Resilient Space-based
Software Defined
Networks
27 June 2017
Integrity « Service « Excellence
James Lyke
Air Force Research Laboratory
([email protected])
FutureSpaceCommunications(FSC)
• Vision
• Give warfighters resilient and ambient connectivity
(enabled with space technology)
• HowCommunicationsShouldWork
• “Always on” – there when you need it
• “Path agnostic” – exploit any all links and pathways
• “Automagic” – sort connectivity autonomously for
human /machine clients, apps
• Secure – in a fluent, natural way – make and break
virtual networks on demand cryptographically solid
isolation (“polychromatic” – multi-level security) and
cyber-secure (minimize attack surfaces)
• Invisible – Low probability of intercept (LPI) –
elastically trade bandwidth and non-detection
Goal:Ambientconnectivity
2
DesiredUseCases
Missionwith
comm relay
Opportunistic
relay
Tacticalcomm
augmentation
payload
mission
data
ground(inter)networks
3
AchievingAmbientConnectivity
•Self-formingnetworks(topologyagnostic)
•Everyplatformcanbeacontributingnode
•Everyphysicallayer(rf,optical)canbeacontributing
link
•Everynodeispotentiallyanopportunisticrelay(with
configurable“optin/out”settings)
•Networksessionsarevirtualizedovertheactual
physicalnetworkon-demand
4
ATaleofTwoGraphModels
Scale-free
network
Random
network
Scale-freenetworks(e.g.Power-Lawdistribution)
• Fewsuper-bigpipes
• Lotsoflittlepipes
• Vulnerabletoattacksonbiggestpipes
Randomnetworks(e.g.Erdos-Renya)
• Egalitarian,uniformnode-degreehubs(mayrelaxthisconstraint)
• Noperseweakestlinks
• Nodesandedgesconnectedwithhighprobability(percolation)
Recommendedreading:A.Lazlo-Barabasi,Linked
5
Resilienceand“GNATS”
•Onepathtoachievingresilienceistoexploitthe
Erdos-Renya randomgraphbyenrichingspacewith
asmanyspacecraftaspossible(cubesats,bigsats,..)
•Theconceptofaflexible,self-hostingnetworkhub
witharbitrarymixtureofphysicallayersweterma
GlobalNetworkAccessTerminal(GNAT)
•AsatellitewithaGNATisaGNATsatellite(GNATS)
•Liketheannoyinginsects,“gnats”areeasytokill
individually,buthardasanensemble– resilience!
6
SoftwareDefinedNetworksin
ContestedEnvironments
•Buildinganetworkofconnectedlinksisonething,
harnessingiteffectivelyisanother
•Considerconceptsfromsoftware-definednetworks
(SDN)andnetworkfunctionalvirtualization(NFV)
withacoupleoftwists:
– Heterogeneousnodesandedges
– Designforresiliencytofailures
7
NetworkGraphs– SupplyandDemand
• Real-worldconnectivityis
asupplygraph
• Wecreateasetofvirtual
networksondemand
(demandgraph)
• They“seemreal”inevery
significantway
Real-worldnetwork
A
B
C
– HavedefinedQualityof
Service(QoS)
– Cryptographicallyisolated
(seemair-gapped)
G
D
F
E
• Theyworkonlysolongas
the“real-world”networks
haveadequate
connectivityandQoS
Virtualnetworks
C
A
G
F
E
A
B
Contestedenvironment
A
B
D
E
C
A
G
F
•
Oneormorelinks
Real-worldnetwork
mightbreakdueto
jamming,weather,
C
orviolationofQoS
(notenough
bandwidth)
G
• Example,welostSF
bandlinkandit
wastheonly
Virtualnetworks
connection,soall
VNsinvolvingthat
linkarebroken
E
A
B
9
WhatisDifferentinSDNsfor
ContestedEnvironments
• farmoredynamic:complexevolvinggeometricrelationships,
linksmaybeattackedorimpaired,nodesmaybedestroyed,
• disparate/heterogeneousnodeandlink(wired,RF,optical)
structure;
• flexibleanddistributedprovisioning(itmaybenecessaryto
allowhierarchicaldelegationofprovisioningandcontrol
policies,processes);
• diverseQoS withafocusonthebesteffortcanbeprovidedin
acontestedarea(fromdisruptiontoleranttoreal-time)
• QoE (QualityofExperience)fortheuser’smissionplanningin
acontestedarea;
• egalitarian(mostly)networknodes(notnecessarily
distinguishedmasters)
10
Contestedenvironmentexample2
Real-worldnetwork
A
B
C
G
D
F
E
Virtualnetworks
C
A
G
F
E
A
B
• Ka-bandbroke(A<->B)
buttheopticallink
didn’t
• Therefore,theVNscan
maintainconnectivity
• Callthisidea“SDNin
contested
environments”
• Seekthepurist
representationof
networksintermsof
flows,data(data
forwarding)plane,
controlplane,
managementplane
11
RevisitingOpenSystemsInterconnectModel
Application
Programming
Interface
ProvisioningFacility
Provisioning/
Managementplane
Application
Software-DefinedVirtualNetwork(SVDN)
Presentation
AppB
AppA
AppX
Session
SVDN
Transport
CP
NetworkLayer
Network
DataLink
Physical
CP
Waveform
Layer
CP
Waveform
Layer
DesiredUseCases
“SmartphonesforSpace”
GlobalNetworkAccessTerminal(GNAT)
* *
Missionwith
comm relay
Opportunistic
relay
Tacticalcomm
augmentation
payload
mission
data
ground(inter)networks
13
GNAT:
ASmartphoneforSpace
•
•
•
•
Wehavesmartphones
Spacecraftdonot
Thephonestheyhaveareverylimited
Wecanchangethat
14
ExampleRealWorldSmartphone- SamsungGalaxyS5
GNAT/Smart-PhoneAbstraction
(InformationConvergenceDevice)
Communication
modems(physical
layerdevices)
Router/Backplane
Informationprocessing
APossible“GNATRoadmap”
Cubesat /nanosat
VNX,half-heightVNX
Photonicbackplane
Conventionalspacecraft
OpenVPX/SpaceVPX
Gen0
Feature
BackPlane
Cube
I/O
Gen0
Custom/OEM
---100MbE,SpW
Gen1
Gen2
Gen1
SpaceVPX(SVPX)
"SpaceVNX"
1GbE,SRIO
Gen3
Gen2
SVPX
Half-heightVNX
1/10/40GbE
Gen3
Photonic/SVPX
Half-heightVNX
1GbE+optical
A testbed for GNATS
The “RESINATE” testbed at Kirtland AFB, NM
Dashboard
ControlPlane
* *
* *
Site2
Cosmiac
• Phy0
• Phy14400MHz
• Phy2optical
• Possiblesitefor
space-facing
optical*
• PossiblesiteforSbandspacefacing*
Site7
Site3
B595(GPS/AGT?)
• Phy0
• Phy14400MHz
• Phy2optical
Site4
Site5
Site6
SiteY1
Examplerandom
B425
instrumentat
• Phy0
randomlocation
• Phy1
4400MHz
• Control
center
• Possible
B472roof
crosslink
• Phy0
network
• Phy14400MHz
emulator
B570roof
• Phy0
• Phy14400MHz
• Phy2optical
Site1
Manzano
• Phy0
• Phy14400MHz
• Phy2optical
• Possiblesitefor
space-facing
optical
• PossiblesiteforSbandspace-facing
SandiaCrest
• Phy0
• Phy14400MHz
• Phy2optical
Phy0– Controlplane
Phy1– 4400/4500MHz
Phy2-- Optical
Phy3– S-band
Emulatednodes
Crosslinkinhouse networkfarm
18
Reconfigurability,AdaptiveHierarchy,andtheRoleof
Cognitivex
19
WhatareReconfigurableSystems
?
• Involvesanabilityto
alterstructureand/or
functionunder
softwarecontrol
• Software-defined
hardware
• Functional
reconfigurabilityvs.
Physical
reconfigurability
20
Whyhavereconfigurability?
• Flexibility
• x – ondemand(developmentspeed)
• Reduceinventory(replacen partswith1)
• Fieldupdates(features,bugfixes)
• Resiliency(workaroundfaults,self-heal)
• Adaptive,dynamicreconfiguration(time-share
silicon)
21
TheComputer…
22
TheFieldProgrammableGateArray
(FPGA)
Uncommitted
“blank” function
Design
capture
Synthesis
Program
Device
23
Past -1977
Today
• Ahandfuloflogical
“truthtables”
n
n
Billionsoftransistors
Millionsoflogical“truth
tables”
GeneralReconfigurableSystems
Programmablemechanisms,materials
Programmable
Analogarrays
thermal
Digitalsystems
Computers
FPGAs
Programmable
microwave
Programmablepathways
Analogsystems
configurablepower
25
ReconfigurableRF
Basedon
Physical
Alteration
26
ReconfigurableLaserCommunications
andPhotonicNetworks
FIXEDDIGITAL
LASERCOMM
TRANSCEIVER
RECONFIGURABLE
LASERCOMM
TRANSCEIVER
ANALOG
LASERCOM
TRANSCEIVER
RECONFIGURABLE
PHOTONIC
PROCESSOR
CHANNELPROCESSOROPTIONS
OPTICALTIME
TRANSFERMODULE
•
Core#1
Core#2
TXchannels
Core#n
Aperture/
Telescope
PHOTONICDATABACKPLANE
RXchannels
•
•
Work(underARAP)
towardsa“photonic
softwaredefinedradio”
withreconfigurable
networks
Agilelytunablebeam
andwavelength
Reconfigurable
waveform/protocol
ControlPlaneProcessor(managesindividualCOREsand
CORE-to-COREinteractions)
OpticalReconfigurableChannelProcessor(ORC-P)
27
AdaptiveHierarchy
Increasing adaptiveness
Autonomous Systems
Self-configuring
Systems
Configurable Systems
Programmable
Systems
Fixed Systems
28
29
AutonomyHierarchy
Increasing adaptiveness
Autonomous Systems
Self-configuring
Systems
Configurable Systems
Programmable
Systems
Fixed Systems
30
31
AutonomyHierarchy
Increasing adaptiveness
Autonomous Systems
?
Self-configuring
Systems
Configurable Systems
Programmable
Systems
Fixed Systems
32
CognitiveEngines
PossibleFrameworks
AdaptiveApproach
Cognitive
Engine
AdHoc/Intellectual
propertyblocks
(pre-determined
waveforms)
Apps/App
Frameworks
Factoryset
(not
adaptive)
PortabilityPOSSIBLE
Portability
generallynot
•
POSSIBLE
Manually
(humans
turnknobs)
•
Reconfigurability
Middleware(DDS,
CORBA,…)
REST
ApplicationProgramming
Interfaces(APIs)
Library
Tightlycoupled
Howknobsare
physicallyaccessed
Reconfigurablesystem(has
knobsthataresoft-defined)
•
Mechanically
/Manually
•
Mostsoftwareradios
aren’t(setatfactory)
Havesoftware,but
can’tchange
APIsestablishthe
centralmechanismfor
cognition
APIsarenon-unique
andnotmutually
exclusive(RESTcancoexistwithlibrariesand
manuallyturned
knobs)
33
Distributedcognitiveengine
Possible
provisioning
system
Cognitive
Engine
– Apoint-to-pointlinkmightapply
cognitivemorphing(frequency
hopping)
– Asetofnetworknodesmightapply
cognitivenetwork
Cognitive
Engine
Cognitive
Engine
Cognitive
Engine
Cognitive
Engine
Cognitive
Engine
• Cognitivecommunicationsisa
cross-layerproblem
Cooperative
Morphing
• Assuch,wemustconsidernot
onlyhowtodesignportable
cognitiveenginesbuthowto
– Maketheminteroperable
– Allowelasticitytopush-down
(delegate)ruleswhenpossible
34
CooperativeMorphingConcept
Hypothesis 1
{λ1,α1}
Hypothesis 2
ex.max(BW)
experimentsubjectto
parametricwaveform
variation
(0<r<1)
{λ2,α2}
Hypothesis 3
{λ3,α3}
•Promotesthecreationofsingle-purpose(disposable)
waveforms
35
Summary
•Ambientconnectivityistheaspirationofcommunicationsresearch
•Patternsofcommunicationsshouldbeondemand
•“Software-definednetworksincontestedenvironments”isthevon
Neumann-esque questforreliablenetworkswithunreliablelinks
andnodes(cyber-resilientwithpoly-chromaticsecurity)
•Weneedeveryplatformtohave/beasmartphone
– EveryplatformisaGNAT,someless/morecapablethanothers
– GNATsseekeachtobuildopportunisticnetworks
•BesidesbuildingtheGNAT-SDNinfrastructure,the“smartphones”
canbeextendedtoinformationconvergence,EMconvergencehubs
•GNATnetworksaregreattestbedsforcognitiveengines
– Reconfigurability andadaptivenessarenotthesameconcepts
– Considerationofcooperating/cross-layercognitiveengines
36