Transcript Slides INTERACT-9

Santander (SPAIN) - September 22-24, 2010
OLSRp: Predicting Control Information to
Achieve Scalability in OLSR Ad Hoc Networks
Esunly Medina ф
Roc Meseguer ф
Carlos Molina λ
Dolors Royo ф
ф Dept. Arquitectura de Computadors
λ Dept. Enginyeria Informàtica i Matemàtiques
Universitat Politècnica de Catalunya
Universitat Rovira i Virgili
Barcelona, Spain
Tarragona, Spain
[email protected]
{esunlyma, meseguer, dolors}@ac.upc.edu
Outline
OLSR
• Motivation
• Potentiality
• OLSRp
• Conclusions & Future Work
Motivation
Motivation
• Ad-hoc networks:
– Need for maintaining network topology
– Control messages consume network resources
• Proactive link state routing protocols:
– Each node has a topology map
– Periodically broadcast routing information to neighbors
… but when the number of nodes is high …
… can overload the network!!!
OLSR: ControlOLSR
Traffic and Energy
OLSR is one of the
most intensive
energy-consumers
Traffic and energy do
NOT scale !!!
… can we increase scalability of routing protocols
for ad-hoc networks? …
DQ OLSR
principle
• Data per query × Queries per second →constant
– For routing protocols:
• D = Size of packets
• Q = Number of packets per second sent to the network
• We focus on Q:
– Reducing transmitted packets
– Without adding complexity to network management
• HOW?
PREDICTING MESSAGES !!!!
We propose a mechanism for
increasing scalability of ad-hoc networks
based on link state proactive routing protocols
– Called OLSRp
– Predicts duplicated topology-update messages
– Reduce messages transmitted through the network
– Saves computational processing and energy
– Independent of the OLSR configuration
– Self-adapts to network changes.
Potentiality
Experimental
OLSR Setup
• NS-2 & NS-3
• Grid topology, D = 100, 200, … 500 m
• 802.11b Wi-Fi cards, Tx rate 1Mbps
• Node mobility:
• Static, 0.1, 1, 5, 10 m/s
• Friis Propagation Model
• ICMP traffic
• OLSR control messages:
– HELLO=2s
– TC=5s
OLSR: Messages
OLSRdistribution
TC vs HELLO
Ratio of TC messages is significant for low density of nodes
Control Information
OLSR Repetition
Number of nodes does not affect repetition
Control Information
OLSR Repetition
Density of nodes slightly affects repetition
Control Information
OLSR Repetition
Repetition is mainly affected by mobility
Control Information
OLSR Repetition
Repetition still being significant for high node speeds
OLSRp
OLSRp:
Basis
OLSR
Prevent MPRs from transmitting duplicated TC
throughout the network:
– Last-value predictor placed in every node of the network
– MPRs predicts when they have a new TC to transmit
– The other network nodes predict and reuse the same TC
– 100% accuracy:
• If predicted TC ≠ new TC  MPR sends the new TC
– HELLO messages for validation
• The topology have changed and the new TC must be sent
• The MPR is inactive and we must deactivate the predictor
OLSRp:
Layers
OLSR
Upper Levels
Upper Levels
OLSR
OLSR
Input
Output
OLSR
OLSR
Input
Output
OLSRp OLSRp
Lower Levels
Input
Output
Wifi Input
Lower Levels
Wifi Output
Input
TCWifi
TCOLSR
Wifi 
if (TC[n]=TC[n-1]): TCOLSRp  TCOLSR
else: TCWifi TCOLSR
WifiTCOutput
if MPR:
OLSR  TCWifi
if MPR if(TC[n]=TC[n-1]): TCOLSRp
else: TCOLSR  TCWifi
OLSRp:
Basis
OLSR
– Each node keeps a table whose dimensions depends on the
number of nodes
– Each entry records info about a specific node:
• The node’s @IP
• The list of @IP of the MPRs (O.A.) that announce the node in
their TCs and the current state of the node (A or I). (HELLO
messages received).
• A predictor state indicator for MPR nodes (On or Off):
– On when at least one of the TC that contains information
about the MPR is active
– Off when the node is inactive in all the announcing TC
messages (new TC message will be sent)
OLSRp:
Example
OLSR
B
B
OLSRp:
Example
OLSR
B
B
NODE D TABLE
OLSRp:
Example
OLSR
X
X
B
X
X
B
NODE D TABLE
OLSRp:
Example
OLSR
X
X
B
X
X
B
NODE D TABLE
OLSRp:
Example
OLSR
X
X
B
X
X
B
NODE D TABLE
OLSRp:
Benefits
OLSR
• Reduction in:
– Control traffic
– CPU processing
– Energy consumption
OLSRp: OLSR
Some Results
Conclusions & Future Work
ConclusionsOLSR
& Future Work
• Conclusions:
– OLSRp has similar performance than standard OLSR
– Can dynamically self-adapt to topology changes
– Reduces network congestion
– Saves computer processing and energy consumption
• Future Work:
– Further evaluation of OLSRp performance
– Assessment in real-world testbeds
– Application in other routing protocols
Santander (SPAIN) - September 22-24, 2010
OLSRp: Predicting Control Information to
Achieve Scalability in OLSR Ad Hoc Networks
Questions?