Landmark Routing for Large Ad Hoc Wireless Networks Globecom 2000 San Francisco, Nov 30, 2000
description
Transcript of Landmark Routing for Large Ad Hoc Wireless Networks Globecom 2000 San Francisco, Nov 30, 2000
Landmark Routing for Large Ad Hoc Wireless Networks
Globecom 2000San Francisco, Nov 30, 2000
Mario Gerla, Xiaoyan Hong and Gary Pei
Computer Science Department
University of California, Los Angeles
http://www.cs.ucla.edu/NRL/wireless/
Ad Hoc vs Cellular Wireless Nets
Multihop (Ad Hoc)
Single Hop (Cellular)
Base BaseBaseBase
Scalability in ad hoc wireless routing
• Scalability to network size– Potentially, thousands of nodes (e.g., battlefield, sensor networks)
• Scalability to mobility– mobility critical in battlefield and vehicular applications
Do Existing Routing Protocols Scale?
• Proactive routing:– Distance Vector based: DBF, DSDV, WIRP
– Link State
Main limitations: routing table O/H; control traffic O/H
• On-demand, reactive routing:– AODV, TORA, DSR, ABR etc
Main limitations: search-flood O/H with high mobility and many short lived flows
Distance Vector
0
5
1
2
4
3
Destination Next Hop Distance
0 2 31 2 2… … …
Routing table at node 5 :
Tables grow linearly with # nodes
Control O/H grows with mobility and size
Link State Routing
• At node 5, based on the link state packet, topology table is constructed:
• Dijkstra’s Algorithm can then be used for the shortest path
0
5
1
2
4
3
{1}
{0,2,3}
{1,4}
{2,4}
{2,3,5}
{1,4,5}
0 1 2 3 4 50 1 1 0 0 0 01 1 1 1 1 0 02 0 1 1 0 1 13 0 1 0 1 1 0
4 0 0 1 1 1 15 0 0 1 0 1 1
0
5
1
2
4
3
query(0)
query(0)
query(0)
query(0)
query(0)
query(0)
query(0)
reply(0)
reply(0)
reply(0)
On-demand Routing
Advantages:– on-demand request & reply
eliminates periodic update O/H (channel O/H)
– routing table size is reduced (it includes only routes in use) (storage O/H)
Limitations:– not scalable with traffic load– mobility may trigger frequent
flood-searches
Hierarchical Routing
• Traditional solution in large scale networks (eg, Internet):
hierarchical routing
• Unfortunately, hierarchical routing implementation problematic in ad hoc nets
• In a mobile ad hoc network the hierarchical addresses must be continuously changed to reflect movements
• Some ad hoc routing schemes recently proposed use an “implicit” hierarchy (eg, Fisheye, Zone routing, etc)
Wireless Hierarchical Routing(addresses change with motion)
5
1
7
6
11
4
23
10
98
Level = 0
(1,1)
(1,2)(1,3)
(1,4) Level = 1
(2,1) (2,3)Level = 2
DestID
1
6
7
(1,2)
(1,4)
(2,3)
Path
5-1
5-1-6
5-7
5-1-6-(1,2)
5-7-(1,4)
5-7-(1,4)-(2,3)
HSR table at node 5
Implicit hierarchical routing: Fisheye State Routing
11
1
2
3
4
5
67
8
9
9
1012
14 1516 17
18 19
20
21
2223
2425
26
27
28
29
30
31
3234
35
36
Hop=1
Hop=2
Hop>2
13
Fisheye Routing
• In Fisheye routing, routing table entries for a given destination are updated (ie, exchanged with the neighbors) with progressively lower frequency as distance to destination increases
• Property 1: the further away the destination, the less accurate the route
• Property 2: as a packet approaches destination, the route becomes progressively more accurate
• Major “scalability” benefit: control traffic O/H is manageable even for very large network size
• Unsolved problems: route table size still grows linearly with network size; out of date routes to remote destinations
Update O/H Reduction in FSR (optional)
0
5
1
2
4
3
0:{1}1:{0,2,3}2:{5,1,4}3:{1,4}4:{5,2,3}5:{2,4}
101122
LST HOP
0:{1}1:{0,2,3}2:{5,1,4}3:{1,4}4:{5,2,3}5:{2,4}
212012
LST HOP
0:{1}1:{0,2,3}2:{5,1,4}3:{1,4}4:{5,2,3}5:{2,4}
221101
LST HOP
Ad Hoc “Group” Hierarchical Solution: Landmark Routing
• Main assumption: nodes move in groups• Three components in LANMAR:• (1) a “local ” proactive routing algorithm that
keeps accurate routes from a source to all destinations within scope N (e.g., Fisheye alg truncated to scope N, Bellman Ford, DSDV, etc)
• (2) a Landmark selection alg for each logical group
• (3) a routing algorithm that maintains accurate routes to landmarks from all mobiles in the field
Logical SubnetLogical Subnet
• Logical subnet: group of nodes with functional affinity with each other (eg, they move together)
• Node logical address = <subnet, host>
Landmark Routing: the Concept
LandmarkLandmark
• A Landmark is elected in each subnet• Every node keeps Fisheye Link State table/routes
to neighbors up to hop distance N• Every node maintains routes to all Landmarks
Landmark Routing (cont’d)
• A packet to local destination is routed directly using Fisheye table based on MAC address
• A packet to remote destination is routed to corresponding Landmark based on logical addr
• Once the packet gets within Landmark scope, the direct route is found in Fisheye tables
• Benefits: dramatic reduction of both routing overhead and table size; scalable to large networks
LandmarkLandmark
Logical SubnetLogical Subnet
Landmark Routing: Dynamic Election
• Dynamic landmark election a must in a mobile environment and in presence of enemy attacks
• Node with largest number of group members in its scope proclaims itself Landmark for group; ties broken by lowest ID
• “Oscillation” of landmark role is eliminated by hysteresis.
• Multiple landmarks may coexist if group spans several “scopes” (they can be hierarchically organized)
Landmark Election (detail - may skip)
• Landmark election algorithm:– No landmark exists initially, only FSR progresses.
– A node proclaims itself as a landmark when it detects > T number of group members in its FSR scope.
– An election is required to select the winner in the group.
• Simple election winner algorithm– A node with the largest number of group members wins and the
lowest ID breaks a tie.
• Hysteresis election winner algorithm– The current election winner replaces the old landmark when its
number of group members is larger than the old one by an extra fraction.
– Or, the old landmark gives up the landmark role when its number of group members reduces to a value smaller than a threshold T.
Drifting nodes (detail - may skip)
• Drifters are nodes outside of the scope of their landmark
• Drifters periodically “register” with Landmark• Registration message creates reverse path from
Landmark to drifter• A packet directed to a drifter must be first
received by the Landmark and then forwarded to drifter
• Routing table entries to drifters increase routing table OH; however, the extra O/H is low if drifter fraction is low
Illustration by Example
A
B
C D
HI
JK L
O
P
LM1
LM2
LM3
LM4
Simulation Environment
• GlomoSim platform• 100 nodes• 1000x1000 square meter simulation area• 150m radio range• UDP sessions between random node pairs• CBR traffic ( one 512 byte pkt every 2.5 sec)• # of logical groups = 4• 2-level Fisheye with radius = 2 hops• IEEE 802.11 MAC layer; 2Mbps link rate• Reference Point Group Mobility model
– random waypoint model is used for both individual and group component of the mobility vector
Throughput and Delay
mobility = 6 m/s
0
200
400
600
800
1000
1200
1400
10 30 50 100 200 300 400 500
number of communication pairs
dela
y (m
sec)
AODV
LANMAR
FSR
mobility = 6 m/s
0
50
100
150
200
250
300
10 30 50 100 200 300 400 500
number of communication pairs
Thro
ughp
ut (k
bits
/sec
)
AODV
LANMAR
FSR
Routing Load with and w/o Election
Conclusions
• Accuracy of the route to Landmark nodes proves to be adequate
• LANMAR exhibits good scalability with increasing communication pairs
• LANMAR provides a dramatic reduction in routing table storage overhead with respect to FSR
• Dynamic Landmark Election introduces only a moderate increase in routing O/H (with respect to fixed Landmark)
Work in Progress (optional)
• Independent (instead of group) mobility• Very small groups (in the limit, all isolated
nodes)• “Optimal” scope of local routing• Hierarchical Landmark organization• Membership change from one group to
another • Landmarking in a heterogeneous
structure: directive antennas, UAVs etc
The End
Thank You !
www. cs.ucla.edu/NRL