Link capacity estimation in QOLSR Ignacy Gawdzki <i@lri.fr> - - PowerPoint PPT Presentation

Link capacity estimation in QOLSR

Ignacy Gawędzki <i@lri.fr> Hakim Badis <badis@lri.fr> Khaldoun Al Agha <alagha@lri.fr>

LRI laboratory

July 29th, 2005

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion

1

Introduction What is QOLSR?

2

Link capacity in radio networks Correlated capacities Simple example

3

Clique constraints Conflict graphs Cliques of conflict graphs

4

Clique constraints with OLSR Additional topology information Routing with capacity

5

Conclusion

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion What is QOLSR?

QOLSR = OLSR with QoS routing

Routing metrics: Hop count Delay Capacity (available bandwidth) Packet loss probability Security Constraint: independence with respect to NIC driver More info: http://qolsr.lri.fr

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Radio link vs. wired link capacity

One antenna per interface = ⇒ sending XOR receiving Omnidirectional antennas = ⇒ implicit broadcasting Capacities of links are correlated and not independent.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example without flow control

a b c d e f

c wants to transmit to d

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example without flow control

a b

DATA c DATA d

e f

c transmits the data frame to d

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example without flow control

a b

DATA c DATA d

e f

b and d cannot emit anymore

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example without flow control

a b c

ACK

d

ACK

e f

d acknowledges the frame

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c d e f

c wants to transmit to d

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c d

RTS RTS

e f

c sends the RTS frame

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c d

RTS RTS

e f

b and d cannot emit anymore

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c

CTS

d

CTS

e f

d sends the CTS frame

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c

CTS

d

CTS

e f

now b, d and e cannot emit anymore

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b

DATA c DATA d

e f

c transmits the data frame to d

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Correlated capacities Simple example

Transmission example with flow control

a b c

ACK

d

ACK

e f

d acknowledges the frame

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Conflict graphs Cliques of conflict graphs

Some links do share a common resource

Conflict graph: vertices are links edges are conflict relations

a b c d e ab de cd bc

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Conflict graphs Cliques of conflict graphs

Necessary condition

Links in the same clique of conflict graph share a common capacity resource.

a b c d e ab de cd bc

Necessary condition: sum of loads on links of one clique must be lower than the common capacity.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Conflict graphs Cliques of conflict graphs

Necessary but not sufficient

Perfect graph: chromatic number = size of greatest clique Conflict graph perfect = ⇒ Clique constraints sufficient Unit Disc Graph (UDG): edge {a, b} exists iff d(a, b) ≤ 1 Topology graph is UDG = ⇒ β × Clique constraints sufficient

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to do that?

General idea: Nodes advertise transmitted load in HELLOs Each node performs capacity estimation on incident links Nodes calculate routes using capacities But: Nodes need total (all nodes and all links) vs. partial topology Two-hops neighborhood is not enough

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

(N+1)-hops neighborhood

(N+1)-hops neighborhood is necessary to compute the conflict graph with N-hop correlations.

a c e d b b d e c a

We need to retransmit HELLO messages.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion Additional topology information Routing with capacity

How to calculate routes?

A flow is possible if it satisfies the clique constraints on its path. Simple heuristics: 1-hop-long paths are trivial 2-hops-long paths No more than one link used simultaneously. 3-hops-long or more N=2: no more than 1/3 of the links used simultaneously.

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion

What’s next?

Capacity estimation already implemented (to be released soon). Broadcast capacity implementation is planned. More routing metrics?

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR

Introduction Link capacity in radio networks Clique constraints Clique constraints with OLSR Conclusion

Any questions?

?!

Ignacy Gawędzki, Hakim Badis, Khaldoun Al Agha Link capacity estimation in QOLSR