Routing and Performance Evaluation of Disruption Tolerant Networks. Mouhamad IBRAHIM Ph.D. defense Advisor: Philippe Nain INRIA Sophia Antipolis 14 November, 2008. Thesis outline. Part I: Design and performance evaluation of routing protocols for disruption tolerant networks - PowerPoint PPT Presentation
Welcome message from author
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
1
Routing and Performance Routing and Performance Evaluation of Disruption Tolerant Evaluation of Disruption Tolerant
Part I: Design and performance evaluation of routing
protocols for disruption tolerant networks
Part II: Design and performance evaluation of medium
access control protocol for IEEE 802.11 standard
3
Routing in mobile ad hoc networks
Mobile Ad Hoc Networks (MANETs) No fixed infrastructure Nodes communicate in a peer to peer mode
with other nodes Nodes work as routers: Store-Forward
Routing in MANETs: Main assumption Existence of end-to-end paths between
Source-Destination pairs
4
Routing challenges in MANETs Instability of wireless paths: node mobility, low
node density, interferences,… Does not help to establish and maintain routes
Appearance of Disruption/Delay Tolerant Networks (DTNs): disconnected mobile networks Often there is no end-to-end path among
Source-Destination pairs
take advantage of node mobility to perform routing
Store-Carry-Forward
5
Store-Carry-Forward: how does it work?
S
V1
V3D
V2
R
6
Routing approaches for DTNs
Classification based on the degree of knowledge that nodes have about their future contact opportunities
Four classes of routing techniques: Scheduled-contact based routing Controlled-contact based routing Predicted-contact based routing Opportunistic-contact based routing
7
Opportunistic-contact based routing
Flooding mechanism Epidemic routing protocol
Limit the number of hops Multicopy Two-hop Relay protocol
Limit the number of copies Spray-and-Wait protocol
Question: To what extent we can push the performance if we
increase number of contact opportunities: Throwboxes
8
Throwboxes (1)
Throwboxes are fixed relays with better storage and energy capabilities
Battery powered for short term use or solar panel for long term use
Photos are taken from http://prisms.cs.umass.edu/dome/
9
Throwbox (2)
Operate in Store-Forward paradigm
Promising approach to route messages in DTNs Adding one throwbox on UMass DieselNet
improves packet delivery by 37% and reduces message delivery delay by 10%[1]
Research still in its early stage!!Part I: Evaluate and design routing techniques for
opportunistic DTNs augmented by throwboxes
[1] N. Banerjee et al. An energy-efficient architecture for DTN throwboxes. Infocom 2007.
10
Opportunistic DTNs: Inter-meeting times
Characteristic of inter-meeting times among nodes
Random mobility: Inter-meeting times mobile/mobile have
shown to follow an exponential distribution [Groenevelt et al.: The message delay in mobile ad hoc networks. Performance Evaluation, 2005]
Human mobility: Inter-meeting times mobile/mobile have
shown to follow power law distribution [Chaintreau et al.: Impact of human mobility on the design of opportunistic forwarding algorithms. Infocom, 2006]
11
Opportunistic-contact: Random mobility
X1
X2
V1
V2
Directions (αi) are uniformly distributed (0, 2π) Speeds (Vi) are uniformly distributed (Vmin,Vmax) Travel times (Ti) are exponentially /generally distributed
Directions (αi) are uniformly distributed (0, 2π) Speeds (Vi) are uniformly distributed (Vmin,Vmax) Travel times (Ti) are exponentially /generally distributed
R
T1, V1
T2, V2
R
Next positions (Xi)s are uniformly distributed Speeds (Vi)s are uniformly distributed (Vmin,Vmax)
Next positions (Xi)s are uniformly distributed Speeds (Vi)s are uniformly distributed (Vmin,Vmax)
α1
α2
Random Waypoint model (RWP)
Random Waypoint model (RWP)
Random Direction model (RD)
Random Direction model (RD)
12
Mobile/box inter-meeting timesCCDF on a linear-log scale: log(Pr(τ > x)) = log(e - μ x )= - μ x )CCDF on a linear-log scale: log(Prτ((x > log(e =μ x -μ x - =(
Simulation N = 1Exponential –μx Simulation N = 5Exponential –5μxSimulation N = 10Exponential –10μx
Simulation N = 1Exponential –μx Simulation N = 5Exponential –5μxSimulation N = 10Exponential –10μx
13
E[c].μ]E[
E[c]πc
Parameter μ (1) Stationary probability to find the mobile within
neighborhood of a box
f(.,.) stationary spatial pdf of the mobility model
Using Renewal theory, we have
y),f(x,πrdudvv)f(u,πκ
2c )( Lr
Contact time
C1 C2 C3
Time
τ1 τ2 τ3
14
Parameter μ (2) Unconditioning on throwbox location within the network area
LxL
Case of Random Direction model: mobile nodes are uniformly distributed[1]
and hence
independent of throwboxes pdf distribution!!
LxL1
cdxdyy)g(x,y)f(x,
]E[V
r2μ
pdf of throwboxes distribution
pdf of throwboxes distributionStationary pdf of location
for mobility model
Stationary pdf of location for mobility model
2L
1y)f(x,
]E[VL
r2μ
1c
2
[1] P. Nain et al. Properties of random direction models. Infocom 2005.
15
Parameter μ (3) Case of Random Waypoint model: mobile
nodes are distributed around the center[3]
μ depends on throwboxes spatial distributionThrowboxes uniformly distributed
Throwboxes generally distributed, e.g.
2
1),(
Lyxg
),(),( yxfyxg ]E[VL
1.36r2μ
1c
22
]E[VL
r2μ
1c
21
[3] J.-Y. Le Boudec and M. Vojnovic. Perfect simulation and stationarity of a class of mobility models, Infocom 2005.
16
Performance evaluation of relaying protocols in DTNs with throwboxes
Copies make at MAX two hops between Source/Destination
Copies make at MAX two hops between Source/Destination
Multicopy two-hop protocol (MTR)
19
S
V1
V3D
V2
R
B1
B2
Network model
Source node
Source node
Destination node
Destination node
N-1 mobile relay nodes
N-1 mobile relay nodes
M throwboxes
M throwboxes
Mobile/mobile: Exponential with λ[4]
Mobile/mobile: Exponential with λ[4]
Mobile/box: Exponential with μ
Mobile/box: Exponential with μ
[4] R. Groenevelt, P. Nain, and G. Koole. The message delay in
mobile ad hoc networks. Performance Evaluation, 2005.
20
Metrics of interest
Distribution and mean value of
Delivery delay T user side
Total number of generated copies G when one packet is to be send from source to destination network operator side
21
Markov analysis Two-dimensional continuous time absorbing
Markov chain I(t) = (R(t),B(t)) as follows:
For t < T:
R(t) {1,2,…,N} number of mobile nodes holding a copy of the packet (source included)
B(t) {0,1,2,…,M} number of throwboxes holding a copy of the packet (assumed fully disconnected)
For t > T, I(t)= {a} absorbing state, i.e. when destination receives the packet
22
MTR protocol: Delivery delay (1) Approach to solve: Stochastic analysis
Delivery delay TMTR is the minimum of N + M mutually independent R.V.s
TMTR = (DSD, Dr1, Dr
2,…, DrN-1, DB
1,…, DBM)
Hence distribution of TMTR reads asMNtMN
MTR ttetTr )1()1()( 1)(
source destination: exponential with rate λsource relay destination: sum of two
exponentials with rate λ source throwbox destination: sum of two exponentials with rate μ
23
MTR protocol: Delivery delay (2) and mean of TMTR reads as
Using fluid model, we obtained also asymptotic expression for E[TMTR] when N or M go large
mn1N
0n
M
0mMTR )
MμNλ
μ()
MμNλ
λ(m)!(n
m
M
n
1N
MμNλ
1]E[T
24
MTR protocol: # of generated copies
Define Pra(n,m) as probability that last visited state before absorption is state (n,m)
Pra(n,m) is sum of probabilities of different paths joining state (1, 0) to state (n,m) These probabilities are all equal. Their total number is
The probability distribution of GMTR reads as
1m
0j
1n
1ia MμNλ
μj)(M
MμNλ
λi)(N
MμNλ
mμnλ
1n
1mnm)(n,Ρr
1n
1mn
MNknknkGNk
MknaMTR
,...2,1,),(Pr)(Pr),min(
),1max(
25
Epidemic protocol: Delivery delay Approach to solve: Theory of absorbing Markov
chain
Delivery delay TER represents time to absorption
Q = infinitesimal generator of Markov chain
M = transition matrix among non-absorbing states
00
RMQ
,j)(1,m]E[T1)(MN
1jER
m*(i,j) is the (i,j)th entry of M-1
[1,1,...1],[1,0,...0]bε,be1x)Pr(T MxER
26
Epidemic: # of generated copies
Define Pra(n,m) as probability that last visited state before absorption is state (n,m)
Case of epidemic protocol: transition rates are state dependent approach reported by [Gaver et al.: Finite Birth-And-Death Models in Randomly Changing Environments, 1984]
The probability distribution of GER follows then
a)m),q((n,mN)n(1,mm)(n,Ρra
MN1,2,...k,n)k(n,Prk)(GPrN)min(k,
M)kmax(1,naER
27
Case of connected Throwboxes
Underlying assumption: Pass a copy to one throwbox to let all the others infected
Same expressions hold by substitutingM 1μ M μ
28
Model validation: Delivery delay
Epidemic protocol
RWP model
Epidemic protocol
RWP model
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
29
Model validation: Delivery delay
MTR protocolRWP model
MTR protocolRWP model
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
30
Performance evaluation framework for throwboxes-augmented DTNs
Objective: Framework to evaluate and analyze performance of various routing strategies for DTNs extended with throwboxes
31
Proposed five routing strategies (1)
Main idea: define possible message forwarding interactions among the Source, Mobile relays, Throwboxes and the Destination
Ultimate goal: exploit throwboxes presence to minimize copies generations at mobile nodes
32
Proposed five routing strategies (2) Common forwarding interactions:
Source Relay
Relay Destination
Relay Destination
Source Relay
Relay Throwbox
Relay Throwbox
Relay Throwbox
Relay Destination
Relay Destination
Strategy VStrategy IVStrategy IIIStrategy IIStrategy I
Infected throwbox Mobile relay
Destination
Particular interactions for each strategyParticular interactions for each strategy
Infected mobile relay Mobile relay
Source ThrowboxDestination
33
Metrics of interestUnder a given routing strategy s:
1- Mean delivery delay between a Source/Destination E[Ts]
Mean number of valuable transmissions E[Gs], i.e. those made only by mobile nodes plus the source
2- Mean number of mobile relays infected by the source, Is
3- Mean number of infected throwboxes, Ks
4- Proba. Source delivers message to destination, PrSs
5- Proba. Mobile relay delivers message to destination, PrRs
34
Modeling framework (1) Three-dimensional continuous time absorbing
Markov chain As(t) = (Is(t), Js(t), Ks(t)) as follows:
For t < Ts, As(t) = (Is(t), Js(t), Ks(t)):
Is(t) Number of mobile nodes infected by the source
Js(t) Number of mobile nodes infected by the
throwboxes
Ks(t) Number of infected throwboxes
For t > Ts: As(t) = {a} absorbing state
35
Modeling framework (2)
)1kj,(i,Fk)j,S(i,
k)j,γ(i,k),1j(i,F
k)j,S(i,
k)j,β(i,
k)j,,1(iFk)j,S(i,
k)j,α(i,k)j,(i,uk)j,(i,F
ss
sss
i,j,k i,j+1,k
i+1,j,k
i,j,k+1
a
α(i,j,k)
β(i,j,k)
θ(i,j,k)
γ(i,j,k)
S(i,j,k)
Fs(i,j,k) is mean value of metric Fs till absorption starting from (i,j,k)
Fs(i,j,k) is mean value of metric Fs till absorption starting from (i,j,k)
Mean value of metric Fs at (i,j,k)
Mean value of metric Fs at (i,j,k)
k)j,S(i,
1k)j,(i,u s 1- Mean sojourn time Ts
2- Mean number of mobile relays Is
k)j,S(i,
k)j,θ(i,ik)j,(i,u s 3- Mean number of throwboxes
Ks k)j,S(i,
k)j,θ(i,k)j,(i,us k4- Proba. delivery by source
PrSs k)j,S(i,
1k)j,(i,us 5- Proba. delivery by relay PrRs
k)j,S(i,
)1(k)j,(i,us
ji
36
Modeling framework (3) Values of Fs are known at last states only
one possible transition to state {a}, e.g.
Iterating recursive equation till initial state (1,0,0): (1,0,0)T]E[T
ss
M)(N,0,uM)(N,0,F ss N,0,M
a
θ(N,0,M)
(1,0,0)PrR
(1,0,0)PrS(1,0,0)K1)(1,0,0)(I]E[G
s
ssss
Known!
37
Modeling framework (4)
To compute E[Ts] and G[Ts] under a given strategy Define corresponding state space Es and infinitesimal generator Qs(t)
38
Framework validation
N = M Metric Analytical Simulation Rel. error %
10
T(1,0,0) 8.15 103 7.95 103 2.54
I(1,0,0) 3.21 3.27 1.75
K(1,0,0) 1.48 1.37 7.83
PrS(1,0,0) 0.27 0.28 4.21
PrR(1,0,0) 0.55 0.56 1.15
100
T(1,0,0) 2.30 103 2.6 103 2.75
I(1,0,0) 8.35 8.46 1.29
K(1,0,0) 4.34 4.61 5.81
PrS(1,0,0) 0.07 0.08 6.48
PrR(1,0,0) 0.79 0.78 3.0
Strategy II: Analytical versus simulation results Strategy II: Analytical versus simulation results
39
Comparing E[T] and E[G] with respect to Epidemic protocol
Strategy II Strategy IV Strategy V
Strategy II Strategy IV Strategy V
40
Diameter of epidemic protocol
Context: Opportunistic DTNs running epidemic protocol WITHOUT throwboxes
Objective: Examine the mean length of forwarding
path
41
Diameter of epidemic protocol Instance of epidemic tree:
XS,D denote number of intermediate hops between S and D
Aim is to compute E[XS,D]: diameter of epidemic protocol
R5
S
R2 R1
D R3
R4
42
Diameter computation (1)
Approach to solve: Theory of recursive tree
Recursive tree is like any tree on a graph, however, nodes are labeled with their joining instants to the tree
Example: recursive tree of order 4 1
2
3 4
2
4
3
1
2
3
4
1
2 3 4
1
2
3
4
1
3
2
4
1
E[Xi,j] is known for random tree
E[Xi,j] is known for random tree
43
Diameter computation (2)
Conditioning on possible labels of the destination among the N nodes
Look to the impact of limiting number of forwarding hops on relaying performance
Using the framework, we analyze different dissemination algorithm with limited number of hops
1
11 1
1]log[][
1][
N
i
N
nnSD i
i
NNH
NXE
44
Epidemic protocol: Limiting # of hops
Max. hop = 2Max. hop = 3 Max. hop = 4Max. hop = 5
Max. hop = 2Max. hop = 3 Max. hop = 4Max. hop = 5
45
Part II
46
Adaptive Backoff Algorithm for IEEE 802.11
Motivation: IEEE 802.11 performs poorly in congested network Following a successful transmission, source
station chooses backoff duration randomly in {0,…,CW0}
Objectives: Adaptive algorithm aware of active stations
Maximize system throughput and minimize end-to-end delay
Inadequate for large networks
Inadequate for large networks
47
σ2T
1τc
*
N
How to transmit at optimal transmission probability τ*
Bianchi model[5]: Transmission probability
Our idea:
)p)(2(1pCW1)p)(CW2(1
p)22(1τm
00
m= log(CWmax/CW0)
m= log(CWmax/CW0)
)p)p(2p(1τ
p)2)(1τ-(2CWm*
*
min
[5] G. Bianchi. Performance analysis of the IEEE 802.11 distributed coordination function. JSAC 2000.
48
Estimating # of active stations Active stations are decoding all transmitted
packets on the channel identify emitting stations
Stations counts signs of life coming from others stations signs of life: error free data and RTS
packets Measured during virtual transmission
times
Samples used as input to a corrected WMA filter
β2
0iCW
2
0iNCW
αN
ik
ikik
Ňk: sample at kth periodCWk: window at kth periodα, β : correcting factors
49
Algorithm performanceAdaptiveStandard
AdaptiveStandard
Group entrance
Group departure
Group entrance
Group departure
50
Conclusions (1)
Accurate approximation for meeting rate between a mobile/throwbox: For two common mobility model For general throwboxes spatial distribution
Explicit expressions for the distribution and the mean of delivery delay and number of generated copies Under epidemic and MTR protocols Asymptotic expressions for these means under
MTR
51
Conclusions (2)
Proposed various routing strategies for DTNs augmented with throwboxes
Markovian framework to evaluate performance of various routing strategies Can be extended to evaluate other
performance metrics and routing techniques
Explicit expression for the diameter of forwarding path under epidemic protocol
52
Conclusions (3)
Proposed an efficient MAC protocol for IEEE 802.11 Adapt starting value of contention window
to network size Original mechanism to estimate number of
active stations
53
Future research direction
Analyze correlation and heterogeneous movement patterns in real mobility traces
Elaborate corresponding mobility models and evaluate proposed routing strategies over them e.g. markovian model for community based
mobility, bus mobility
Analyze impact of different buffer management techniques on routing under heterogeneous mobility model
54
The end …
Thank you!
55
Publications M. Ibrahim, A. Al Hanbali, P. Nain, "Delay and Resource
Analysis in MANETs in Presence of Throwboxes", Performance Evaluation, Vol. 64, Issues 9-12, P. 933-947, October 2007.
Al Hanbali, M. Ibrahim, V. Simon, E. Varga, I. Carreras "A Survey of Message Diffusion Protocols in Mobile Ad Hoc Networks", Inter-Perf 2008, Athens, Greece, Octobre 2008.
M. Ibrahim, S. Alouf, "Design and Analysis of an Adaptive Backoff Algorithm for IEEE 802.11 DCF mechanism", Networking 2006, Coimbra, Portugal, Mai 2006.
Under submission: M. Ibrahim, P. Nain, I. Carreras. "On routing trade-offs in throwbox-embedded DTN networks".
56
57
ZebraNet: mobility based routing
Objective track zebras in wildlife Collars attached to zebras Base stations move sporadically to collect data
58
Model validation: Delivery delay
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
Throwboxes disconnected and uniformly distributedThrowboxes disconnected and RWP stationary distributedThrowboxes connected and uniformly distributedThrowboxes connected and RWP stationary distributed
MTRMTR EpidemicEpidemic
59
Virtual transmission time
Virtual transmission time = time separating two successful random transmissions