Neighborhood Changing Rate: A Unifying Parameter to Characterize and Evaluate Data Dissemination Scenarios Institut Eurécom † Department of Mobile Communications Sophia Antipolis, France Jérôme Härri † , Biao Zhou ‡ , Mario Gerla ‡ , Fethi Filali † , Christian Bonnet † {haerri,filali,bonnet}@eurecom.fr {zhb,gerla}@cs.ucla.edu University of California ‡ Department of Computer Science Los Angeles, USA 4th IEEE/IFIP Wireless On demand Network Systems and Services (WONS) Obergurgl, Austria January 24th 2007
34
Embed
Institut Eurécom † Department of Mobile Communications Sophia Antipolis, France
Neighborhood Changing Rate: A Unifying Parameter to Characterize and Evaluate Data Dissemination Scenarios. Institut Eurécom † Department of Mobile Communications Sophia Antipolis, France. J é r ô me H ä rri † , Biao Zhou ‡ , Mario Gerla ‡ , Fethi Filali † , Christian Bonnet † - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Neighborhood Changing Rate:
A Unifying Parameter to Characterize and Evaluate
Data Dissemination Scenarios
Institut Eurécom†
Department of Mobile Communications
Sophia Antipolis, France
Jérôme Härri†, Biao Zhou‡, Mario Gerla‡, Fethi Filali†, Christian Bonnet†
• A car shares its packet with all vehicles reachable within its transmission range.
• Objective: Disseminating the packet throughout the network
• Example:
Spreading factor: 2
4Härri et al. Neighborhood Changing Rate (NCR)
Data Dissemination
• At each encounter, the more vehicles the car meets, the more efficient is the spreading factor.
• Example:
Spreading factor: 5
t-3
C1
t-3
C3
t-3
C2
t-3
C4t-2
t
C4
t
C3
t-1
C1
C5
t-2
C6
C6
C6
t
t
C1
C5
t
C5
t-2
C1 C4
C3C2
5Härri et al. Neighborhood Changing Rate (NCR)
Data Dissemination
• In order to reduce the broadcast storm effect, no relaying.
• Each car that receiving the set of data may in turn share it with any encountered vehicle.
• Best dissemination Strategy: At each encounter point, a single car with data shares it with a large set of vehicles.
• Group mobility does not help data dissemination, as in that case, a large set of cars containing data shares it with a potentially smaller set of vehicles.
Data Dissemination Efficiency : Time needed to spread a given set of data to the entire network.
6Härri et al. Neighborhood Changing Rate (NCR)
Data Dissemination
• The data dissemination efficiency in therefore dependant to a large set of parameters:– The rate a car encounters other neighbors.– The number of vehicles met that do not follow a similar trajectory.– …
Objective: Define a single universal metric including all these parameters
In other terms, data dissemination efficiency may depend on a
Neighborhood Changing Rate
7Härri et al. Neighborhood Changing Rate (NCR)
Agenda
• Data Dissemination in Mobile Ad Hoc Network
• NCR– Definition– Example– Justification
• The Mobeyes Protocol
• Performance Results
• Conclusion
8Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood Changing Rate (NCR)
• Let’s define– : Sampling interval equal to the time needed for
a node to move a distance equal to its transmission range
– : Expected Neighbor entering node i’s neighborhood during the time interval
– : Expected Neighbor leaving node i’s neighborhood during the time interval
– : Node i’s nodal degree at time t.
• Then,
[ ] [ ][ ] [ ])(#)(
)(#)(#)(
tNbEtDegE
tNbEtNbEttNCR
inew
i
inew
ileavei
Δ+
Δ+Δ=Δ+
[ ])(# tNbE inew Δ
[ ])(# tNbE ileave Δ
tΔ
tΔ
tΔ
)(tDeg i
9Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood ChangingRate (NCR)
• Example:
The NCR of Car 1 as a function of time, with Δt=1
t-3
t-2 t-1t
t-3t-2 t
t-1t
C2
C2
C1
C3
C3
C1
NCR=0
NCR=_
NCR=_
NCR=1
C4
C4
C1 C1
C4
C4
10Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood ChangingRate (NCR)
Definition (Uniform Mobility Model) :A Uniform Mobility Model (UMM) is a model preserving uniformly distributed velocities and densities
Theorem :Defining speedav- representing and densityav– representing the average node density both generated by an UMM, NCR has the following features
1. 0 ≤ NCR(t) ≤ 12. NCR speedav
3. NCR densityav
Proof: See paper
11Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood ChangingRate (NCR)
• The performance of protocols using data dissemination usually depends on multiple criteria– Speed– Velocity– Mobility pattern– …
• Evaluating a protocol depending on multi-criteria is hard and gives arguable results.
• More specifically, Mobility Patterns are not easily quantifiable because they depend on a too large set of parameters.
• It would be preferable to evaluate it depending on a single
criterion.
12Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood ChangingRate (NCR)
• As NCR is independent to speedav and densityav, In all models where the real speed and density diverge
from the initial speedav and densityav,
NCR controls the set of parameters that generates the complex spatial and temporal dependencies we may observe in realistic mobility patterns
• Specific Topologies or Mobility Patterns may become less relevant to evaluate the performance of dissemination protocols
• With a given speedav , densityav, and NCR, we can perform cross-topology and cross-mobility patterns performance evaluation.
13Härri et al. Neighborhood Changing Rate (NCR)
Neighborhood ChangingRate (NCR)
• A similar situation also exists in Transportation Planning:– How to represent traffic flows in transportation that depend
on multi-parameters such as:• Speed, density, volume/capacity ?
– Level of Service (LOS) : Works like an American report card grade, using the letters A through F, with A being best and F being worst.
– By using LOS classification and referring to a traffic situation as having a particular LOS , engineers can have a global knowledge of traffic condition in a particular area.
• NCR is designed to have the same usage:– By referring to data dissemination as having a particular
NCR, we can have an intuitive vision of its efficiency, and thus evaluate accurately VANET Protocols using this feature.
14Härri et al. Neighborhood Changing Rate (NCR)
Agenda
• Data Dissemination in Mobile Ad Hoc Network
• NCR– Definition– Example– Justification
• The Mobeyes Protocol
• Performance Results
• Conclusion
15Härri et al. Neighborhood Changing Rate (NCR)
The Mobeyes Protocol
• Mobeyes [1] is a protocol for sensed data mining in vehicular environments:– Periodic diffusion of a summary of sensed data– On demand harvesting of sensed data
• Mobeyes Architecture– Mobeyes Sensing Interface (MSI) : Interface responsible for the
access to the sensors or GPS– Mobeyes Data Processor (MDP): Reads raw data and generates the
Harvesting rate for scenarios with same density, speed and NCR, and different mobility models and topologies
33Härri et al. Neighborhood Changing Rate (NCR)
Cross-Topology Comparison
• All Models having the same NCR
Time before which 100% of the data is harvested
Mobeyes + Map
Mobeyes + Triangle
RWM + Triangle
15 m/s 58.43s 50.26s 61.58 s
velocityav
topology
34Härri et al. Neighborhood Changing Rate (NCR)
Conclusion
• NCR is a novel parameter describing data dissemination
• NCR is able to describe spatial and temporal dependencies, not covered by speed or density.
• NCR is an unifying parameter, as it regroup mobility patterns and topology parameters.
• Data dissemination in any kind of topology and for any type of mobility pattern, can be the fully control by three parameters:– Average Speed– Average Density– NCR