sterbenz, et al. ittc cross-layering and metrics resilinets and resumenet 07 october 2009 james p.g....
Post on 21-Dec-2015
218 views
TRANSCRIPT
Sterbenz, et al. ITTCCross-Layering and Metrics
ResiliNets and ResumeNet
07 October 2009
James P.G. Sterbenz*†
Джеймс Ф.Г. Стербэнз 제임스 스털벤츠 司徒傑莫David Hutchison,
Abdul Jabbar, Radovan BrunčákJustin P. Rohrer, Egemen Çetinkaya
*Department of Electrical Engineering & Computer Science
Information Technology & Telecommunications Research Center
The University of Kansas†Computing Department, Infolab 21
Lancaster University
http://www.ittc.ku.edu/~jpgshttp://wiki.ittc.ku.edu/resilinets
© 2009 Sterbenz
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 2
Cross-Layering and MetricsAbstract
Cross-layer theory: mechanisms and algorithms to control state transactions among several state machines within required level
Metrics: a notation to describe state transactions of particular state machine
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 3
ResiliNetsMotivation for Cross-Layering
• D2R2 strategy motivates…• Principles for resilient networking, e.g.
– Adaptability– State management– Multilevel– Context awareness– Autonomic (self-optimization)
• Requires cross-layering• Theory, algorithms, mechanisms and
methodologies to support resilience
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 4
Knobs and DialsFormalism
• Set of all knobs = K k– union of out-of-band K and in-band k
• Set of all dials = D d– union of out-of-band D and in-band d
• Set of knobs or dials between layers Li and Lj
– where i and j are either numbers, e.g. {1,1.5,2,2.5,3,4,7}or designators, e.g. {HBH, net, PoMo, E2E, app, …}
– Kij = KK Kij
– vertical when i≠j; horizontal when i=j
• Individual knob or dial between layers Li and Lj
– Kij(d) where d is a descriptor, e.g. BER
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 5
Knobs and DialsFormalism
• A protocol instance has– state at time t of s(t)
– context at time t of cn(t)
• At Ln, state is– s(t+1) = f (n+1n, nn–1, s(t), cn)
SM
kn+1n
Ln
Kn+1n
dn+1n
Dn+1n
knn–1 dnn–1
Knn–1 Dnn–1
K,Dnn
k,dnn
cn
s(t)
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 6
Cross-Layer ModelProtocol Instance Model
• Node state– state machine– memory
• Node inputs/outputs– vertical data– p2p virtual data– context– horizontal signalling– cross-layer signalling
SM
E2EE/H payloadHBH
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 7
ResiliNetsCross-Layer Model
• E2E and HBH layers
HBH HBH HBHHBH HBH
E2E E2E
E2EE/H payloadHBH
E2Econtext
HBHcontext
E2Econtext
HBHcontext
physicalenvironmentdials/knobs
applicationknobs/dials
cross-layerin-band
signallinghorizontal
explicit signalling
verticalexplicit signalling
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 8
End-to-End Communication Knobs and Dials
• Knobs and dials between upper layers and PoMo– support heterogeneous subnetworks
• e.g. lossy wireless vs. reliable wired
– explicit signalling of path diversity and multipath• geographic location of realms, nodes, channels
Knobs Dials Layer
application
E2E transport
PoMo internetwork
network realm
HBH link
service characteristics
path char., geography
realm characteristics
link characteristics
service classreliability mode
PoMo knobs, FD, motiv.
realm oper. parameters
link type and codingerror control type/strength
Sterbenz, et al. ITTC
Goal
• Make progress in understanding fundamental principles of cross-layering
• Deliver cross-layer theory– Cross-layer formalisms
• State space, variables, operations, transitions– Cross-layering calculus
• Cost and benefit calculation• Theory to compare cross-layer mechanisms, reason about
costs and benefits• Mechanisms build up on standard building blocks (in-baud vs.
out-of-baud, open vs. control loop, E2E vs. HBH) – Cross-layer evaluation model
• Dynamics of the cross-layer system, avoid instabilities, oscillations, etc.
07 October 2009 9
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 10
Case study: error control
• Functional alternativesN noneO open loop (FEC)C closed loop (ARQ)
• S&W, GB-N, SelRep
• Location– HBH– E2E
• App requirements– unreliable– quasi-reliable– reliable
None
OFEC
CARQ
NoneO
FECC
ARQ
E2E
HBH
Sterbenz, et al. ITTC
Approaching problem
• Cross-layer control: between E2E and HBH error control • One hop
– Select error control (ARQ,FEC,NONE) – Setup strength– Local decision only– Inputs: state, context, knob, dials
• Hop-by-Hop– Calculating dials
• End-to-End– Selecting error control and setup strength– Calculating knobs
• Cost contrasting methodology07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 11
Sterbenz, et al. ITTC
Selecting ER control: Reasoning about costs
• Variables: Capacity, Speed, BER, BER interval, dynamics of BER
• Capacity: High => FECLow => ARQ
• Latency: High => ARQLow => FEC
• BER: HIGH => FEC, ARQLOW => FEC, ARQ
• BER interval: HIGH => FEC, ARQLOW => FEC, ARQ
• BER dynamics: HIGH =>ARQLOW => FEC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 12
€
cos tFEC = cos ti
i=1
n
∑
cos tARQ = cos ti
i=1
n
∑
Sterbenz, et al. ITTC
Cost-Contrasting Methodology
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 13
BER interval
BER dynamics
HIGHLOW
ARQFEC
BER
Sterbenz, et al. ITTC
Selecting ER control: Reasoning about benefits
• Guarantee: – HIGH because there is feedback => ARQ – LOW because there is not feedback => FEC
However, it can be HIGH, if estimation of the linkcharacteristics is good. However we need to deal withProbabilities
• Latency– Global remedy (E2E) is slow– Local remedy (HBH) is quick
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 14
Sterbenz, et al. ITTC
Setup strength
p – error rate, proportional to BERs – secure constant – depends on state, context (can go even to negative
numbers)m – mode, (unreliable ->0, quasi-reliable -> 0.5, reliable -> 1)?
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 15
€
strenghtmechanism = fmechanism = f (knob,state,context,dial) = m(pΔber
Δt+ s)
Sterbenz, et al. ITTC
Dial calculation: possible approaches
• Max BER value• Significant BER change • Average BER value• Number of unmanaged packets
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 16
Sterbenz, et al. ITTC
Knob calculation: possible approaches
• Knob– Reliable– Quasi-reliable– Unreliable
• Calculation?
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 17
Sterbenz, et al. ITTC
Dynamics of cross-layer system
• Knob can be change changed in real-time• How to describe the dynamics of the underline
system? • How can we describe the state transitions?• How can control theory be applied?
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 18
Sterbenz, et al. ITTC
Control theory in cross-layering?
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 19
Sterbenz, et al. ITTC
Control theory in cross-layering?
• Dynamics of the cross-layer system?• How knobs can change state of the underline
system ?
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 20
Sterbenz, et al. ITTC
Cross-layering calculus
• Decomposition?• Towards a cross-layer calculus:
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 21
€
min f (si − oi)i=1
d
∑
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 22
Resilience and HeterogeneityEvaluation Methodology: Simulation
• Resilience strategy and principles• Postmodern Internet Heterogeneity• Example realms
– WDTN– highly mobile airborne ad-hoc networking
• Evaluation methodology– simulation– experimentation
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 23
Evaluation MethodologyFlexible and Realistic Topology
Generation • Hierarchical topology generation
– evaluation of PoMo mechanisms– network engineering for resilience
• Level 1: backbone realms– nodes distributed based on location constraints– links generated using various models under cost constraints
• Level 2: access network realms– distributed around backbone nodes– access network connectivity: ring, star, mesh
• Level 3: subscribers– distributed around access network node
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 24
Evaluation MethodologyChallenge Simulation Module
• Separate challenge from network simulation• Simulate challenges to any network over time interval
– natural disaster destroy network infrastructure (e.g. Katrina)or large-scale grid failure (e.g. Northeast US 2003)
– attack: {node|link} down, wireless link attenuated
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 25
Evaulation MethodologyExample: Sprint Actual Topology
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 26
Evaluation MethodologyExample: Sprint Synthetic Fragile
Topology
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 27
Evaluation MethodologyChallenge Simulation Module
# of nodes down
agg
reg
ate
pa
cke
t de
live
ry r
atio
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 28
Evaluation MethodologyChallenge Simulation Module
• KU-CSM Challenge Simulation Module– challenge specification describes challenge scenario– network coordinates provide node geo-locations– adjacency matrix specifies link connectivity– input to conventional ns-3 simulation run– generates trace to plot results
adjacencymatrix.txt 0 0 1 0
simulation description.cc
challenge specification.txt
node coordinates.txt
ns-3 simulator
(C++)
plotted trace
KU-LoCGen
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 29
Resilience MeasureTowards a Resilience Metric
• Need: analyse and understand level of resilience• Problem: how to measure
– ideal: = [0,1]– but too many metrics to combine into a single number
• perhaps a small set of objective functions
– need to separate service spec. from operational parms.
• Model resilience as two-dimensional state space– operational state– service level
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 30
Resilience MetricsMultilevel Approach
• Resilience defined at any layer boundary– operational state describes system below layer boundary– service states represents behaviour above boundary
• Operational range arbitrarily divided into 3 regions– normal operation– partially degraded– severely degraded
• Service delivered arbitrarily divided into 3 regions – acceptable service– impaired service– unacceptable service
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 31
Resilience MetricsStates
• State of the system represented by (N, P) tuple– multivariate operational state N– multivariate service state P
• Initial state described by– acceptable service– during normal operations
• Challenges perturb N and P– resilience trajectory
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 32
Resilience State SpaceOperational Resilience
NormalOperation
PartiallyDegraded
SeverelyDegraded
Operational State N
S
• Operationalresilience– minimal
degradation– in the face of
challenges
• Resilience state– remains in
normal operation
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 33
Resilience State SpaceService Resilience
NormalOperation
PartiallyDegraded
SeverelyDegraded
Acceptable
Impaired
Unacceptable
Operational State N
Ser
vice
Par
amet
ers
P
S
• Serviceresilience– acceptable
service– in the face of
degraded operation
• Resilience state– remains in
acceptable service
S
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 34
Resilience State SpaceResilience Trajectories
• Choose scenario– network– application
• Metrics– choose– aggregate
• Observe– under
challenge
Unacceptable
Impaired
Acceptable
P
[del
ay, p
kt d
eliv
ery
rati
o]
Normal operation
Partially degraded
Severely degraded
S5
S1
S4
p1 > 10 secp2 > 0.85
p1 < 10 secp2 > 0.50
p1 < 1 secp2 < 0.50
n1 ≤ 1.5ρ0 n2 ≥ 3.0
n1 ≤ 3ρ0 n2 ≥ 1.0
n1 > 3ρ0 n2 < 1.0
N [load, node degree]
S3
S2
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 35
Resilience EvaluationBasic Approach
• Define service boundary and metrics– boundary Bij between two adjacent layers Lij
– k operational metrics characterize network below Bij
– Nij = {N1, N2, … Nk}
– l service parameters define service across Bij
– Pij = {P1, P2, … Pk}
• Evaluate resilience– operational and service space divided in to regions
– resilience Rij at the boundary Bij is evaluated as ….
… the transition of the network through the state space
– Rij = f (Nij,Pij)
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 36
Resilience EvaluationMultilevel
• Mapping of service to operations – service parameters become operations at layer
above• Ni+1,j+1=Pij
– resilience can be evaluated at any arbitrary layer boundary
– resilience as seen by the application is at the L67 boundary
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 37
Resilience EvaluationTime Scale
• Steady state analysis– long term view of resilience– theoretical evaluation or simulation study to
understand• impact of perturbations on operations to service being
provided
– best, worst, and mean case type of studies are possible
• Transient analysis– based on instantaneous state of the network– plot service parameters vs. operational metrics in
real time– resilience is characterized by state transitions – shows the impact of resilience mechanisms
• defense, remediation, and recovery
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 38
Resilience EvaluationPreliminary Results
• Resilience study at boundary B23
– operational state of network is characterized by• vertices, edges, and edge failures
– service provided across the boundary is a connected graph
– steady state analysis based on simulations
• Sprint case study– operational metric:
• link failures f (fixed nodes and links)
– service parameters: (preliminary, proof of concept)• avg. node degree d
• relative size of largest connected component |cmax|– normalised to number of nodes in graph
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 39
Resilience EvaluationPreliminary Results: Sprint Case Study
• 1000 runs• Conventional plot
– limited to1 metric/axis
• Resiliencestate space– discrete
approach– map into
N, P regions
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 40
Resilience EvaluationPreliminary Results: Sprint Case Study
normall = 0
partially degraded 0 < l ≤ 5
severely degraded 5 < l ≤ 68
unacceptable d < 2 & |cmax| < 0.95
impaired 4 > d ≥ 2 & 1 > |cmax| ≥ 0.95
acceptable d ≥ 4 & |cmax| = 1
# of link cuts
aver
age
nod
e de
gre
e, c
ompo
nen
t s
ize
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 41
Resilience EvaluationMultilevel Operational vs. Service State
Operational State Ni
Ser
vice
Par
ms.
Pi+
1
Operational State Ni+1 Pi+1
Bi,i+1
Bi+1,i+2
Ser
vice
Par
ms.
Pi+
2
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 42
Resilience EvaluationMultilevel Operational vs. Service State
Layer Service Operational State
Application(7)
application behaviour E2E performance (delay, goodput, …)
E2E Transport(4)
E2E data transfer stable E2E path
Inter-realm path(3.5 PoMo)
E2E inter-realm path(routing and forwarding)
stable intra-realm path
Subnetwork path(3)
e2e intra-realm path(routing and forwarding)
topology
Topology connected graph #link cuts, nodes out
HBH(2)
stable linkHBH data transfer
channel conditions
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 43
Resilience EvaluationPreliminary Results: Sprint Case Study
• Conclusions– shows an example of resilience characterization at
L23
– best case scenario, the topology is highly resilient– worst case, single link failure results in unacceptable
service
• Future work– evaluate resilience R34 at boundary B34,
• d and |cmax| become the operational metrics at B34
– compare multiple topologies against each other• Sprint, GÉANT2, synthetic topologies generated with
KULocGen
– conduct transient analysis • see the impact of D2R2
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 44
End
Sterbenz, et al. ITTC
07 October 2009 Resilience, Survivability, Heterogeneity in Postmodern Internet 45
End-to-End Communication Knobs and Dials in PoMo Header
• In band: knobs and dials in data packet• Explicit signalling: knobs and dials in signalling packet
realm realm realmrealm
E2E E2E
realminstrumentation
dials/knobs
applicationknobs/dials
cross-layerin-band
signallinghorizontal
explicit signalling
verticalexplicit signalling