time and the the internet of things - stanford...
TRANSCRIPT
2016SCPNTSymposium
Time and the The Internet of Things
Bob Iannucci, Ph.D. Distinguished Service Professor, ECE Director, CyLab Mobility Research Center http://sv.cmu.edu/bob
November 3, 2016
1
Key Points
• TheIoTistheFi7hpla:orm• Itisinherentlycyber-physical• Itwillbedistributed,federated,andenergy-poor• Timewillbeafirst-orderconceptinprogrammingit• TradiFonalFme-keepingapproacheswon’twork• TimeislandsmayofferaworkableabstracFon
2016SCPNTSymposium2
What is the IoT?
• internetofTHINGS
• WiFi/Bluetoothgadgets• Cloud-connectedstuff• Phone-as-mediator• Fitbit,Nest,…• Nosignificantnetworkeffect
• INTERNETofthings
• CiFes,communiFes• Sharing• Openness• Programmability• Significantnetworkeffect
• Cyber-physical
2016SCPNTSymposium
These differing views lead to wildly different thoughts about what is important
3
IoT Programing Today
2016SCPNTSymposium4
Embedded programming • Fixed functionality • Focused on low-power operation • Local power OR recharged regularly
Cloud programming • Aggregation • Machine learning • Visualization
Effectively closed
Sensors in the Real World: the Out-of-Building Experience
• Aninsidethebox(building)IoTiseasybutlimiFng• Power,cableplant,WiFi,tamephysicalcondiFons
• Outsidethebox• Offersnoneofthese• ButisessenFalforlogisFcs,powersystems,fleets,smartcampuses,smartciFes
• Costofdeploymentandmaintenance• Self-sufficientdevices• 5-10yearlifeFme• “OHIO”
2016SCPNTSymposium5
An Hour in the Life of an OHIO Device: Energy is at a Premium
• “Perfect”1000mAhba\erywithnoenergyharvesFng• LP-WANradio• FiveyearlifeFme
2016SCPNTSymposium6
State Secondsperhour
Sensing&processing 10
Receiving 10
Transmi`ng 5
Shutoff 3575
But is the Ultimate Value in Devices?
Transducers Bo7om-of-PyramidDevices
MiddlewareandMiddleDevices
AggregaEon FederaEon AnalyEcs BusinessLogic
1-10trillion 100billion 1billion 10,000 10’s 1000’s millions
InnovaFvemanufacturing,longevity,electronicdatasheets,reliability,repeatability,lowpower
Lowpower,lowpower,lowpower(“OHIO”)
Remoteprogrammability,reliablewirelesscomms,easeofdeployment,authenFcaFon
Programmingtoolsforfleetdeploymentandmanagement,versioncontrol,provenancemanagement,generaFonalcontrol
Managementofenormousvolumesofmostlyuselessdata(needleinahaystack),abilitytosummarizeacrossFmeandspace
Managementofdissimilarstandards,privacy,security
“Marketplace”
ExtracFonofsignals(usefuldata)fromnoise,hierarchicalprocessing,real-Fmeresponse
TradiFonalservicesbusiness,tailoredtotheenterprise
2016SCPNTSymposium7
An Uncomfortable Transition to the Future State
• ValueisinaggregaFonofsensedinformaFon• Sensingmustbepervasiveandflexible• Pervasivesensingdependsonubiquitouscoverage• CreaFngcoverageisasubstanFalinvestment• InvestmentjusFfiedbyabroadbaseofapplicaFons• ThefirstwaveofIoTapplicaFonswillbeverFcal• TherealvaluewillcomewiththetransiFontohorizontally-organizedpla:orm
• Similartothepla:ormsthatprecededit
2016SCPNTSymposium8
Platfomization and Iannucci’s Law
Ineachgenera*onofcompu*ng,theemergenceofastandardpla*ormtransformstheindustrybyshi7ing
valuefromhardwaretoso7wareandservices.
2016SCPNTSymposium9
Computing Platforms: Past and Present
GoldenAgeStandardPlaHorm CompaEbles
ValueShiLtoSoLware
ValueShiLtoServices
Mainframe
Mini
PC
Mobile
Magnuson Computer
WhataboutthePlaHormfortheInternetofThings?
2016SCPNTSymposium
Mobile Apps plus cloud services
10
IoT: the Fifth Platform (P5): What to Expect?
P5willbeacombina*onof1. Flexible,powerful,programmable,opendevicefamilies2. Flexible,powerful,programmable,opennetworks
Observa1on:#1exists.#2doesnot
2016SCPNTSymposium11
Challenges for P5
• ConsiderableearlyIoTresearchfocusedoncreaFngconnecFvity• Meshnetworkingvs.devicepower
• LP-WANsshi7awayfrommeshestostars• SigFox,LoRa,IngenuRPMA
• Under-addressed:programmingmodel• Device-centric?No,distributed.
• Cloud-centric?No,toofaraway.
• Emerging:enablingthenetworksideoftheIoTplaHorm• CompuFngmovestominimizelatency
• Be\ermanagementoftheTimeLine
2016SCPNTSymposium12
IoT Programing Future?
2016SCPNTSymposium13
USING [ e23, e97 ]�DO [ b1 ]�SYNCHRONIZED TO c19 WITH TOLERANCE x ONERROR [ b2 ]
DecomposiFon
Implications • One program on multiple devices • Multiple programs on one device
Cloud has unlimited • Computing • Storage • Energy • Space
Network-as-participant “Near to Near” (vs. peer-to-peer)
Today: latency too high Future: embed virtualized, �movable, real-time �computing capability in the network
Cyber-physically enabled future network: Automated marshaling of computing
Phone, sensor, actuator layer: devices have limited
• Computing • Storage • Energy • Space
Physical World
Sense-Compute-Actuate�Latency
Sens
e Actuate
Compute
The Cyber-Physical Network
2016SCPNTSymposium14
The�Time�Line
The Time Line
• Belowtheline:Fmeisreal• Concurrency,simultaneity,causality
• Abovetheline:Fmeismeta-data• ProgrammercanmanipulatetherepresentaFonofFme
• Crossingcybertophysical?• Translateprogrammer’sintenttoreality–notalwayspossible
• Crossingphysicaltocyber?• Inadistributedworld,howdoweFme-labeleventsinawaythatiscorrect
enoughastonotgiverisetosubsequenterrors?
2016SCPNTSymposium15
Crossing the Line: Time Accuracy Requirements • Crowdsourcingdatafromsmartphones
• Seconds!milliseconds
• Sensorfusion:situatedandairbornesensors• Milliseconds!microseconds
2016SCPNTSymposium16
Networked UAV Sensors1
!
Antofagasta
Tocapilla
Iquique
Arica
Calama
BOLIVIA
CHILE
PERU
50 k
m
��Ý:
��Ý:
��Ý6
��Ý6
��Ý6 ~M
8.8 18
68
~M8.
8 18
77
M8.
2 20
14
0 20 40 60
depth (km)
0246
6OLS��P�
85Ý:
80Ý:
75Ý:
70Ý:
65Ý:
45Ý6
40Ý6
35Ý6
30Ý6
25Ý6
20Ý6
15Ý6
10Ý65Ý60Ý
5ÝN
Qui
to
Ant
ofag
asta
La P
az
Valp
arai
so
Sant
iago
Con
cepc
ion
Lim
a
Puer
to M
ontt
Mot
ion
ofN
azca
Pla
te
1979
M 8
.2
1958
M 7
.719
42 M
7.8
1960
M 7
.6
1996
M 7
.519
70 M
7.8
1940
M 8
.019
66 M
8.1
1974
M 8
.020
07 M
8.0
1942
M 8
.119
96 M
7.7
2001
M 8
.5
2007
M 7
.719
95 M
8.1
1922
M 8
.4
1943
M 8
.319
71 M
7.8
1985
M 8
.019
28 M
8.3
2010
M 8
.8
1960
M 9
.5
Nazca
Ridg
e
Chi
le R
ise
RSA
SENS
OR
FRXUWHV\�0��6
LPRQV
Figure!1.!A)!M
ap!of!South!
America!show
ing!historic!great!
earthquake!ruptures!(red!
ellipses).!!B)!Zoom!of!the!Arica!
Gap!region!showing!historic!
ruptures!(yellow!ellipses),!slip!
model!from
!M8.2!2014!Pisagua!
earthquake,!and!aftershocks.!The!
orange!squares!are!the!RSA!
Sensors!to!be!deployed!in!this!
proposal.!Depiction!of!actual!sites!
is!schematic,!total!num
ber!,!~250.!!
B!A!
Crowdsourcing using �Smartphones2
Networked Situated�Sensors
2Minson,S.E.,Brooks,B.A.,Glennie,C.L.,Murray,J.R.,Langbein,J.O.,Owen,S.E.,Heaton,T.,Iannucci,R.A.andHauser,D.L.
CrowdsourcedEarthquakeEarlyWarning.ScienceAdvances,1(3),1–7.h\p://advances.sciencemag.org/content/1/3/e1500036
1Coverage:Teng,E.,Falcao,J.D.,Dominguez,C.R.,Mokaya,F.,Zhang,P.,&Iannucci,B.AerialSensingandCharacteriza1onofThree-DimensionalRFFields.
InSecondInternaFonalWorkshoponRoboFcSensorNetworks.Sea\le,WA.2015.
Sensor-Side Timekeeping is Costly
• Coin-cell-powered,OHIO,with50¢billofmaterials• GPS
• Currenttech@1minuteon/hour:>5years!<2years• Cellularmodem
• Currenttech:no(protocoloverhead,power,cost)• NB-IOT:TBD
• CSAC• Currenttech:125mW,$1k• Future:TBD
• TCXO• CostofFmekeepinghastobesensor-appropriate• CostofFmekeepingisapplicaFon-dependent
2016SCPNTSymposium17
TickTalk: Time-Oriented Meta-Language for P5
• Concept:FmecomplexityofIoTprogrammingmustbereducedsothatnon-specialistprogrammerscancreateFme-correctIoTprogramsincluding• Cloudelements• Networkelements• Sensor/actuatorelements
• OneprogramperapplicaFoninsteadofthree• PrimiFvesforexpressingsyntonizaFon,synchronizaFonand
simultaneity• BoundedinFmeandspace–Emeislands
2016SCPNTSymposium18
Joint work with Aviral Srivastava, Arizona State U
TickTalk: Core
• Aneasily-programmedFederaFonofCyber-PhysicalSystems(FSCS)
• AnFSCSisatuple(C,E,B):• C={c1,c2,c3,…}referenceclocks• E={e1,e2,e3,...}compuFng/storage/communicaFonensembles.Eachhasatleastonelocalclockcapableofbeingsynchronizedtoareferenceclock.Eachmayhavesensorsoractuators.
• B={b1,b2,b3,...}codeblocks(programfragments)withinwhichacFonscanbescheduledagainstanensemble’slocalclock
2016SCPNTSymposium19
Joint work with Aviral Srivastava, Arizona State U
TickTalk: A Program Fragment
2016SCPNTSymposium20
USING [ e23, e97 ]�DO [ b1 ]�SYNCHRONIZED TO c19 WITH TOLERANCE x ONERROR [ b2 ]
Wherethecodeblock
runs
Thecodeblocktorun
Referenceclock Syncspec
Joint work with Aviral Srivastava, Arizona State U
TickTalk: Runtime
2016SCPNTSymposium21
Meta-Language
Intermediate Representation
Code Blocks
Reference Clocks
Cloud
Network
Device
USING [ e23, e97 ]!DO [ b3; SIMULTANEOUSLY [ a1 ] WITH TOLERANCE y ONERROR [ b5 ]; b4 ] SYNCHRONIZED TO c19 WITH TOLERANCE x ONERROR [ b2 ]
Run-TimeManager
Sensors and Actuators
Blockb3onensemblee23
Blockb3onensemblee97
Clockc19
Synchronize
Joint work with Aviral Srivastava, Arizona State U
The TickTalk Network Ensemble
22
LocalClock
Time-AwareCompu@ng
Network-EmbeddedEnsembleInstances
NetworkUplinkTraffic+PerformanceMetrics
Downlinktosensorsandactuators
Labe
ls
2016SCPNTSymposium
Joint work with Aviral Srivastava, Arizona State U
Summary
• TheIoTistheFi7hpla:orm(P5)• P5isinherentlycyber-physical• P5willbedistributed,federated,andenergy-poor• Timewillbeafirst-orderconceptinprogrammingP5• TradiFonalFme-keepingapproacheswon’twork• TimeislandsmayofferaworkableabstracFon
2016SCPNTSymposium23
Further Reading
• Minson,S.E.,Brooks,B.A.,Glennie,C.L.,Murray,J.R.,Langbein,J.O.,Owen,S.E.,Heaton,T,Iannucci,B.,andHauser,D.L.CrowdsourcedEarthquakeEarlyWarning.ScienceAdvances,1(3),1–7.2015.
• Teng,E.,Falcao,J.D.,Dominguez,C.R.,Mokaya,F.,Zhang,P.,andIannucci,B.AerialSensingandCharacteriza*onofThree-DimensionalRFFields.InSecondInterna*onalWorkshoponRobo*cSensorNetworks.Sea\le,WA.2015.
• Weiss,M.,Eidson,J.,Barry,C.,Broman,D.,Goldin,L.,Iannucci,B.,Lee,E.,andStanton,K.Time-AwareApplica*ons,Computers,andCommunica*onSystems(TAACCS)Time-AwareApplica*ons,Computers,.2015.
• Athreya,A.P.,Chan-Maestas,H.,Katz,E.,Tague,P.,&Iannucci,B.Energy-GovernedResilientNetworkedSystems.In11thAnnualIEEEConsumerCommunica*onsandNetworkingConference.LasVegas,NV:IEEE.2014.
• Mengshoel,O.J.,Iannucci,B.,&Ishihara,A.(2013).MobileCompuFng:ChallengesandOpportuniFesforAutonomyandFeedback.In8thInterna*onalWorkshoponFeedbackCompu*ng.SanJose,CA.2013.
• Zhang,J.,Iannucci,B.,Hennessy,M.,Gopal,K.,Xiao,S.,Kumar,S.,…Rowe,A.SensorDataasaService--AFederatedPlaYormforMobileData-centricServiceDevelopmentandSharing.2013IEEEInterna*onalConferenceonServicesCompu*ng,446–453.2013.
• Zhang,J.,Li,Z.,Sandoval,O.,Xin,N.,Ren,Y.,MarFn,R.A.,Iannucci,B.,…Rowe,A.SupporFngPersonalizableVirtualInternetofThings.In2013IEEE10thInterna*onalConferenceonUbiquitousIntelligence&Compu*ng(UIC2013).SorrentoPeninsula,Italy.2013.
2016SCPNTSymposium24