an analysis of the information spreading delay in heterogeneous mobility dtns
TRANSCRIPT
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 1/45
An Analysis of the Information Spreading Delay inHeterogeneous Mobility DTNs
Andreea Picu Dr. T. Spyropoulos Theus Hossmann
Computer Engineering and Networks Laboratory (TIK)Communication Systems Group (CSG)
July 24, 2012
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 2/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Outline
1
Motivation & Goals2 Our New Analysis
Model and Delay Def.sDelay Derivation: The Gist
The Bounds3 Empirical Results and Insights
Bounds ValidationScaling with Network Size
4 Applications
5 Conclusion & Outlook
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 2 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 3/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Epidemic Information Spreading
Some node generates a message m...
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 3 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 4/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Epidemic Information Spreading
... “infected” nodes replicate m to every encountered node...
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 3 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 5/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Epidemic Information Spreading
... “infected” nodes replicate m to every encountered node...
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 3 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 6/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Epidemic Information Spreading
... until the whole network is “infected”.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 3 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 7/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Epidemic Information Spreading
Simple
Robust
Works in any local information exchange context
Sometimes, the only option...
In DTNs:
broadcast routing / flooding
unicast routingdistributed estimationetc
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 4 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 8/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Common assumptions in existing analyses
1 Permanent network connectivitysynchronous operationability to communicate with all neighbors at will
2 Specific graph geometriescomplete, bipartite, Erdos-Renyi, regular etc.
3 Homogeneous communication rate λ
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 5 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 9/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
In the DTN context
All nodes are identical ⇒ can be treated as group
Real DTN experiments ⇒ heterogeneous mobility
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 6 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 10/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
In the DTN context
All nodes are identical ⇒ can be treated as group
Real DTN experiments ⇒ heterogeneous mobility
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 6 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 11/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Results
and Insights
Applications
Conclusion &Outlook
What we do...
All nodes are beautiful snowflakes
Heterogeneous communication rates λi j
N -state Markov chain ⇒ N -state Markov chain
Bounds for delay distribution
Dependency on contact graph properties
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 7 of 23
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 12/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Results
and Insights
Applications
Conclusion &Outlook
Heterogeneous Mobility
(i , j) meeting probability: pci j = λ i j
∑ λ i j
Contact probability matrix:
Pc=
pc
pc
N
pc pcN ⋮ ⋮ ⋮
pcN
pci j estimated from pairwise intercontacts of (i, j)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 8 of 23
l d
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 13/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Results
and Insights
Applications
Conclusion &Outlook
Delays We Are Interested in
Total spreading delay:DSSS = inf {n > ∶ S(T n) = V } (1)
Partial spreading delay:
DSSS(α) = inf {n > ∶ S(T n) = α} (2)
Transition delay:
Dα = DSSS(α + ) − DSSS(α) (3)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 9 of 23
D l D i ti Th Gi t
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 14/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Results
and Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
D l D i ti Th Gi t
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 15/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
D l D i ti Th Gi t
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 16/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
D l D i ti Th Gi t
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 17/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
Delay Derivation The Gist
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 18/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
Delay Derivation The Gist
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 19/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
Delay Derivation: The Gist
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 20/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Delay Derivation: The Gist
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 10 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 21/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 11 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 22/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 11 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 23/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 11 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 24/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Expected delay from state x to state y :
pxy
pxy = pcia + pcib + pc jc + pckb
= i∈Sα ; j∉Sα
pci j
= cut volume of vertex set Sα
= ∂(Sα)
Expected transition delay: E[Dα] maxx , y
pxy =
minSα
∂(Sα)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 12 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 25/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Expected delay from state x to state y :
pxy
pxy = pcia + pcib + pc jc + pckb
= i∈Sα ; j∉Sα
pci j
= cut volume of vertex set Sα
= ∂(Sα)
Expected transition delay: E[Dα] maxx , y
pxy =
minSα
∂(Sα)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 12 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 26/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Expected delay from state x to state y :
pxy
pxy = pcia + pcib + pc jc + pckb
= i∈Sα ; j∉Sα
pci j
= cut volume of vertex set Sα
= ∂(Sα)
Expected transition delay: E[Dα] maxx , y
pxy =
minSα
∂(Sα)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 12 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 27/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Expected delay from state x to state y :
pxy
pxy = pcia + pcib + pc jc + pckb
= i∈Sα ; j∉Sα
pci j
= cut volume of vertex set Sα
= ∂(Sα)
Expected transition delay: E[Dα] maxx , y
pxy =
minSα
∂(Sα)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 12 of 23
Transition Delay
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 28/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Transition Delay
Expected delay from state x to state y :
pxy
pxy = pcia + pcib + pc jc + pckb
= i∈Sα ; j∉Sα
pci j
= cut volume of vertex set Sα
= ∂(Sα)
Expected transition delay: E[Dα] maxx , y
pxy =
minSα
∂(Sα)
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 12 of 23
Total Delay Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 29/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
y
Graph α-conductance: Φα = minSα
∂(Sα)α ⋅
(N − α
)
Graph conductance: Φ = minα
Φα
Key Inequality
minSα
∂
(Sα)
Φα ⋅ α
(N − α
)
Φ⋅ α
(N − α
)
E
[DSSS
] =
N −
α=
E
[Dα
]
N −
α=
Φα ⋅ α(N − α)
N −
α=
Φ ⋅ α(N − α) = ln(N − )
N Φ
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 13 of 23
Total Delay Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 30/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
y
Graph α-conductance: Φα = minSα
∂(Sα)α ⋅
(N − α
)
Graph conductance: Φ = minα
Φα
Key Inequality
minSα
∂
(Sα)
Φα ⋅ α
(N − α
)
Φ⋅ α
(N − α
)
E
[DSSS
] =
N −
α=
E
[Dα
]
N −
α=
Φα ⋅ α(N − α)
N −
α=
Φ ⋅ α(N − α) = ln(N − )
N Φ
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 13 of 23
Total Delay Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 31/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
y
Graph α-conductance: Φα = minSα
∂(Sα)α ⋅
(N − α
)
Graph conductance: Φ = minα
Φα
Key Inequality
minSα
∂
(Sα) Φ
α ⋅ α
(N − α
) Φ ⋅ α
(N − α
)
E
[DSSS
] =
N −
α=
E
[Dα
]
N −
α=
Φα ⋅ α(N − α)
N −
α=
Φ ⋅ α(N − α) = ln(N − )
N Φ
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 13 of 23
Total Delay Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 32/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
y
Graph α-conductance: Φα = minSα
∂(Sα)α ⋅
(N − α
)
Graph conductance: Φ = minα
Φα
Key Inequality
minSα
∂
(Sα) Φ
α ⋅ α
(N − α
) Φ ⋅ α
(N − α
)
E
[DSSS
] =
N −
α=
E
[Dα
]
N −
α=
Φα ⋅ α(N − α)
N −
α=
Φ ⋅ α(N − α) = ln(N − )
N Φ
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 13 of 23
What Just Happened?
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 33/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
pp
Exact delay formula based on cut volumes ∂(Sα)st bound:replace all α−cut volumes with resp. α−conductance Φα
nd bound:
replace all cut volumes with overall graph conductance Φ
Implications:
Bounds’ tightness depends on graph structure
No discernable communities ⇒ very tight bounds
Clear community structure ⇒ less tight bounds
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 14 of 23
Distribution Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 34/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
P[ DSSS = k] =N
α=
wα pα( − pα)k−,i.e., sum of geometrics.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 15 of 23
Distribution Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 35/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
P[ DSSS = k] =N
α=
wα pα( − pα)k−,i.e., sum of geometrics.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 15 of 23
Distribution Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 36/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
P[ DSSS = k] =N
α=
wα pα( − pα)k−,i.e., sum of geometrics.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 15 of 23
Distribution Bound
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 37/45
Outline
Motive & Goals
Our New Analysis
Model and Delay Def.s
Delay Derivation: The Gist
The Bounds
Empirical Resultsand Insights
Applications
Conclusion &Outlook
P[ DSSS = k] =N
α=
wα pα( − pα)k−,i.e., sum of geometrics.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 15 of 23
The NCP: Indicator of Community Structure
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 38/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Bounds Validation
Scaling with Network Size
Applications
Conclusion &Outlook
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 16 of 23
Expected Spreading Delays
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 39/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Bounds Validation
Scaling with Network Size
Applications
Conclusion &Outlook
Expected spreading delays Coefficients of variation
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 17 of 23
Spreading Delay CCDFs
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 40/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Bounds Validation
Scaling with Network Size
Applications
Conclusion &Outlook
Infocom MIT
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 18 of 23
Spreading Delay CCDFs
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 41/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Bounds Validation
Scaling with Network Size
Applications
Conclusion &Outlook
HCMM SLAW and MIT
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 19 of 23
Algorithms Based on Epidemic Spreading
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 42/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Randomized forwarding
Distributed estimation
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 20 of 23
Conclusion
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 43/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
New analysis for epidemic spreading in DTNs:
Much more generic mobility
Bounds for expected delay & delay distribution
Bound tightness linked to graph structure
Delay scaling behaviour linked to graph structure
Applies to algorithms based on epidemic spreading
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 21 of 23
Outlook
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 44/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Further relax mobility assumptionsSemi-MarkovNo exponential assumption
Further explore the applications domain
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 22 of 23
The End
7/22/2019 An Analysis of the Information Spreading Delay in Heterogeneous Mobility DTNs
http://slidepdf.com/reader/full/an-analysis-of-the-information-spreading-delay-in-heterogeneous-mobility-dtns 45/45
Outline
Motive & Goals
Our New Analysis
Empirical Resultsand Insights
Applications
Conclusion &Outlook
Thank you for your attention.
Please, ask questions.
Picu, Spyropoulos, Hossmann Information Spreading in DTNs 23 of 23