Dominating Set Based Support for Distributed Services in Mobile Ad Hoc Networks
Károly Farkas, Florian Maurer, Lukas Ruf,
Bernhard Plattner
5
thApril, 2006, Vancouver
Introduction
Popularity of mobile devices and ubiquitous environments
Emergence of self-organized and ad hoc communication paradigms
- Popularity of mobile devices and ubiquitous environments
- Emergence of self-organized and ad hoc communication paradigms
Support for Distributed Services
in Mobile Ad Hoc Networks
Contribution Overview
Zone-Based Game Architecture
By the aid of graph theory
● Zone Server selection via Dominating Set (DS) computation
● Maintenance of the zones and Zone Servers in case of topology changes
Support for Distributed Services in Mobile Ad Hoc Networks
But How?
Zone-Based Game Architecture
By the aid of graph theory
● Zone Server selection via Dominating Set (DS) computation
● Maintenance of the zones and Zone Servers in case of topology changes
Outline
Introduction
Zone-Based Game Architecture
Priority Based Selection (PBS) Algorithm
Performance of PBS
Conclusions and Future Work
Zone-Based Game Architecture
Zone Red Zone Green
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Challenges
Create the zones in the network
Select the Zone Server nodes
Maintain the zones and Zone Servers in case of dynamic topology changes
- Create the zones in the network
- Select the Zone Server nodes
- Maintain the zones and Zone Servers in case of dynamic topology changes
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Our Approach
Compute and maintain a Dominating Set (DS) of the network graph ⇒ Priority Based
Selection (PBS) algorithm
Dominator Dominatee S={3,5}
Dominating Set: For a graph G=(V, E), a DS is a subset S ⊆ V, such that for all nodes v ∈ V, either v ∈ S or a neighbor u of v is in S
1
2 4 6 7
3 5
3 5
Zone Green
Zone Red
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Assumptions for Zone Server Selection
Ad hoc network
● Not fragmented
● Knowledge about the neighbor nodes (routing protocol)
Nodes
● Unique ID
● Node weight
● Co-operative behavior
Ad hoc network
● Not fragmented
● Knowledge about the neighbor nodes (routing protocol)
Nodes
● Unique ID
● Node weight
● Co-operative behavior
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Requirements for Zone Server Selection
Dominating Set
● Minimum 2 nodes
● Node weighted
● Good Minimum Dominating Set (MDS) approximation
Algorithm
● Completely distributed
● Update mechanism for mobility maintenance
● Low time and message complexity
Dominating Set
● Minimum 2 nodes
● Node weighted
● Good Minimum Dominating Set (MDS) approximation
Algorithm
● Completely distributed
● Update mechanism for mobility maintenance
● Low time and message complexity
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Priority Based Selection (PBS) Algorithm
A node can be in one of the following states:
● DOMINATOR (Server)
● DOMINATEE (Client)
● INT_CANDIDATE (Internal Candidate)
● EXT_CANDIDATE (External Candidate)
A node can be in one of the following states:
● DOMINATOR (Server)
● DOMINATEE (Client)
● INT_CANDIDATE (Internal Candidate)
● EXT_CANDIDATE (External Candidate)
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Main Idea Behind the PBS Algorithm
status = INT_CANDIDATE or EXT_CANDIDATE;
while v has INT_CANDIDATE neighbors within distance 2 do - send neighborlist;
- receive neighborlist;
- change to DOMINATEE, if neighbor is DOMINATOR;
- change to DOMINATOR, if v has highest priority within distance 2 among the nodes with INT_CANDIDATE status;
od
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Priority Computation
A node has higher priority than another if
● the node has higher node weight;
● if tie: the node has higher span value (INT_CANDIDATE neighbors);
● if tie: the node has more neighbors with DOMINATOR status;
● if tie: the node has a lower ID.
A node has higher priority than another if
● the node has higher node weight;
● if tie: the node has higher span value (INT_CANDIDATE neighbors);
● if tie: the node has more neighbors with DOMINATOR status;
● if tie: the node has a lower ID.
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Mobility Maintenance in the PBS Algorithm
msgsent
roundfinished
finished
s1
s2
s3
Receive neighborlists
Status determined Send neighborlist
Send neighborlist
Determine status
Change happened, Send neighborlist
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Analysis of the PBS Algorithm Building an initial DS:
O(k(Δ+1)) Message size per round:
Lower bound: Ω(3) Upper bound: O(n) Rounds:
Optional Connected DS
O(n) Messages per round:
O(log Δ) MDS approximation:
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
State-Of-The-Art DS Computation Algorithms
*The values in the highlighted area are the same as in case of PBS
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
N/A N/A
PBS O(n) O(n) messages, size O(k(Δ+1)) No O(logΔ)
Performance of the PBS Algorithm
Simulation settings*
● Network Simulator NS-2, Version 2.28
● 3 scenarios (School Yard, Train, Test)
● 15 and 35 nodes forming the network
● Adapted mobility models
● Typical real-time multiplayer game traffic between player nodes
● Background traffic
● Game sessions of 900 seconds
Simulation settings*
● Network Simulator NS-2, Version 2.28
● 3 scenarios (School Yard, Train, Test)
● 15 and 35 nodes forming the network
● Adapted mobility models
● Typical real-time multiplayer game traffic between player nodes
● Background traffic
● Game sessions of 900 seconds
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Simulation Results in Detail (1/2)
Scenario Avg.
[kbps]
S. dev.
[kbps]
School Yard w/15 nodes 0.040 0.012 0.090 1.322 0.042 0.182 Train w/15 nodes
0.008 0.003 0.015 0.180 0.011 Test w/15 nodes
School Yard w/35 nodes Train w/35 nodes
Test w/35 nodes 0.053
Scenario Avg.
[ms]
S. dev.
[ms]
School Yard w/15 nodes 159 111 185 283 153 510 Train w/15 nodes
0.172 0.178 0.188 0.251 0.193 Test w/15 nodes
School Yard w/35 nodes Train w/35 nodes
Test w/35 nodes 0.311
Bandwidth (total = 2 Mbps): Determination delay:
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Simulation Results in Detail (2/2)
Scenario Avg. S. dev.
School Yard w/15 nodes 0.6 1.6 0.8 3.6 4.4 Train w/15 nodes
0.5 1.3 0.6 0.8 1.2 Test w/15 nodes
School Yard w/35 nodes Train w/35 nodes
Number of DS changes:
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Bandwidth
● 0.02-0.6 % of 2 Mbit/s
● Will not cause any problems
Determination delay
● ~100-500 ms depending on the number of 1-hop neighbors
Number of DS changes
● Relatively small in scenarios with few nodes
● Increases with the increasing number of nodes and higher mobility level
Summary of Simulation Results
Bandwidth
● 0.02-0.6 % of 2 Mbit/s
● Will not cause any problems
Number of DS changes
● Relatively small in scenarios with few nodes
● Increases with the increasing number of nodes and higher mobility level
Determination delay
● ~100-500 ms depending on the number of 1-hop neighbors
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
*More simulation results in the Proceedings
Conclusions
PBS computes a close to minimal node weighted DS with similar performance to the state-of-art DS computation algorithms
PBS offers continuous maintenance of the computed DS
The computed DS can be used for creating a zone-based distributed service architecture
- PBS computes a close to minimal node weighted DS with similar performance to the state-of-art DS computation algorithms - PBS offers continuous maintenance of the
computed DS
- The computed DS can be used for creating a zone-based distributed service architecture
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions
Future Work
PBS enhancements
● Include a more sophisticated weight computation method
● Include mobility prediction
Scenarios
● Test PBS in more realistic mobility scenarios
Testbed measurements
● Investigate the performance of PBS in real environment
PBS enhancements
● Include a more sophisticated weight computation method
● Include mobility prediction
Scenarios
● Test PBS in more realistic mobility scenarios
For more information visit: www.siramon.org
Testbed measurements
● Investigate the performance of PBS in real environment
Introduction Zone-based Arch. PBS Alg. Performance of PBS Conclusions