Extending Network Knowledge:
    Making OLSR a Quality of
    Service Conducive Protocol

	Pedro Villanueva pvillanu@site.uottawa.ca
	Thomas Kunz tkunz@sce.carleton.ca
	Pramod Dhakal pdhakal@eion.com

Outline

	Introduction
	OLSR Protocol
	Extended Topology Knowledge (TK)
	Methodology
	Results
	Conclusions
	Future Work

Introduction

	MANETS require robust and efficient routing protocols.
	Proactive protocols are preferred over reactive protocols to support critical systems and QoS.
	OLSR reduces overhead. But, its partial topology view constrains path computation.
	We gradually extend the partial network view of OLSR.

OLSR Protocol

	OLSR (Optimized Link State Routing) allows manipulating the amount of advertised topology information.
	OLSR reduces the amount of advertised links, advertising nodes and forwarding nodes; by using the MPR mechanism.
	Periodic Hello and Topology Control messages advertise the known topology.

MPR Mechanism

Extended Topology Knowledge

	The partial view of the network topology represents a severe shortcoming when constructing routing paths.
	Lack of network (i.e. load), node (i.e. battery) and link (i.e. quality) status information acts against robustness and reliability
	Multiple paths could be provided.

Extended Topology Knowledge

Redundant Topology Information

	Five strategies defining the links to be advertised (TC_Redundancy).
		TC=0. Links to its MPR Selectors (MPRS).
		TC=1. Every node, links to MPRs and MPRSs
		TC=2. Every node, links to every one-hop neighbour.
		TC=3. Selected MPRs advertise links to MPRs and MPRSs
		TC=4. Selected MPRs advertise every link to       every one-hop neighbour.

MPR Redundancy

	Defines the desired number of one-hop neighbours to reach every two-hop neighbour (MPR Coverage).
	Analyzed values: MPR =1,2,3.
	Effect: Increases the number of advertising nodes, advertised links and forwarders.

Methodology

	NS-2 simulation using the OOLSR implementation of OLSR (Hipercom project).
	Three different sets of scenarios
		Static scenarios without data traffic.
		Static scenarios with data traffic.
		Mobile scenarios with data traffic.
	Same scenarios were used for each strategy.

Simulation parameters for static scenarios

Generated TC Messages

Overhead vs TK

Simulation parameters for data traffic

MPRs vs TK

Delivery rate on static networks

Simulation parameters for mobile scenarios

Accuracy of TK

Per Packet Delay

NS-2 Bug

	Generated outliers on packet delay
	Statistical analysis was used to remove them

Conclusions

	High levels of TK can be achieved at different costs without impacting delivery rate
	TC=2 maximizes TK. TC=0 minimizes overhead. TC=4 offers a fair trade-off
	With data traffic no strategy keeps large TK

Conclusions

	Extended TK may support construction of better paths and QoS
	Higher reliability and robustness may be achieved if new metrics are applied
	Results can be used as a guideline to tune up OLSR

Future Work

	Advertise additional node and link status information to apply more powerful metrics
	Use alternative metrics for path computation
	Design new criteria to select the links to be advertised (i.e. quality of links, node’s resources and network load)

Slide 24

Extending Network Knowledge:
    Making OLSR a Quality of
    Service Conducive Protocol

	Pedro Villanueva pvillanu@site.uottawa.ca
	Thomas Kunz tkunz@sce.carleton.ca
	Pramod Dhakal pdhakal@eion.com