260 likes | 369 Views
Author Ljubica Blazevic Author Jean-Yves LeBoudec Author Silvia Giordan Presenter Douglas Pepelko. Location Based Routing for MANETS. Problem: Holes Holes are areas where the are few/no nodes Old solution was to route in geographic direction until hole was hit
E N D
Author Ljubica Blazevic Author Jean-Yves LeBoudec Author Silvia Giordan Presenter Douglas Pepelko Location Based Routing for MANETS
Problem: Holes • Holes are areas where the are few/no nodes • Old solution was to route in geographic direction until hole was hit • Then route around hole in “perimeter mode” • Not the best way. Ex Going from to Egypt from France would cause us to travel down to Italy, then Greece, and around.
Problem: Mobility • When nodes move finding them with old routing techniques no longer work • A nodes location may be very transient. Suppose the node is in a car moving 60mph
Solution: Terminode Routing • Link state mechanism • Nodes maintain some routing information • Anchor areas • Allows nodes to route around holes • Restricted Local Flooding (RLF) • Allows us to “search” a limited area using flooding • If a node has moved, it can still be found
Overview of Routing Methods • Terminode Remote Routing (TRR) • Routing when destination is more than 2 hops • Link state routing • Similar to geographic routing except that anchors are used • Terminode Local Routing (TLR) • Broadcast using RLF or use TLR table when destination is within two hops
Anchors • Anchors are geographic areas • Similar to a zip code • Not a specific node • Used to select a route to a destination • I would have called them “zip codes” or “anchor areas” or something else to avoid conflicting with the common use of the term “anchor”
Friends • Friend Assisted Path Discovery (FAPD) • Helps nodes find routes (Knows how to get places) • Certain percentage of node preconfigured to be FAPD responders • Can find other friends
Other Terms • EUI - End-system Unique Identifer • An address (like an IP address) • LDA - Location Dependent Address • Geograpical Coordinates (x, y) • RLF - Restricted Local Flooding • Described later • TLR,TRR - Terminode Remote/Local Routing • Routing techniques used throughout
Restricted Local Flooding • RLF is used when local communication is needed but the exact location is not known • Packets are sent with a source location, a max distance and the RLF flag set • Any node that receives a packet at a distance greater than the max distance from source location and the RLF flag set does NOT repeat the message • Otherwise the message is repeated just as in regular flooding
Friends Finding Friends • All preconfigured FAPD nodes use RLF • Broadcast a get_friends_request • Four (4) packets are sent in opposite directions • Presumably North South East and West • Sent with no destination address EUI • Sent with a location LDA • Other friends respond with their lists • Lists are merged and sent to other friends on request
Nodes in the 'hood' • A node must know its neighbourhood • Broadcasts a HELLO with LDA and EUI • Listens for other HELLOs and builds a TRL table' • Table has links up to two (2) hops away • Creates a Gabriel Graph • Spatial proximity • A node may posses density maps to assist in Geographic Map-based Path Discovery (GMPD)
There and Back Again • A node wants to send a packet • Destination is local: Easy, just turn on “Use TLR” bit and send via the TLR table • Destination is remote: Several steps • Obtain an LDA (a geographic location) • Send using TRR without Anchors. • Send to closest neighbour in the TLR table • Obtain feedback about route taken • Attempt to optimize route using anchors
Forwarding a Packet • A node receives a packet • For me? Keep it • Local? Use the TRL table • TRR termination • If location is in transmission range but packet is not in TRL table • Attempts to find node using RLF (flooding) • Sends six (6) packets in different direction • Max distance is set to twice the transmission range
What about Anchor Paths • Geographic routing produces poor path in some cases • If a node sends a packet and then determines that the path taken was non-optimal it can attempt to compute a better “anchor path” • Node sends an anchored path request using RLF in four (4) directions.
An FAPD Receives an Anchor Request • A FAPD gets a request • If it is closer to the destination or has an anchor path to the destination it appends its geographic location (and the rest of the path if available) to the list • Otherwise it sends the packet on to the destination • TABU index • How far “backwards” the packet can move from the destination • If no friends are closer to destination, head backwards, but increment tabu • Check that tabu never exceeds max_tabu of two (2)
Destination Receives Anchor Path Search Packet • Destination gets a long list of anchors “accumulated” in packet • Destination simplifies path • Reduce number of anchors • Set timers on path freshness
Source uses GMPD to find path • If a node knows network density it can use this determine a good path. • High node density areas are “towns” • Towns are connected by “highways” • Much like a second layer of connectivity • Use an atlas to get you across the country • Use a city map to get you to a street
Performance • The authors claim that this method works better than all other methods • As good as GPSR when location accuracy is high, but better when location accuracy is low • Better than GPSR when there are holes • Same as AODV and LAR1 with small nets but better than both in a large network or when dealing with mobility
What did they test? • Packet deliver fraction • Ratio of packets delivered / total packets generated by CBR (constant bit rate) source • Average end-to-end delay • How long did it take to get the data there • Includes delay in route discovery and waiting for LDA • Normalized routing load • Number of control packets per data packet delivered at destination
Nodes movin' and groovin' • To simulate a real network nodes must move • Random nodes • Node chooses a random waypoint then moves at a random speed (1-20m/sec) • Node pauses at destination for a random time • Restricted Random nodes • Some nodes are town nodes and are more likely to “stay in town” • Some nodes are commuters and move between towns. • Finally some are stationary nodes. They don't move.
Small unobstructed network It certainly seems better tan LAR1 or AODV
RLF (restricted local flooding) helps quite a bit If a node has moved, it usually has not gone too far. RLF can find it.
Conclusions • Terminode is a mish-mash of several things • Should have been separated into different papers • RLF Assumes nodes are going to move after being located. Assumes that they don't move too far. Broadcasts in the area of the last known location • Could be improved by calibrating the distance based on the measured mobility of nodes • Could use the LDA dest as center of Broadcast area • TRR using towns/highways (GMPD) seems like a good idea • Results show it is one of the better algorithms • Requires density maps though
References [1] [BLAZEVIC05] http://people.cs.vt.edu/~irchen/6204/pdf/BLA05-sensor-routing.pdf [2] [SAVCHENKO] http://cgm.cs.mcgill.ca/~godfried/teaching/projects.pr.98/sergei/project.html [3] [MAURO] http://www.i-cherubini.it/mauro/blog/2005/10/17/the-gabriel-graph/