Interference-Aware QoS OLSRfor Mobile Ad-hoc Network Routing
SAWN 2005, May 24
P. Minet & D-Q. Nguyen
SAWN 2005 May 24, 2005 2
Outline
1. Introduction2. QoS framework for ad-hoc networks3. Interference-aware QoS OLSR4. Performance evaluation5. Conclusion
SAWN 2005 May 24, 2005 3
1. Introduction
Transmissions and receptions in ad-hoc networks are subject to radio interference.=> Bandwidth resource is affected.
SAWN 2005 May 24, 2005 4
1. Introduction
Ad-hoc networks have scarce resources.=> QoS management in ad-hoc networks is more difficult than
in wired networks. Admission control is needed.
SAWN 2005 May 24, 2005 5
2. QoS framework
Class 2
Bandwidth Controland Reservation
Application(Bandwidth)
Path Computation
MAC layer metrics
QoS Advertisements
SAWN 2005 May 24, 2005 6
2. QoS frameworkApplication(Bandwidth)
Routing on theReserved Path
Class 2 Marking
SAWN 2005 May 24, 2005 7
3. Interference-aware QoS OLSR
QoS signalisation
Measure local available bandwidth (LAB) At each node. Based on values obtained from MAC layer.
LAB dissemination Any node broadcasts in Hello message : its LAB and
the LAB of each neighbor. Any MPR (multipoint relay) broadcasts in TC message
the LAB of each MPR selector.
MPR selection based on LAB Any node selects its MPRs so that it can reach any
two-hop neighbor by a largest path; i.e. path with maximum bandwidth.
SAWN 2005 May 24, 2005 8
3. Interference-aware QoS OLSR
i
m n
x y z
5 1
N1(i) = {m, n}N2(i) = {x, y, z}
OLSR native MPR selection MPR selection with bandwidth
MPR selection example :
SAWN 2005 May 24, 2005 9
3. Interference-aware QoS OLSR
Interference-aware admission control
Accept a new flow iff: QoS required by this flow can be met. QoS of already accepted flows must not be altered.
Perform in 2 steps: Step 1: Selection of an acceptable path
Any node on the path must provide the amount of bandwidth required by the new flow.Can be checked locally by the source node.
Step 2: Path feasibility with interferencesAny node in the interference zone of a node on the path must have enough bandwidth to support this new flow.A message is sent from source to destination.
SAWN 2005 May 24, 2005 10
3. Interference-aware QoS OLSR
Admission control example :
S
DS
D
This path forthe Yellow flow
is not acceptable!
SAWN 2005 May 24, 2005 11
3. Interference-aware QoS OLSR
Interference-aware QoS routing algorithm
The shortest routes tend to minimize network resources required for transmission of each packet from its source to destination in a wireless multihop environment. => Minimize the number of hops as first criterion.
Some flows require bandwidth as QoS parameter.=> Consider local available bandwidth at each node as second criterion.
Route computation is called upon any topology change.=> Complexity must be similar to Dijkstra algorithm.
Network resources is scarce.=> Algorithm is based on partial knowledge of topology.
SAWN 2005 May 24, 2005 12
3. Interference-aware QoS OLSR
Interference-aware QoS routing algorithm
Algorithm 1: Default algorithm used to compute routing table. Unconstrained, widest-shortest path. Called upon any change in the one-hop neighborhood,
two-hop neighborhood or topology table.
Algorithm 2: Constrained by a bandwidth request. Used to compute a route offering the requested
bandwidth from a source to a destination if the default route, from algorithm 1, cannot provide that bandwidth.
Called by the admission control for a new flow with bandwidth demand.
SAWN 2005 May 24, 2005 13
3. Interference-aware QoS OLSR
Admission control example with routing algorithm 2 :
S
DS
D
Flow Yellowacceptable!
SAWN 2005 May 24, 2005 14
4. Performance evaluation
Simulation plan 250 static nodes uniformly located on a 2500x2500m2 square. 7 CBR flows, each requires 175Kbps at application level. MAC 802.11b, no RTS/CTS. Native OLSR and Interference-aware QoS OLSR routing.
SAWN 2005 May 24, 2005 15
4. Performance evaluation
Results obtained with native OLSR
SAWN 2005 May 24, 2005 16
4. Performance evaluation
Results obtained with Inteference-aware QoS OLSR routing
SAWN 2005 May 24, 2005 17
4. Performance evaluation
Loss rate comparison
FlowNative OLSR (with 4 flows)
(%)
Native OLSR (with 7 flows)
(%)
Interference-Aware (with 7
flows) (%)
1 23.56 41.56 0.982 7.91 50.44 1.343 1.65 38.58 2.054 28.81 66.99 1.645 - 66.06 3.466 - 39.07 1.957 - 68.86 1.55
Average 15.48 53.08 1.85
SAWN 2005 May 24, 2005 18
5. Conclusion
Interference-aware routing can accept more flows into the network.
It offers better stability to the accepted flows, better bandwidth guarantee to the applications than native OLSR.
If it cannot find a route meeting the bandwidth requested,then such a route does not exist.
Main drawback: more MPRs selected => more control overhead, broadcasting using MPR becomes less efficient.
Perspective: Reduce control overhead, improve broadcasting.