Idit Keidar, Technion Intel Academic Seminars, February Octopus A Fault-Tolerant and Efficient Ad-hoc Routing Protocol Idit Keidar, Technion Joint work with Roie Melamed
Idit Keidar, TechnionIntel Academic Seminars, February Ad-Hoc Networks A collection of mobile wireless nodes No pre-existing infrastructure Peer-to-peer routing: nodes relay each other's packets toward their ultimate destinations
Idit Keidar, TechnionIntel Academic Seminars, February Applications of Ad-Hoc Networks Military: tactical communications Rescue missions: without adequate wireless coverage Commercial use: sales presentations Local Area Networks (LANs): in limited- coverage areas
Idit Keidar, TechnionIntel Academic Seminars, February Challenges in Ad-Hoc Networks Lack of Infrastructure Limited wireless transmission range Rapid movement constantly changing topology Battery constrains Intermittent node disconnections
Idit Keidar, TechnionIntel Academic Seminars, February Multi-Hop Routing A B C D
Idit Keidar, TechnionIntel Academic Seminars, February Position-Based Ad-Hoc Routing Each node knows its location e.g., using GPS To send a packet– source discovers target location packets forwarded to this location Knowing location can eliminate flooding, improve scalability
Idit Keidar, TechnionIntel Academic Seminars, February Location Severs Location servers for node n: nodes storing n’s location need to be updated whenever n moves To lookup t’s location– discover a location server of t All-for-some: each node has some location servers no flooding for update or lookup each node acts as location server for some nodes e.g., Grid Location Service (GLS) [Li et al.]
Idit Keidar, TechnionIntel Academic Seminars, February Goals and Tradeoffs Low location update overhead want to send few update packets do not want to send many far away (many hops) Fault-tolerance (overcome disconnections) need many location servers need information to be fresh (frequently updated) Challenge: have many fresh location servers without inducing high load
Idit Keidar, TechnionIntel Academic Seminars, February Observation In most protocols, each location update packet contains the location of a single node, and updates a single location server The key to a better fault-tolerance/overhead tradeoff is aggregation Challenge: locate location servers as to allow efficient aggregation and cheap location discovery
Idit Keidar, Technion Intel Academic Seminars, February Octopus
Idit Keidar, TechnionIntel Academic Seminars, February Octopus in a Nutshell Space divided into horizontal and vertical strips Nodes in same strip store each other’s locations Location updates aggregated in each strip Grid can change over time ( unlike GLS)
Idit Keidar, TechnionIntel Academic Seminars, February Octopus: Key Features Fault tolerant many fresh location servers Efficient aggregation reduces location update overhead Simple Supports dynamically changing area Improved forwarding
Idit Keidar, TechnionIntel Academic Seminars, February Three Sub-Protocols Location update maintains each node’s location at its designated location servers as well as at its radio range neighbors Location discovery discovers a target location (at an appropriate location server) Forwarding forward data packets to this location
Idit Keidar, TechnionIntel Academic Seminars, February Location Update I – Neighbor List Periodically, each node broadcasts HELLO message with its identity and location to radio-range neighbors
Idit Keidar, TechnionIntel Academic Seminars, February Location Update II – End Nodes A north/south end node has no neighbors in direction north/south that reside in its vertical strip Same for east/west horizontal
Idit Keidar, TechnionIntel Academic Seminars, February Location Update II – Strip Update A-C A-FA-KA-M A-P A-WA-S A-I #messages per node- constant # bits- sqrt
Idit Keidar, TechnionIntel Academic Seminars, February Location Discovery Take I
Idit Keidar, TechnionIntel Academic Seminars, February Location Discovery Take II Forwarding Hole Quadratic reduction of failure rate
Idit Keidar, TechnionIntel Academic Seminars, February Location Discovery Alternatives Two opposite directions at a time north and south concurrently, if fails, west and east concurrently One direction at a time try short direction first (use estimate of grid area) Tradeoff between overhead and latency
Idit Keidar, TechnionIntel Academic Seminars, February Forwarding: Geographic Greedy Forward packet to neighbor that is closest to target
Idit Keidar, TechnionIntel Academic Seminars, February Forwarding: Local Maxima Geographic forwarding fails Octopus uses redundant information about strip nodes Forward to strip node closest to target
Idit Keidar, Technion Intel Academic Seminars, February Evaluation
Idit Keidar, TechnionIntel Academic Seminars, February NS-2 Simulations Scalability increasing the network size with fixed density increasing the node density Fault-tolerance Data forwarding Comparison with GLS
Idit Keidar, TechnionIntel Academic Seminars, February Reliability: Query Success Rate
Idit Keidar, TechnionIntel Academic Seminars, February Message Complexity Scalable!
Idit Keidar, TechnionIntel Academic Seminars, February Byte Complexity
Idit Keidar, TechnionIntel Academic Seminars, February Node Density & Reliability
Idit Keidar, TechnionIntel Academic Seminars, February Node Density & Message Complexity Scalable!
Idit Keidar, TechnionIntel Academic Seminars, February Node Density & Byte Complexity Scalable!
Idit Keidar, TechnionIntel Academic Seminars, February Node Disconnections – Simulation Setting Two types of nodes: stable and unstable A stable node is always up An unstable node alternates between being connected and disconnected up 2/3 of the time down 1/3 of the time For a percentage p of unstable nodes we run nodes
Idit Keidar, TechnionIntel Academic Seminars, February Fault-Tolerance
Idit Keidar, TechnionIntel Academic Seminars, February Data Forwarding Reliability
Idit Keidar, TechnionIntel Academic Seminars, February Comparison with GLS Leading solution to date Compare: Reliability Message and byte complexity Fault-tolerance Data forwarding reliability and overhead
Idit Keidar, TechnionIntel Academic Seminars, February Reliability
Idit Keidar, TechnionIntel Academic Seminars, February Message Complexity
Idit Keidar, TechnionIntel Academic Seminars, February Byte Complexity
Idit Keidar, TechnionIntel Academic Seminars, February Fault-Tolerance
Idit Keidar, TechnionIntel Academic Seminars, February Data Overhead
Idit Keidar, TechnionIntel Academic Seminars, February Octopus: Conclusions Highly fault tolerant reliable when all nodes intermittently disconnect many fresh location servers Efficient aggregates: sends much fewer messages saves MACs, hence sends fewer bytes Simple Supports dynamically changing area Forwarding uses location information