Analysis of the Impact and Interactions of Protocol and Environmental
Parameters on Overall MANET Performance
Michael W. Totaro and Dmitri D. PerkinsCenter for Advanced Computer Studies
University of Louisiana at Lafayette
Presentedby
Michael W. Totaro
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Introduction and Motivation Important design challenge of MANETs:
maximize overall performance of protocols operating in a MANET
Impact that one or more factors have on MANET performance?
2k factorial design—an important tool that may aid researchers in analyzing effect of factors on MANET performance
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
2k Factorial DesignProcess
Code each factor to a “+” and a “-” level Design matrix: All possible combinations of factor levels Example for k = 3 factors:
Make the8 simulationruns, andmeasure theeffects ofthe factors!
2k Factorial DesignMain Effect of a Factor
Main effect of a factor is the average difference in the response when this factor is at its “+” level as opposed to its “-” level:
2k Factorial DesignMain Effect of a Factor – cont’d
The main effects measure the average change in the response due to a change in an individual factor, with this average being taken over all possible combinations of the other k-1 factors (numbering 2k-1).
2k Factorial DesignMain Effect of a Factor – cont’d
We can rewrite the above as “Factor 1” column ● “Response” column / 2k-1
-R1 + R2 – R3 + R4 – R5 + R6 – R7 + R8
e1 =
4
2k Factorial DesignFactor Interaction
Two factors A and B are said to interact if the effect of one depends upon the level of the other
Conversely, these two factors, A and B, are said to be noninteracting if the performance of one is not affected by the level of the other
We shall look at examples of interacting factors and noninteracting factors
2k Factorial DesignExamples of Noninteracting and Interacting Factors
A1 A2
B1 3 5
B2 6 8
Noninteracting Factors
Interacting Factors
A1 A2
B1 3 5
B2 6 9
As the factor A is changedfrom level A1 to level A2,the performance increasesby 2 regardless of the levelof factor B
As the factor A is changedfrom level A1 to level A2,the performance increaseseither by 2 or 3 dependingupon whether B is at levelB1 or level B2, respectively
2k Factorial DesignExamples of Noninteracting and Interacting Factors – cont’d
Performance
Graphical representation of interacting and noninteracting factors.
6
2
8
A1 A2
B2
B1
Performance
6
2
8
B1
A2
A1
B2
(a) No Interaction
Performance
6
2
8
A1 A2
B2
B1
Performance
6
2
8
B1
A2
A1
B2(b) Interaction
2k Factorial DesignInteraction Effects
1 x 3 interaction effect: “Factor 1” ● “Factor 3” ● “Response” / 2k-1
R1 - R2 + R3 - R4 – R5 + R6 – R7 + R8
e13 =
4
Addresses the question: “Does the effect of a factor depend on level of others?”
Sign of effect indicates direction of effect on response of moving that factor from its “-” to its “+” level
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Related Work Interest in cross-layer factor interaction in MANETs is not
entirely new. Performance metrics for assessing the behavior of MANETs
are identified, discussed, and, most especially, partitioned into three classification levels [1] Thread-task level metrics (algorithmic level) such as average
power expended and task completion time Diagnostic packet level metrics such as end-to-end throughput,
end-to-end delay, link utilization, and packet loss, which characterize network behavior at the packet level
Scenario metrics that describe the network environment and define the scenario; these include: nodal movement/topology rate of change, number of network nodes, area size of network, density of nodes per unit area, offered load and traffic patterns, and number of unidirectional links
Related Work – cont’d A comprehensive analysis of five factors—node speed, pause-time,
network size, number of traffic sources, and type of routing (source vs. distributed)—was done using a factorial experimental design in an effort to identify and quantify the effects and two-way interactions of these factors on three performance responses: throughput, average routing overhead, and power consumption [2]
Potential benefits may be derived by information exchange between the lower layer, routing layer, and transport layer, which is useful in the design and standardization of an adaptive architecture that can exploit the interdependencies among link, medium access, network, and applications protocols [3]
The underlying premise in cross-layer interaction analyses is that, by learning more about factor and two-way interactions on the performance of MANETs, researchers may want to consider taking these effects into account in the design of future protocols
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Methodology
Effects of factors?
Methodology – cont’dPartial Design Grid (Coded)
Methodology – cont’dPartial Design Grid (Uncoded)
Methodology – cont’d Simulation—QualNet
26 factors = 64 experimental runs Replications = 5 Total running time = 320 seconds LAR1 routing protocol Free-space model 2 Mbps Packet size = 512 bytes Pause time = 25 seconds Transmission range = 250 meters Terrain dimensions*: N = √ [ (MR2π / X) – 1 ]
where X = average number of neighbors M = number of nodes R2 = transmission range
*Using the formula for computing the average number of neighbors for a node, derived by Ihklas Ajbar.
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Analysis, Results, and ModelsScatterplot—Packet delivery ratio
Analysis, Results, and ModelsScatterplot—End-to-end delay
Analysis, Results, and ModelsScatterplot—Control packet overhead
Analysis, Results, and ModelsMain effects—Packet delivery ratio
Analysis, Results, and Models – cont’d
Main effects—End-to-end delay
Analysis, Results, and Models – cont’d
Main effects—Control packet overhead
Analysis, Results, and Models – cont’d
Two-way factor interactions—Packet delivery ratio
Analysis, Results, and Models – cont’d
Two-way factor interactions—Control packet overhead
Analysis, Results, and Models – cont’d
Two-way factor interactions—End-to-end delay
Analysis, Results, and Models – cont’d
Response-surface plots
Analysis, Results, and Models – cont’dContour plots
Analysis, Results, and Models – cont’d
Predictive effects—Packet delivery ratio
Analysis, Results, and Models – cont’d
Predictive effects—End-to-end delay
Analysis, Results, and Models – cont’d
Predictive effects—Control packet overhead
Analysis, Results, and Models – cont’d
Prediction profile
Analysis, Results, and Models – cont’d
Regression Models—Packet delay ratio
Analysis, Results, and Models – cont’d
Regression Models—End-to-end delay
Analysis, Results, and Models – cont’d
Regression Models—Control packet overhead
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Future Work and Open Questions
Validation of prediction (regressions) models
Other factors (e.g., power consumption, etc.)?
Other empirical models (e.g., neural networks, time-series prediction)?
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
Questions
Thank you!
Topics Introduction and Motivation 2k Factorial Design (A Brief Tutorial) Related Work Methodology Analysis, Results, and Models Future Work and Open Questions Questions References
References1. M. W. Subbarao, “Ad Hoc Networking Critical Features
and Performance Metrics”. White paper, Wireless Communications Technology Group, National Institutes of Standards and Technology, September 15, 1999
2. D. D. Perkins, H. D. Hughes, and C. B. Owen, “Factors Affecting the Performance of Ad Hoc Networks”. Proceedings of IEEE International Conference on Communications (ICC 2002), New York, April 2002
3. J. Lee, S. Singh, and Y. Roh, “Interlayer Interactions and Performance in Wireless Ad Hoc Networks,” Internet – Draft, IRTF ANS Working Group; http://www.flarion.com/ans-research/Drafts/draft-irtf-ans-interlayer-performance-00.txt, accessed 1/29/2004