MIT CSAIL
Geographic Routing without Planarization Ben Leong, Barbara Liskov & Robert Morris require planarization
Tree Routing (GDSTR) DOES NOT uses spanning tree, not planar graph low maintenance cost better routing performance than existing algorithms – – – –
Greedy Distributed Spanning New geographic routing algorithm • Overview Background Problem Approach Simulation Results Conclusion • • • • • x-y coordinates
x-y
Geographic Routing can use virtual coordinates (Rao et al. 2003) – Wireless nodes have Nodes know coordinates of immediate neighbors Packet destinations specified with coordinates In general, forward packets greedily • • • • Geographic Routing Geographic Routing Source Destination
Geographic Routing Source Destination
Greedy Forwarding Source Destination
Greedy Forwarding Source Destination
Greedy Forwarding Source Destination
Greedy Forwarding Source Destination Whoops. Dead end!
Geographic Routing:
Dealing with Dead Ends Source Destination
Face Routing Source Destination
Face Routing Source Destination
Face Routing Source Destination
Back to Greedy Forwarding Source Destination
Back to Greedy Forwarding Source Destination
Back to Greedy Forwarding Source t work ’
Planarization is Costly! GG and RNG don CLDP (Kim et al., 2005) – – Planarization is hard for real networks Planarization is complicated & costly! • • ” summarize “
Tree Routing (GDSTR) convex hulls tells us what points are possibly reachable reduces the subtree that must be traversed (smaller search problem) – – the area covered by a subtree Greedy Distributed Spanning Route on a spanning tree Use convex hulls to • • Hull Tree Hull Tree Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
GDSTR Example Source Destination
Revert to Greedy Forwarding Source Destination
Revert to Greedy Forwarding Source Destination
Revert to Greedy Forwarding Source Issues multiple hull trees conflict Hulls – – Choosing forwarding direction Undeliverable packets • • Destination
Using Multiple Trees Source Destination direction.
Using Multiple Trees Source ” bad With one tree, may be forced to route in “ Destination a void ” approximate “
Using Multiple Trees Source Two extremal-rooted trees are usually sufficient to Destination
Using Multiple Trees Source Pick tree with root closest to the destination Summary: Routing choose tree record start node traverse subtree – – – Try greedy forwarding Dead end: If possible, revert to greedy forwarding Back to start node: packet undeliverable • • • • . t s to
s
Theorem in connected graph t , GDSTR guarantees
Given a pair of nodes G and packet delivery from messages
keepalive
Building Hull Trees minimal and maximal x-coordinates minimal hop count from root – – Choose roots: Convex hull info in Want compact trees Aggregate convex hulls from leaves to root Conflict hull info percolates from root to leaves • • • • • )
Simulation Results low/high density low/high obstacle density • • average node degree Path Stretch Hop Stretch range of network densities ( larger networks up to 5,000 nodes – – – – Measured 2 routing metrics: Topologies • • Simulation Results GPSR (Karp, 2001), GOAFR+ (Kuhn, 2003) and GPVFR (Leong et al., 2005) storage bandwidth – – – – – Compare with Measured costs and compared with CLDP: • under CLDP planarization (Kim et al., 2005) • Hop Stretch Hop Stretch O(log n) s scan] ’
Costs convex hull computation: operations [Graham – Computation: Storage: < 1 kb Bandwidth • • • Message Sizes Messages for Startup Messages for Stabilization up to 20% better Summary – Maintenance cost one order of magnitude less than CLDP (face routing) Better routing performance (stretch) • • Large Voids Small Voids Destination Source
Explaining Performance Destination Source
Explaining Performance Destination Source
Explaining Performance Destination Source
Explaining Performance Destination Source
Explaining Performance Destination Source
Explaining Performance Destination Source
Explaining Performance Extra overhead Summary
• Sparse networks – GDSTR chooses correct forwarding direction more often than face routing • Moderately dense networks – Faces are small, forwarding direction is inconsequential – Trees do not “approximate” small voids well • Ultra-dense networks – Greedy forwarding works all the time! Conclusion information allows GDSTR ” Global to choose good forwarding direction more often GDSTR achieves improved routing stretch at lower maintenance cost than CLDP Cheaper to maintain two hull trees than a planar graph “ • • • Future Work convex hulls generalizable to higher dimensions – Evaluate GDSTR in a practical and mobile setting Geographic routing in higher dimensions • • MIT CSAIL
Geographic Routing without Planarization Ben Leong, Barbara Liskov & Robert Morris Reducing Convex Hulls Reducing Convex Hulls Reducing Convex Hulls allow us to avoid
Conflict Hulls : parent nodes tell child Undeliverable packets will be forwarded to the root. forwarding to the root Conflict hulls Key idea about other nodes with intersecting hulls • • • Example: Conflict Hull Example: Conflict Hull Example: Conflict Hull Example: Conflict Hull Example: Conflict Hull …
Example: Conflict Hull Forward to parent Example: Conflict Hull Packet undeliverable! Minimal-x Tree Maximal-x Tree
Example GDSTR Hull Trees Two Trees
Topologies
Comparing Routing (CLDP) Planar Graph