Multi-radio systems are preferable – Next version may have 3 radios, two for the backbone, 1 for access • Foliage has a significant negative impact on data rates – Performance in winter and summer is significantly different Up to factor of 2 – Neighboring mesh nodes have to be approximate in line-of-sight Carefully network planning is necessary Routing doesn’t change often • Directionality can help improve performance Use of sector antennas and polarization capabilities © 2007 Victor Bahl. All Rights Reserved. Courtesy: Krishna Ramachandran UCSB’s Multi-Channel Mesh Network http://moment.cs.ucsb.edu/meshnet/ • • • • • • Motivation: High-capacity multi-channel mesh network design Location: University of California Santa Barbara department offices Scale: 20 802.11a/b nodes in a 5 floor office building Access Speed: > 20 Mbps backhaul Routing: AODV with Link Metrics Application: Indoor broadband access (“Wireless Office”) UCSB MeshNet Map – subscript with node number indicates floor number © 2007 Victor Bahl. All Rights Reserved. UCSB’s Discoveries and Innovations Discoveries • Commodity radios have imperfect band filters – Inter-radio interference, such as board-crosstalk and radiation leakage, prevents construction of a single-unit multi-radio router – Physical separation alleviates inter-radio interference Innovations .11a & .11b • Split Wireless Router Architecture – Individual radios on separate processing nodes – Modular wireless router architecture – Backhaul formed using Wired / UWB .11a • TIC Multi-Channel Architecture – Avoids external interference Ramachandran-Infocom-2006 – Results in high-throughput, frequency-diversified routes between APs and mesh gateways © 2007 Victor Bahl. All Rights Reserved. .11a 4 radio Split Wireless Router consisting of 3 processing units and a 100 Mbps switch as as interconnect Courtesy: Saumitra M. Das Mesh@Purdue: Purdue’s Wireless Mesh Testbed http://engineering.purdue.edu/MESH/ • Motivation: Design, study and evaluate throughput improving architectures for mesh networks • Location: Purdue University CS Dept. • Scale: 32 dual-radio 802.11a/b routers deployed across 4 academic buildings with a few outdoor links • Access Speed: radio 1: 802.11b (~5 Mbps) radio 2: 802.11a (>20Mbps) • Application: Broadband internet access and community networking Purdue MAP testbed © 2007 Victor Bahl. All Rights Reserved. Purdue’s Discoveries and Innovations • Directional antennas in multi-radio mesh networks Das-JSAC-2006 – Spatial + frequency separation (directional antennas with channel assignment) is more powerful than spatial or frequency separation alone – Conservative channel assignment based on rough antenna characteristics is effective • Mitigating the gateway bottleneck using cooperative caching in mesh networks – Mesh routers can be leveraged as caches through expandable storage – Internet traffic exhibits locality. Hit rate 40~50% in traces of small user populations – Fetching cached content from other mesh routers can improve throughput (reduced gateway bottleneck, reduced hop count, multiple choices) Das-mobicom_wintech-2006 • Interference measurements in wireless mesh networks – Study of the impact of multi-way interference on a link (multiple simultaneous transmitters) and whether pair-wise interference measurements are adequate – Significant multi-way interference does occur although in a small fraction of cases. Protocols that schedule multiple transmitters need to take into account such effects by being conservative. © 2007 Victor Bahl. All Rights Reserved. Standards The contents of the next set of slides are derived directly from the IEEE 802.11s presentations made during the various IEEE meetings. The draft standard is a living document that is constantly being revised. Therefore it is possible that the contents of these slides may differ from the standard that finally issues. Reference for these slides: IEEE Document 802.11-06/0329r2, March 2006 What is IEEE 802.11s? Protocol specification for unmanaged mesh networks. – Primary target is a self-configuring multi-hop wireless distribution system that supports broadcast / multicast and unicast traffic – Mesh nodes interoperate with legacy nodes. Guidelines – Low complexity and to the extent possible, reuse previous technology Reuse 4-address frame format for exchanging packets between APs Build on top of 802.11i security Build on top of 802.11e MAC and power saving options Initiated in late 2003; Call for proposal ended in July 2005. Current Status – Draft Specification – (hopefully) Letter ballot review in November 2006 – Target completion: First half of 2008 © 2007 Victor Bahl. All Rights Reserved. IEEE 802.11s Background History – Started with 35 proposals, narrowed it down to 6 by July 2005; to 4 by Sept. 2005, to 2 by Nov. 2005 and to 1 by Mar. 2006 Final Proposal is co-authored by 38 companies. It’s a merge of: – SEEMesh Proposal (SEE – Simple, Efficient, Extensible) Intel, Nokia, Motorola, NTT DoCoMo, and TI – Wi-Mesh Proposal Accton, ComNets, InterDigitial, NextHop, Notel, Phillips, Extreme Networks, MITRE, Naval Research Laboratory, Swisscom Innovations and Thomson http://wi-mesh.org/ The proposal includes flexibility to define protocols / mechanisms and scenario-specific optimizations. Allows future extensions. The design supports both single-radio and multi-radio platforms. © 2007 Victor Bahl. All Rights Reserved. IEEE 802.11s in a Nutshell Discovery and Network Formation – Single-hop & multi-hop neighbor discovery via beacons & probe responses – Combinations of single radio / multi radio & single channel / multi-channel possible Inter-networking – 802.1 (STP) bridging support included Path Selection and Forwarding – Hybrid wireless mesh protocol default – Support for optional protocols (e.g. RA-OLSR) Security – 802.11i link security, support for distributed and centralized authentication MAC Enhancements – 802.11e (EDCA) with support for congestion control and optional multi-channel operation Optional Power Save – Reduced beaconing, single hop, APSD and ATIM/DTIM © 2007 Victor Bahl. All Rights Reserved. 802.11s Nomenclature Device Classes Station (STA) “Legacy” .11 nodes, allowed to initiate sessions Mesh Point (MP) Mesh node Mesh AP (MAP) MP + access point for communication with “legacy” nodes Mesh Portal Commonly referred to as ITAP or Internet Gateway in research literature but a little more generic Light Weight MP (LWMP) Subset of MP: primarily for neighbor-link communication © 2007 Victor Bahl. All Rights Reserved. 802.11s Mesh Formation • Neighbor Discovery is via beacons and probe response frames. A new Information Element is defined – Mesh ID – Name of the mesh – Mesh Capability Element – Summary of active protocol & metric – Channel coalescence mode and channel precedence indicators (see next) • MPs inform one another about the supported mesh services via IEs – e.g. Link state announcement, path selection information etc. • Membership is determined by secure association with neighbors – Mesh data services can be used by LWMPs without association © 2007 Victor Bahl. All Rights Reserved. 802.11s Multi-Radio Support MP may have one or more logical radio interface: – Each logical interface belongs to one “Unified Channel Graph” – Two possible modes for each interface: Simple channel unification mode – Follows rules to form a fully connected graph on one channel Advanced mode – Framework for flexible channel selection, algorithms/ policy (out of scope) – Each Unified Channel Graph shares a channel precedence value Channel precedence indicator – for coalescing disjoint graphs and supporting channel switching © 2007 Victor Bahl. All Rights Reserved. 802.11s Routing • All implementations support default mandatory protocol and metric – Any vendor may implement any protocol and/or metric – Only one protocol/metric is active on a particular link at a time – A particular mesh has only one active protocol • MP uses the Mesh Capability IE to indicate which protocol is in use • A mesh that is using other than mandatory protocol is not required to change its protocol when a new MP joins – (Algorithm to coordinate such a reconfiguration is out of scope) • Proposal for Path Selection Protocol – Default: HWMP (RM-AODV or Distance Vector Routing Tree) – Optional: Radio Aware OLSR (RA-OLSR) Based on RFC 3626 © 2007 Victor Bahl. All Rights Reserved. 802.11s Interworking with LAN Segments Scope – Connect a 802.11s mesh to a 802.1D bridged LAN Support – Transparent forwarding (“Broadcast LAN”) when bridge receives packets destined for outside the LAN – Learning by overhearing packets to certain destinations – Bridge-to-bridge communications using stand bridging techniques Note: 802.11s meshes support – Broadcast delivery – Unicast delivery – Multicast delivery © 2007 Victor Bahl. All Rights Reserved. 802.11s Bridging Functionality Bridges handle packet delivery to a node – – – in the mesh outside the mesh In a the mesh but reachable via another bridge over a LAN segment leverage layer-2 mesh path discovery to determine if the destination is inside or outside of the mesh IF destination inside the mesh – ELSE – – Use layer-2 routing to discover route & forward packets // destination outside the Mesh Identify the “right” gateway, and deliver packets via unicast If not known, deliver packets to all mesh portals / gateways © 2007 Victor Bahl. All Rights Reserved. 802.11s Packet Forwarding at the Bridge When a packet reaches the MAP, it makes bridging decisions IF destination is not known – Send packet on all ports ELSE IF destination is reachable via a non-mesh port – Deliver to that port ELSE IF destination is reachable via a mesh port IF packet came from the mesh – Drop it ELSE IF the packet came from outside the mesh – Use mesh delivery Note: bridge has to know where the destination is located. (out of scope) © 2007 Victor Bahl. All Rights Reserved. 802.11s Security Goals and Requirements • Builds on top of 802.11i security mechanisms – Both distributed & centralized authentication schemes supported – 802.11i provides link-security, .11s provides link-by-link security End-to-end security can be layered on top, e.g. using IPSEC • Allows association/authentication between neighboring MPs / MAPs – Pre-requisite to participating in the mesh • Protects mesh management & control messages exchanged between MPs / MAPs – Allows routing information to be authenticated © 2007 Victor Bahl. All Rights Reserved. 802.11s Basic Security Model • Mesh nodes belong to single logical administrative domain • Authentication schemes – Distributed Credentials are derived from certificates or PSK – PSK limits security -- no ability to reliably identify source of messages – Centralized At least one MP can directly access AAA server – other MPs authenticate via AAA-connected MPs via EAP Connection to AAA server built up hop-by-hop • Authenticated MP can be trusted – Mesh services (path selection protocol, data forwarding, etc.) © 2007 Victor Bahl. All Rights Reserved. 802.11s Basic Security Model (2) • Each MP acts as supplicant & authenticator for each of its neighbors – Similar to IBSS security model in 802.11i • Each MP uses 4-way handshake with each neighbor to establish session keys – Each MP uses its own group session key to broadcast/multicast and pair-wise session keys for unicast © 2007 Victor Bahl. All Rights Reserved. 802.11s Security Model Authenticator Supplicant security bubble New participant Notes: – Pair-wise keys for unicast communications – Group key for broadcast / multicast communications – Authentication can be distributed or centralized Compatible with 802.1X and PSK © 2007 Victor Bahl. All Rights Reserved. 802.11s Media Access Control Builds on 802.11e (EDCA) – Compatible with “legacy” devices – Simultaneously handles multi-hop mesh & single-hop BSS traffic • Prioritization policies left to vendor Mesh specific enhancements – Intra-mesh Congestion Control • Hop-by-hop congestion control mechanism implemented at each MP – Common Channel Framework (Optional) • Support for multi-channel MAC operation © 2007 Victor Bahl. All Rights Reserved. 802.11s Congestion Management Monitoring [Informative] – Each node actively monitors local channel utilization – If congestion detected, notifies previous-hop neighbors and/or the neighborhood Signaling mechanisms are defined – Congestion Control Request (unicast) – Congestion Control Response (unicast) – Neighborhood Congestion Announcement (broadcast) Rate control [Informative] – On receiving a congestion notification MP should adjust its traffic generation rate – Rate control (and signaling) on per-AC basis – e.g., data traffic rate may be adjusted without affecting voice traffic – Example: MAPs may adjust BSS EDCA parameters to alleviate congestion due to associated STAs © 2007 Victor Bahl. All Rights Reserved. 802.11s Common Channel Framework (Optional) Common channel is: – Unified Channel Graph on which MPs and MAPs operate. – In single radio MPs MPs switch from CC to a destination channel & return back. – In multi-radio MPs MPs may use a separate common channel for each interface Common Channel Framework (CCF) supports channel switching – After RTX/CTX exchange on common channel, MP pairs switch to a destination channel and then switch back – Groups of MPs may switch to a negotiated destination channel Neighbors discover support for CCF during association – Using the Mesh Capability IE in the beacon © 2007 Victor Bahl. All Rights Reserved. 802.11s: Channel Negotiation Protocol Control Frames – Request to Switch (RTX) – Clear to Switch (CTX) Kyasanur-BroadNets-2005 Channel Negotiation Protocol – Using RTX, the transmitter suggests a destination channel – Receiver accepts/declines the suggested channel using CTX – After a successful RTX/CTX exchange, the transmitter and receiver switch to the destination channel – Switching is limited to channels with little activity – Existing medium access schemes are reused © 2007 Victor Bahl. All Rights Reserved. 802.11s Accommodating Legacy Behavior • For devices that do not implement CCF, the common channel appears as a conventional channel • Common channel can be used for data transmission • A MAP with a single radio may use the common channel for WDS as well BSS traffic © 2007 Victor Bahl. All Rights Reserved. 802.11s Power Save Mechanisms (Optional) Mechanisms focused on powersave between neighbors – MPs that support powersave may enter sleep state – Sleep-wake cycles are not coordinated across multiple hops Two approaches – The Automatic Power Save Delivery (APSD) approach Similar to 802.11e APSD – Periodic APSD – Aperiodic APSD – The Announcement Traffic Indication Message (ATIM) / Delivery Traffic Indication Message (DTIM) approach – Well known wake times coordinated with well-known specific beacon times © 2007 Victor Bahl. All Rights Reserved. 802.11s: ATIM based Sleep-Wake Operation • Guaranteed window of awake time after periodic DTIM beacons • DTIM interval is a parameterized multiple of beacon intervals; globally unique to the mesh • Control communication in ATIM window – Indicating pending traffic – Indicating change in PS state – Re-instating of stopped flows • Remain awake after ATIM window depending on the control communication in it DTIM Interval ATIM window Beacon DTIM Interval ATIM window Beacon © 2007 Victor Bahl. All Rights Reserved. Time 802.11s: APSD based Sleep-Wake Operation • Similar to 802.11e APSD solution for wireless LANs • Periodic-APSD: Sleep-wake times coordinated with each neighbor separately and independently – Used for QoS traffic such as VoIP – Pairs of neighbors setup periodic schedules to wake up at set times • Aperiodic-APSD : MP in powersave state sends a packet to an ‘always awake’ neighbor to indicate it is awake – Used only with neighbors that are awake all the time – PS state MP sends a packet to the neighbor to indicate it is awake any time it wishes © 2007 Victor Bahl. All Rights Reserved. 802.11s Lightweight Mesh Points (LWMPs) • Participate in communications with neighbors – Association is not required when source and destination are neighbors • Use and provide a subset of mesh services – Incl. powersave, security, and unicast and broadcast delivery • Do not provide or use any distribution system services – Maximum allowed associations = 0 – Advertised routing profile = “Null” • Lightweight in terms of memory and processing requirements © 2007 Victor Bahl. All Rights Reserved. 802.11s Lightweight Meshing Association is not a pre-requisite – When the source and destination are neighbors When distribution system service is not required – When maximum number of associations allowed/possible at a MP is exhausted; lack of association should not preclude direct communication Mixed mode operation is okay – An MP may associate with some of its neighbors, and may communicate without association with other neighbors The DS service may be used through associated neighbors Neighbor communication possible with un-associated neighbors © 2007 Victor Bahl. All Rights Reserved. Additional Areas of Research Topics not covered in this tutorial Due to lack of time I was not be able cover several important research results that you should also be aware of. Some of these are: Modulation and PHY techniques like spread spectrum, OFDM etc. Adaptive antenna technologies like MIMO, beam forming, etc. TCP over meshes Multicast routing and group communications Topology control and power management Networking technologies including UWB, IEEE 802.16,… Spectrally-agile cognitive radios & networking Directional MACs etc © 2007 Victor Bahl. All Rights Reserved. Active Areas of Research – – – – – – – – – – – – Analytical tools for calculating mesh capacity Flow-level and packet-level fairness Network management & automatic diagnosis of faults Network coding for capacity improvement Routing with directional antennas / routing for network coding Supporting VoIP & video traffic over meshes Inexpensive software steerable directional antennas Smart medium access control Meshing with cognitive radios Multi-spectral meshes Delay tolerant meshing Usage scenarios © 2007 Victor Bahl. All Rights Reserved. Thanks! For prior work & updates, check out: http://research.microsoft.com/mesh/ References Papers are categorized under subject area. Duplication is possible because some papers include more than one problem/solution pair. This is not a exhaustive list. References - Testbeds UMASS’s DieselNet • • • • • • • [Zhao-MASS-2006] Wenrui Zhao, Yang Chen, Mostafa Ammar, Mark Corner, Brian N. Levine, and Ellen Zegura. Capacity Enhancement using Throwboxes in DTNs, IEEE Intl Conf on Mobile Ad hoc and Sensor Systems (MASS), Oct 2006. [Partan-WUWNet-2006] Jim Partan, Jim Kurose, Brian N. Levine, A Survey of Practical Issues in Underwater Networks. ACM Intl Wkshp on Underwater Networks (WUWNet), September 2006 [Jun-ACHANTS-2006] Hyewon Jun, Mostafa Ammar, Mark Corner, Ellen Zegura. Hierarchical Power Management in Disruption Tolerant Networks with Traffic-Aware Optimization, ACM CHANTS. September, 2006. [Burgess-INFOCOM-2006] John Burgess, Brian Gallagher, David Jensen, and Brian N. Levine. MaxProp: Routing for Vehicle-Based Disruption-Tolerant Networks, IEEE INFOCOM, April 2006. [Burns-ICRA-2006] Brendan Burns, Oliver Brock, and Brian N. Levine. Autonomous Enhancement of Disruption Tolerant Networks, IEEE Intl Conf on Robotics and Automation (ICRA), May 2006. [Burns-INFCOMM-2005] Brendan Burns, Oliver Brock, and B.N. Levine. MV routing and capacity building in disruption tolerant networks., IEEE INFOCOM, March 2005. [Hanna-ICNP-2003] Kat Hanna, Brian N. Levine, and R. Manmatha. Mobile Distributed Information Retrieval For Highly Partitioned Networks, IEEE ICNP, November 2003. © 2007 Victor Bahl. All Rights Reserved. References - Testbeds IITK’s Digital Gangetic Plains • [Chebrolu-MobiCom-2006] Kameswari Chebrolu, Bhaskaran Raman, and Sayandeep Sen, LongDistance 802.11b Links: Performance Measurements and Experience, 12th Annual International Conference on Mobile Computing and Networking (MOBICOM), Sep 2006, Los Angeles, USA. • [Raman-MobiCom-2005] Bhaskaran Raman and Kameswari Chebrolu, Design and Evaluation of a new MAC Protocol for Long-Distance 802.11 Mesh Networks, 11th Annual International Conference on Mobile Computing and Networking (MOBICOM), Aug/Sep 2005, Cologne, Germany. • [Bhagwat-HotNets-2003] Pravin Bhagwat, Bhaskaran Raman, and Dheeraj Sanghi, Turning 802.11 Inside-Out, Second Workshop on Hot Topics in Networks (HotNets-II), 20-21 Nov 2003, Cambridge, MA, USA. MIT’s RoofNet • • • • [Briket-MobiCom-2005] John Bicket, Daniel Aguayo, Sanjit Biswas, and Robert Morris, Architecture and Evaluation of an Unplanned 802.11b Mesh Network, ACM MobiCom 2005. [Biswas-SIGCOMM-2005] Sanjit Biswas and Robert Morris, Opportunistic Routing in Multi-Hop Wireless Networks, ACM SIGCOMM 2005 [Aguayo-SIGCOMM-2004] Daniel Aguayo, John Bicket, Sanjit Biswas, Glenn Judd, Robert Morris, Link-level Measurements from an 802.11b Mesh Network, SIGCOMM 2004, Aug 2004 [Couto-MobiCom-2003] Douglas S. J. De Couto, Daniel Aguayo, John Bicket, Robert Morris, A High-Throughput Path Metric for Multi-Hop Wireless Routing, ACM Mobicom 2003 © 2007 Victor Bahl. All Rights Reserved. References - Testbeds MSR’s Mesh Network • • • • • • • • • [Qiu-CCR-2006] Lili Qiu, Paramvir Bahl, Ananth Rao, Lidong Zou, “Troubleshooting Wireless Meshes”, ACM Computer Communications Review 2006 [Eriksson-MobiSys-2006] Jacob Eriksson, Sharad Agarwal, Paramvir. Bahl, Jitendra Padhye, Feasibility Study of Mesh Networks for All-Wireless Offices, ACM/USENIX MobiSys, Uppsala, Sweden, June 2006 [Kyasanpur-Broadnets-2005] Pradeep Kyasanur, Jitendra Padhye, Paramvir Bahl, On the Efficacy of Separating Control and Data into Different Frequency Bands, IEEE BroadNets 2005 , Boston, Massachusetts, USA (October 2005) [Padhye-IMC-2005] Jitendra Padhye, Sharad Agarwal, Venkata Padmanabhan, lili Qiu, A. Rao, Brian Zill, “Estimation of Link Interference in Static Multi-hop Wireless Networks”, ACM Internet Measurement Conference 2005, October 2005 [Kuga-NRSM-2004] Y Kuga, J. Cha, J. A. Ritcey, James Kajiya, Mechanically Steerable Antennas Using Dielectric Phase Shifters, IEEE AP-S International Symposium and USNC/URSI National Radio Science Meeting, June 2004 [Qiu-ICNP-2004] Lili Qiu, Ranveer Chandra, Kamal Jain, M. Mahdian, Optimizing the Placement of Integration Points in Multi-hop Wireless Networks, IEEE ICNP 2004. [Draves-MobiCom-2004] Richard Draves, Jitendra Padhye, Brian Zill, Routing in Multi-radio, Multihop Wireless Mesh Networks, ACM MobiCom, Philadelphia, PA, September 2004. [Bahl-MobiCom-2004] Paramvir Bahl, Ranveer Chandra, John Dunagan, SSCH: Slotted Seeded Channel Hopping for Capacity Improvement in IEEE 802.11 Ad-Hoc Wireless Networks, ACM MobiCom, Philadelphia, PA, September 2004. [Draves-SIGCOMM-2004] Richard Draves, Jitendra Padhye, Brian Zill, Comparison of Routing Metrics for Static Multi-Hop Wireless Networks, ACM SIGCOMM, Portland, OR, August 2004. © 2007 Victor Bahl. All Rights Reserved. References - Testbeds MSR’s Mesh Network • • [Qiu-MSRTR-2003] Lili Qiu, Paramvir Bahl, A. Rao, Lidong Zhou, Fault Detection, Isolation, and Diagnosis in Multi-hop Wireless Networks, Microsoft Research Technical Report, TR-2004-11 [Jain-MobiCom-2003] K. Jain, J. Padhye, V. Padmanabhan, and L. Qiu, Impact of Interference on Multi-hop Wireless Network Performance, ACM MobiCom, San Diego, CA, September 2003 Rice’s TFA • [Camp-MobiSys-2006] Joseph Camp, J. Robinson, C. Steger, Edward Knightly, "Measurement Driven Deployment of a Two-Tier Urban Mesh Access Network," in Proceedings of ACM MobiSys 2006, Uppsala, Sweden, June 2006. • [Camp-DC-2005] Joseph Camp, Edward Knightly, W. Reed, "Developing and Deploying Multihop Wireless Networks for Low-Income Communities," in Proceedings of Digital Communities 2005, Napoli, Italy, June 2005. JHU’s SMESH • [Amir-MobiSys-2006] Yair Amir, Claudiu Danilov, Michael Hilsdale, Raluca Musaloiu-Elefteri, Nilo Rivera, “Fast Handoff for Seamless Wireless Mesh Networks”, ACM/USENIX MobiSys 2006, June 2006 © 2007 Victor Bahl. All Rights Reserved. References - Testbeds Purdue’s Mesh • • [Das-JSAC-2006] Saumitra M. Das, Himabindu Pucha, Dimitrios Koutsonikolas, Y. Charlie Hu, Dimitrios Peroulis, “DMesh: Incorporating Practical Directional Antennas in Multi-Channel Wireless Mesh Networks”, IEEE Journal on Selected Areas in Communications (JSAC 06) special issue on Multi-Hop Wireless Mesh Networks, 2006. [Das-mobicom_wintech-2006] Saumitra M. Das, Dimitrios Koutsonikolas, Y. Charlie Hu, Dimitrios Peroulis, “Characterizing Multi-Way Interference In Wireless Mesh Networks”, ACM MobiCom International Workshop on Wireless Network Testbeds, Experimental evaluation and Characterization (MobiCom WiNTECH 06), Los Angeles, CA, September 29, 2006. © 2007 Victor Bahl. All Rights Reserved. References - Fairness • • [Nandagopal-MobiCom-2000] Thyagarajan Nandagopal, Tae-Eun Kim, Xia Gao, Vadhuvar Bharghavan, “Achieving MAC Layer Fairness in Wireless Packet Networks,” ACM MobiCom 2000, August 2000 [Luo-MobiCom-2000] Haiyun Luo, Songwu Lu, Vadhuvar Bharghavan, “A New Model for Packet Scheduling in Multihop Wireless Networks”, ACM MobiCom 2000, August 2000 References – Capacity / Channelization • • • • • • • [MSBA-SIGMETRICS-2006] Arunesh Mishra, Vivek Shrivastava, Suman Banerjee, William Arbaugh, Partially-overlapped Channels not Considered Harmful, ACM SIGMETRICS, June 2006. [Ramachandran-Infocom-2006] K. Ramachandran, E. Belding, K. Almeroth, M. Buddhikot, “Interference-Aware Channel Assignment in Multi-Radio Wireless Mesh Networks”, IEEE Infocom, Barcelona, Spain, April 2006 [Kyasanur-WCM-2006] Pradeep Kyasanur, Jungmin So, Chandrakanth Chereddi, and Nitin H. Vaidya, "Multi-Channel Mesh Networks: Challenges and Protocols", IEEE Wireless Communications, April 2006 [Kyasanur-BroadNets-2005] Pradeep Kyasanur, Jitendra Padhye, and Paramvir Bahl, "On the Efficacy of Separating Control and Data into Different Frequency Bands", in IEEE Broadnets, Boston, MA, October 2005 [Raniwala-Infocom-2005] Ashish Raniwala and Tzi-cker Chiueh, “Architecture and Algorithms for an IEEE 802.11-based Multi-Channel Wireless Mesh Network”, IEEE INFOCOM, 2005 Alicherry-MobiCom-2005] Mansoor Alicherry, Randeep Bhatia, Li Li, “Joint Channel Assignment and Routing for Throughput Optimization in Multi-radio Wireless Mesh Networks”, ACM MobiCom 2005, September 2005 [Kodialam-MobiCom-2005] Muralidharan Kodialam, Thyaga Nandagopal, “Characterizing the Capacity Region in Multi-Radio, Multi-Channel Wireless Mesh Networks”, ACM MobiCom 2005, September 2005 © 2007 Victor Bahl. All Rights Reserved. References – Capacity / Channelization • • • • • • • • [Kyasanur-MobiCom-2005] Pradeep Kyasanur, Nitin Vaidya, Capacity of Multi-Channel Wireless Networks: Impact of Number of Channels and Interfaces”, ACM MobiCom 2005, September 2005 [Adya-BroadNets-2004] Atul Adya, Paramvir Bahl, Jitendra Padhye, Alec Wolman, and Lidong Zhou, “A Multi-Radio Unification Protocol for IEEE 802.11 Wireless Networks”, IEEE BroadNets 2004. [Bahl-MobiCom-2004] Paramvir Bahl, Ranveer Chandra, John Dunagan, “SSCH: Slotted Seeded Channel Hopping for Capacity Improvement in IEEE 802.11 Ad-Hoc Wireless Networks”, ACM MobiCom, Philadelphia, PA, September 2004 [So-MobiHoc-2004], Jungmin So, Nitin Vaidya, “Multi-Channel MAC for Ad Hoc Networks: Handling Multi-Channel Hidden Terminals Using A Single Transceiver”, ACM MobiHoc 2004, Tokyo, Japan, May 2004 [Jain-MobiCom-2003] Kamal Jain, Jitendra Padhye, Venkata Padmanabhan, and Lili Qiu, “Impact of Interference on Multi-hop Wireless Network Performance”, ACM MobiCom, San Diego, CA, September 2003 [Jinyang-MobiCom-2001] Jinyang Li, Charles Blake, Douglas S. J. De Couto, Hu Imm Lee, and Robert Morris, Capacity of Ad Hoc Wireless Networks, ACM MobiCom '01), Rome, Italy, July 2001 [Gupta-IEEEIT-2000] Piyush Gupta, P. R. Kumar, “Capacity of Wireless Networks”, IEEE Transactions on Information Theory, March 2000. [Wu-ISPAN-2000] S.L. Wu, C.Y. Lin, Y.C. Tseng, J.P. Sheu, “A New Multi-Channel MAC Protocol with On-Demand Channel Assignment for Mobile Ad Hoc Networks”, International Symposium on Parallel Architectures, Algorithms and Networks (I-SPAN), 2000 © 2007 Victor Bahl. All Rights Reserved. References - Routing IETF Standards • • • • • [Corson-RFC2501-1999] Scott Corson, Joseph Macket, “Mobile Ad hoc Networking (MANET): Routing Protocol Performance Issues and Evaluation Considerations”, IETF RFC 2501, January 1999 [Perkins-RFC3561-2003] Charlie Perkins, Elizabeth Beilding-Royer, Samir Das, “Ad Hoc On Demand Distance Vector (AODV) Routing”, IETF RFC 3561, July 2003 [Clausen-RFC3626-2003] T. Clausen, P. Jacquet, “Optimized Link State Routing Protocol (OLSR)”, IETF RFC 3626, October 2003 [Ogier-RFC3684-2004] R. Ogier, Fred Templin, Mark Lewis, “Topology Dissemination Based on Reverse-Path Forwarding (TBRPF)”, IETF RFC 3684, February 2004 [Johnson-RFC4728-2007] D. Johnson, Y. Hu, D. Maltz, “The Dynamic Source Routing Protocol (DSR) for Mobile Ad Hoc Networks for IPv4”, IETF RFC 4728, February 2007 Papers & Drafts • • • • [Biswas-SIGCOMM-2005] Sanjit Biswas and Robert Morris, Opportunistic Routing in Multi-Hop Wireless Networks, ACM SIGCOMM 2005 [Draves-SIGCOMM-2004] Richard Draves, Jitendra Padhye, Brian Zill, Comparison of Routing Metrics for Static Multi-Hop Wireless Networks, ACM SIGCOMM, Portland, OR, August 2004. [Johnson-Draft-2004] David B. Johnson, David A. Maltz, Yih-Chun Hu, “The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks (DSR)”, IETF Draft, July 2004 [Ni-MobiCom-1999] S. Ni, Y. Tseng, Y. Chen, J. Chen, “The Broadcast Storm Problem in a Mobile Ad Hoc Network,” ACM MobiCom ’99, Seattle, Washington, August 1999 © 2007 Victor Bahl. All Rights Reserved. References - Routing Papers & Drafts • • • • [Alicherry-MobiCom-2005] Mansoor Alicherry, Randeep Bhatia, Li Li, “Joint Channel Assignment and Routing for Throughput Optimization in Multi-radio Wireless Mesh Networks”, ACM MobiCom 2005, September 2005 [Couto-HotNets-2002] Douglas De Couto, Daniel Aguayo, Benjamin Chambers, Robert Morris, “Performance of Multihop Wireless Networks: Shortest Path is Not Enough”, First Workshop on Hot Topics in Networks (HotNets-I), October 2002 [Broch-MobiCom-1998] Josh Broch, David A. Maltz, David B. Johnson, Yih-Chun Hu, and Jorjeta Jetcheva. “A Performance Comparison of Multi-Hop Wireless Ad Hoc Network Routing Protocol”,. ACM MobiCom'98, Oct. 1998. [Perkins-CCR-1994] Charlie Perkins and Pravin Bhagwat, “Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobile Computers,”, Computer Communications Review 24(4), Oct. 1994 © 2007 Victor Bahl. All Rights Reserved. References – Security & Management • • • • • • • • • • • [Qiu-CCR-2006] Lili Qiu, Paramvir Bahl, Ananth Rao, Lidong Zou, “Troubleshooting Wireless Meshes”, ACM Computer Communications Review 2006 [Badonnei-JNM-2005] Remi Badonnel, Radu State, Olivier Festor “Management of Mobile Ad Hoc Networks: Information Model and Probe-based Architecture”, International Journal of Network Management, Vol. 15, Issue 5, September 2005 [Kyasanur-TMC-2005] Pradeep Kyasanur, Nitin Vaidya, “Selfish MAC Layer Misbehavior in Wireless Networks”, IEEE Transactions on Mobile Computing, Vol. 4, Issue 4, September 2005 [Hu-Infocom-2003] Yih-Chun Hu, Adrian Perrig, David B. Johnson, “Packet Leashes: A Defense Against Wormhole Attacks in Wireless Ad Hoc Networks,”, IEEE INFOCOM 2003 [Qiu-MSRTR-2003] Lili Qiu, Paramvir Bahl, A. Rao, Lidong Zhou, Fault Detection, Isolation, and Diagnosis in Multi-hop Wireless Networks, Microsoft Research Technical Report, TR-2004-11 [Shen-Milcom-2002] C.-C. Shen, C. Jaikaeo, C. Srisathapornphat, Z. Huang, “The GUERILLA Management Architecture for Ad Hoc networks”, IEEE MILCOM 2002, Oct. 2002 [Hu-MobiCom-2002] Yih-Chun Hu, Adrian Perrig, David B. Johnson, “Ariadne: A Aecure OnDemand Routing Protocol for Ad Hoc Networks”, ACM MobiCom 2002, Atlanta, GA [Bucheggar-MobiHoc-2002] S. Buchegger, J. Y. Le Boudec, “Performance Analysis of the CONFIDANT Protocol”, ACM MobiHoc 2002 [Marti-MobiCom-2000] Sergio Marti, T. J. Giuli, Kevin Lai, Mary Baker, “Mitigating Routing Misbehavior in Mobile Ad Hoc Networks”, ACM MobiCom 2000, Boston, MA [Zhang-MobiCom-2000] Yongguang Zhang , Wenke Lww, “Intrusion Detection in Wireless Ad Hoc Networks”, ACM MobiCom 2000, Boston, MA [Chen-JSAC-1999] Wenli Chen, Nitin Jain, Suresh Singh, “ANMP: Ad Hoc Network Management Protocol”, IEEE Journal on Selected Areas in Communications, Volume 17 Number 8, August 1999 © 2007 Victor Bahl. All Rights Reserved. References - General Network Management • [Badonnel-JTS-2005] Remi Badonnel, Radu State, Olivier Festor, “Self-Organized Monitoring of Ad Hoc Networks”, Journal of Telecommunications Systems, Vol. 30, No. 1-3, 2005 General • • • • • [Andel-Computer-2006] Todd R. Andel, Alec Yasinsac, “On the Credibility of MANET Simulations”, IEEE Computer Magazine, pp. 48-54, 2006 [IEEE-Mesh-2006] Joint SEE-Mesh/Wi-Mesh Proposal to 802.11 TGs Overview”, IEEE 802.1106/0329r2, March 6, 2006 [Karn-CNC-1997] Phil Karn, “MACA – A New Channel Access Method for Packet Radio”, in Proceedings of AARL/CRRL Amateur Radio 9th Computer Networking Conference, Sept. 1997 [Powers-Proc-1995] R. A. Powers. “Batteries for Low Power Electronics”, Proceedings of the IEEE,, April 1995 [Tobagi-Comm-1975] F. A. Tobag, L. Klienrock, “Packet Switching in Radio Channels: Part II – The hidden Terminal Problem in a Carrier Sense Multiple-Access Modes and Busy-Tone Solution”, IEEE Transaction Communications, Vol. COM-23, no. 12, 1975 © 2007 Victor Bahl. All Rights Reserved.