1 / 51

High Performance Mobile Ad hoc Networking

High Performance Mobile Ad hoc Networking. Herbert Rubens Baruch Awerbuch herb@cs.jhu.edu baruch@cs.jhu.edu. Johns Hopkins University Department of Computer Science. Wireless Communication Lab wireless.cs.jhu.edu. Presentation Overview. Mobile Ad hoc Networking Overview

Télécharger la présentation

High Performance Mobile Ad hoc Networking

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. High Performance Mobile Ad hoc Networking Herbert Rubens Baruch Awerbuch herb@cs.jhu.edubaruch@cs.jhu.edu Johns Hopkins University Department of Computer Science Wireless Communication Lab wireless.cs.jhu.edu

  2. Presentation Overview • Mobile Ad hoc Networking Overview • Research Contributions • Related Work • The Pulse Protocol • The Medium Time Metric • Wave Relay System Feel free to ask questions throughout the presentation!

  3. Mobile Ad hoc Network • A self configuring network of mobile routers connected by wireless links • The routers may move freely, creating arbitrary network topologies • The network topology can change rapidly and unpredictably • Nodes communicate by wirelessly forwarding or relaying data through intermediate nodes • The network can be connected to the larger Internet or operate independently http://en.wikipedia.org/wiki/Mobile_ad-hoc_network

  4. JHU Wave Relay Network

  5. Node Locations Determine Topology

  6. Mobile Ad hoc Networking Timeline Ad hoc On-demand Distance Vector (AODV) Destination Sequenced Distance Vector (DSDV) Functions and Structure of a Packet Radio Station CD Player Apple Founded Apple IIgs Intel 486 Windows 3.0 Y2K 1991 www 1995 1975 1985 Microsoft Founded Today Herbert Benjamin Rubens 1979 DARPA Packet Radio Networks Dynamic Source Routing (DSR) Optimized Link-State Routing Protocol (OLSR) Burchfiel, J., Tomlinson, R., Beeler, M. (1975). "Functions and structure of a packet radio station". AFIPS: 245. Kahn, R. E. (January 1977). "The Organization of Computer Resources into a Packet Radio Network". IEEE Transactions on CommunicationsCOM-25 (1): 169–178. Kahn, R. E., Gronemeyer, S. A., Burchfiel, J., Kunzelman, R. C. (November 1978). "Advances in Packet Radio Technology". Proceedings of IEEE66 (11): 1468–1496. Jubin, J., and Tornow, J. D. (January 1987). "The DARPA Packet Radio Network Protocols". Proceedings of the IEEE75 (1).

  7. Fundamental Challenges • Complex dynamics of a wireless link • Continuously fluctuating RF environment (without mobility!) • Bit Error Rate • = small packets more reliable then large packets • Modulation • Different modulations work better in different RF environments • Multi-path, channel fading, delay spread • Link Capacity • Mobility • Further increases wireless link dynamics • Creates hard transitions • walk around a corner and everything changes • If all of the links are continuously changing, how do you select a set of links to form a path?

  8. Research Objectives • Scalability • Design routing algorithms which scale to thousands of devices while minimizing control overhead • Routing algorithm must perform under vehicular mobility, urban channel fading, and arbitrary communication patterns • Efficiency • Selected routes must: • Maximize individual path capacity • Minimize network resource consumption • Continuously adapt to changes

  9. Research Contributions • Medium Time Metric (MTM) • First route selection metric to consider multi-rate radios • Provably optimal route selection in small to medium sized networks • Experimental results and simulated results validate approach • Pulse Protocol • Extremely scalable routing protocol designed for mobile networks • Optimized for infrastructure access and peer-to-peer traffic patterns • Protocol extensions provide integrated time synchronization and power saving • Sensor Network Pulse Protocol • Directly trades route activation delay for power saving efficiency • Optimized for infrequently changing sensor network topologies • Optimized for sensor to collector traffic model

  10. MONET Journal –“The Medium Time Metric: High Throughput Route Selection in Multi-rate Wireless Networks” WONS 2005 –“The Pulse Protocol: Mobile Ad hoc Network Performance Evaluation” MILCOM 2004 – “The Pulse Protocol: Sensor Network Routing and Power Saving” INFOCOM 2004 – “The Pulse Protocol: Energy Efficient Infrastructure Access” WONS 2004 – “High Throughput Route Selection in Multi-rate Wireless Networks” ESAS 2006 –“Dynamics of Learning Algorithms for the On-Demand Secure Byzantine Routing Protocol” SECURECOM 2005 –“On the Survivability of Routing Protocols in Ad Hoc Wireless Networks” NDSS 2005 –“Secure Multi-hop Infrastructure Access” INFOCOM 2005 –“Provably Competitive Adaptive Routing” IZS 2004 – “Swarm Intelligence Routing Resilient to Byzantine Adversaries” WiSE 2002 –“An On-Demand Secure Routing Protocol Resilient to Byzantine Failures” Publications Relevant to Thesis Other work

  11. Existing Approaches Urban Channel Environment Receivers • Multi-path fading & shadowing • Rapidly changing channel conditions Reactive On-Demand Protocols (AODV, DSR) • On-demand protocols have no prior knowledge of channels conditions • A RREQ packet provides only a single sample of a complex distribution Proactive Link State Protocols (OLSR, TBRPF) Destination Source • Channel is continuously changing • Continuous flooding from every node in the network • Hello Protocol – detects link changes You can not accurately track channel with control packets!

  12. How Often Does Connectivity Change? • 10% of min-hop paths fail within 1.3 seconds • After 5 seconds 25% of min-hop paths have failed • On-Demand routes may only work for a short period of time • Link State Protocols need to flood every time a link changes • These simulations only consider changes from connected  not connected (in free space) • What about changes in link speed? Reliability? Hard transitions in a real environment? Fast-fading and urban channel effects? • Connectivity is continuously changing at an extremely fast rate! • Simulation: • 100 Nodes • 1000m x 1000m area • Random Waypoint Mobility (Max Speed=20m/s) • Calculate All-to-All shortest path initially, then track how long until the route fails

  13. Pulse Protocol Outline • Pulse Protocol Overview • Scalable multi-hop ad hoc routing protocol • Based on Tree Routing • Tree Routing vs. Direct Routing

  14. The Pulse Protocol • Proactive Component • Tracks minimum amount of information to avoid flooding for route establishment and maintenance • Periodic flood operation (similar to Hello Protocol) • Proactively rebuilds spanning tree • Estimates neighbors, density, SNR, loss rates, capabilities, number of radios, MTM metric • On-Demand Component • Route establishment • Using only UNICASTS! • Gratuitous mechanism • Neighbors promiscuously monitor packets • Metric tracked at the speed of data packets NOT control packets! • Path switches as metrics change • Local changes in connectivity only generate local traffic • Unlike BOTH on-demand and link state protocols

  15. Ad hoc Nodes

  16. Network Connectivity

  17. Pulse Flood

  18. Spanning Tree

  19. Source and Destination Need to Establish a Path

  20. Pulse Response Sent to Root

  21. Destination Paged on Next Pulse

  22. Destination Sends Pulse Response

  23. Path Option 1: Through the Root Through the Root Path Shortest Path 9 Hops 2 Hops This option is inefficient! It is not necessary to go to the root. Better routes already exist!

  24. Path Option 2: Tree Traversal Tree Traversal Path Shortest Path 5 Hops 2 Hops

  25. Path Option 3: Tree Shortcut Tree Shortcut Path Shortest Path 3 Hops 2 Hops This is the initially selected path of the Pulse protocol.

  26. Path Optimization: Gratuitous Reply Selected Path Shortest Path 2 Hops 2 Hops Node sends gratuitous reply

  27. Proactive Route Maintenance

  28. Proactive Route Maintenance

  29. Tree Routing vs. Direct Routing • Direct Routing • Attempts to initially discover the shortest path • Requires large overhead • Link state • tracks every link in the network regardless of whether it is used • a shortest path spanning tree for every node in the network • On-Demand • floods the network to establish a route • re-floods when ever the path breaks • a shortest path spanning tree for all nodes transferring data • Tree Routing • Proactively rebuilds a single spanning tree on top of the network • Boot straps communication off of the tree route • Route are not initially the direct shortest path, but routing mechanism allows the path to converge towards the shortest path • Active destinations can be reached without flooding the network • Efficient operation for realistic traffic patterns

  30. Pulse Protocol Concepts • Aggregation – for scalability • Spanning tree represents a compressed view of the network topology • Pro-active component maintains the minimum amount of information to allow efficient route establishment • De-Aggregation – for efficiency • The routing metric is tracked at the speed of the data flow • Changes to the metric are only reported locally • Routes are continuously adjusted as the metrics change • High speed accurate route tracking is essentially an on-demand decompression of the topology • However, it occurs ONLY in areas of the network with active data flows • Result: a scalable routing structure which tracks paths at the speed of the data flow

  31. Internet Gateway Example • All nodes routing to centrally located internet gateway • Best possible case for Pulse Protocol • Pulse source is designated as the centrally located gateway • Representative of Pulse internet access deployment at JHU • Similar to DoD “Reach Back” model Representative of most common real-world communication model

  32. DSR Pulse Delivery Ratio Simulations • Pure peer-to-peer communication pattern • Pulse source is an arbitrary mobile node

  33. SNS Scalability Simulation 10 km • Size: 10 km x 10 km • Nodes: 5,000 • Speed: 1 m/s • Traffic: 5 Mbps • Delivery Ratio: 97.2% 10 km Links: 50,000 on average • 100 stationary backbone nodes were arranged in a 10 by 10 grid • 5000 nodes were randomly placed and moved randomly • Exponential random traffic pattern was used • A network of 5,000 nodes could contain up to 25 million wireless links.

  34. Medium Time Metric Outline • Why do wireless radios operate at multiple rates? • Minimum Hop Metric shortcomings • Medium Time Metric

  35. Advantage of Multi-Rate? • Direct relationship between communication rate and the channel quality required for that rate • As distance increases, channel quality decreases • Therefore: tradeoff between communication range and link speed • Multi-rate provides flexibility 1 Mbps 2 Mbps 5.5 Mbps 11 Mbps • 802.11g • 1,2,5,6,11,12,18,24,36,48,54 Mbps • 802.11n (draft) • A lot more! Up to 300 Mbps.

  36. Challenge to the Routing Protocol • Must select a path from Source to Destination • Links operate at different speeds • Fundamental Tradeoff • Fast/Short links = low range = many hops/transmissions to get to destination • Slow/Long links = long range = few hops/transmissions

  37. Minimum Hop Metric(Traditional Technique) • Not designed for multi-rate networks • A small number of long slow hops provide the minimum hop path • These slow transmissions occupy the medium for long times, blocking adjacent senders • Selecting nodes on the fringe of the communication range results in reduced reliability

  38. New Approach: Medium Time Metric (MTM) • Assigns a weight to each link proportional to the amount of medium time consumed by transmitting a packet on the link • Enables the Pulse protocol to discover the path that minimizes total transmission time

  39. MTM Example Medium Time Usage Link Throughput Destination 4.55 Mbps 11 Mbps 2.5ms 3.17 Mbps 5.5 Mbps 3.7ms 1.54 Mbps 2 Mbps 7.6ms 0.85 Mbps 1 Mbps 13.9ms Source Path Medium Time Metric (MTM) Path Throughput 11 Mbps 5.5 Mbps 1 0.85 Mbps 13.9ms = 13.9 ms 2 Mbps 1 Mbps

  40. MTM Example Medium Time Usage Link Throughput Destination 4.55 Mbps 11 Mbps 2.5ms 3.17 Mbps 5.5 Mbps 3.7ms 1.54 Mbps 2 Mbps 7.6ms 0.85 Mbps 1 Mbps 13.9ms Source Path Medium Time Metric (MTM) Path Throughput 5.5 + 2 11 Mbps = 11.3 ms 1.04 Mbps 3.7ms 7.6ms 5.5 Mbps 1 0.85 Mbps 13.9ms = 13.9 ms 2 Mbps 1 Mbps

  41. MTM Example Medium Time Usage Link Throughput Destination 4.55 Mbps 11 Mbps 2.5ms 3.17 Mbps 5.5 Mbps 3.7ms 1.54 Mbps 2 Mbps 7.6ms 0.85 Mbps 1 Mbps 13.9ms Source Path Medium Time Metric (MTM) Path Throughput 11 + 2 1.15 Mbps 2.5ms 7.6ms = 10.1 ms 5.5 + 2 11 Mbps = 11.3 ms 1.04 Mbps 3.7ms 7.6ms 5.5 Mbps 1 0.85 Mbps 13.9ms = 13.9 ms 2 Mbps 1 Mbps

  42. MTM Example Medium Time Usage Link Throughput Destination 4.55 Mbps 11 Mbps 2.5ms 3.17 Mbps 5.5 Mbps 3.7ms 1.54 Mbps 2 Mbps 7.6ms 0.85 Mbps 1 Mbps 13.9ms Source Path Medium Time Metric (MTM) Path Throughput 11 + 11 = 5.0 ms 2.5ms 2.5ms 2.38 Mbps 11 + 2 1.15 Mbps 2.5ms 7.6ms = 10.1 ms 5.5 + 2 11 Mbps = 11.3 ms 1.04 Mbps 3.7ms 7.6ms 5.5 Mbps 1 0.85 Mbps 13.9ms = 13.9 ms 2 Mbps 1 Mbps

  43. MTM Advantages • Paths which minimize network utilization, maximize network capacity • Global optimum under complete interference • Excellent heuristic in even larger networks • Avoiding low speed links inherently provides increased route stability • High speed links operate with greater margin and are more elastic under changes • Experimental results show up to 17 times greater throughput using MTM in 802.11g networks

  44. Wave Relay Systemand Test-bed

  45. Wave Relay Test-bed • Over 50 Wave Relay Routers deployed across JHU Campus • Urban City Environment • Internet Access, Ad hoc Access Points, Voice over IP • Mobility testing from automobiles • Over 100 JHU students simultaneously use network each day for Internet Access • System tested at Holcim Industrial Plant (Chicago, IL) • Complex propagation environment • Enabled real-time industrial process control • Currently Deployed Custom Applications • Military Distributed Battlefield Mapping • GPS based interactive map • Eventual reliability • Locality Specific Messaging System • GPS based messaging system • Messages targeted to any user at a specific location

  46. Pulse Protocol [Infocom’04, Milcom’04, WONS’05] Scalable ad hoc routing protocol Active path tracking Based on Tree Routing strategy Medium Time Metric [MONET,WONS’04] High Throughput Path Selection Increased Path Elasticity Efficient Multi-rate Operation Leader Election Algorithm Handles merge, partition, failure Embedded Linux Distribution Less then 8 MB storage requirement Linux Kernel Module 2.4 and 2.6 compatibility Operates at layer 2 Distributed virtual switch architecture provides seamless bridging Embedded Single Board Computer Intel IXP425 Network Processor On-chip Cryptographic Accelerator 64 Mb Ram onboard 4 mini-PCI interfaces Dual 10/100 Ethernet Compact flash interface Serial port / JTAG / GPIO Hardware Watchdog Power over Ethernet +9V to +48V DC Input Atheros 802.11g/b Wireless Card 400 mW (26 dBm) output power 16 MB Intel Strata Flash Stores OS & Wave Relay software Garmin GPS 16 receiver Li-Ion Battery Pack ~20 hours continuous runtime Industrial NEMA 67 Enclosure 4 N-type antenna mounts 2 Ethernet Ports (6) protection against dust (7) water submersible Wave Relay Device Software Hardware

  47. Wireless Shuttle Bus Project http://wireless.cs.jhu.edu/mobile/

  48. Conclusion • The Pulse Protocol provides scalable routing under high levels of mobility • The Medium Time Metric selects high throughput routes and minimizes consumption of the shared wireless medium • The Wave Relay test-bed demonstrates the effectiveness of the Pulse + MTM protocols in a real-world urban environment

More Related