REALTIMECOMMUNICATIONINAN IP/ETHERNETBASED INCARNETWORK - - PDF document

• REALTIMECOMMUNICATIONINAN

IP/ETHERNETBASED INCARNETWORK

ForschungundTechnik.

OUTLINE

1. IntroductionandScope

Motivation VisionandChallengesforthefutureInCarnetwork EthernetinaVehicle ReverseEngineeringofcurrentbussystems:CANandFlexRay Methodology

• 2. Solutions1/2:StandardSwitchedEthernet

InfluenceoftheTopologies

HyungTaek Lim,BMWForschungundTechnik,30.11.2011 Seite2

InfluenceoftheTopologies NonPrioritized(1)vs.PrioritizedNetwork(2)

• 3. Solution3:IEEE802.1EthernetAVB

BackgroundInformation EvaluationofanEthernetAVBbasedInCarNetwork Summary

• 4. FutureWork

Publications

MOTIVATION

CAN CAN Ethernet CAN

Seite3

MOST

AutomotiveNetworkArchitecture[1]

[1]Freymann,R.:AnforderungenandasAutomobilderZukunft.The2ndMobilityForum,Munich,Germany

()heterogenous network differentprotocols,usedcomplexcentralizedapplicationgateway complexcableharness ⇒ Costs slowdowninnovations

ByteFlight

• FlexRay

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

VISIONAND CHALLENGES

InternetProtocol(IP)basedInCarNetwork

ReplacementofcurrentspecificInCarprotocolsbystandardIP

Seite4

• SuitabletechnologytotransportIPfortheautomotiveuse

Futureapplicationshavehigherbandwith demandduetotheincreasing numberofadvanceddriverassistance(ADADS)systemsandECUs

• WhichoftheexistentInCartechnologiesarecapabletotransportIPandcanfulfillthe

highbandwidthdemandoffutureapplications?

Ethernet??

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

LEGACYETHERNET

Applications

Ethernet MAC IP TCP / UDP

7:Application 6:Presentation 5:Session 4:Transport 3:Network 2:DataLink

? ? ?

(+)MatureTechnology (+)Fast,easytouse (+)Twowireunshieldedavailableforautomotive use (+)Nosinglesource ()Realtimedatatransmissionisnotsupported ()Framescanbedelayedorlost(Switch) ()Efficiencyproblemswithsmallpackets

Seite5

Ethernet Phy

2:DataLink 1:Physical

()Efficiencyproblemswithsmallpackets Ethernetiscurrentlyusedonlyfortwoareas:

• Diagnosisandflashing(OBD)
• Remotediscaccess(CICRSE)

CanEthernetalsobeusedforRealTimeinthecar?? Currently:norealtimeapplications;EthernetwithoutanyQoS mechanisms

Research Questions

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

REVERSEENGINEERINGOF CAN,FLEXRAY [*]

• TraceAnalysis:

AnalysisofcontrolmessagesbasedonrealInCarCANandFlexRaydataderivedfrom aBMWvehicle (1)Messagelengthandtheirdistribution (2)CycleTimes–Timebetweentwoconsecutivemessagesforcyclebasedmessages

CDF:Probabilityofusedmessagelength(CAN) CDF:Probabilityofusedmessagelength(FlexRay)

>60%

Seite6

ProbabilityP(x) ProbabilityP(x) x:MessageLength[Byte] x:MessageLength[Byte]

[*]WorkwaspresentedattheNets4Cars2011Conference,Oberpfaffenhofen,Germany

Mostoftheincarcontrolmessageshaveamessagelengthlessthan8Byte.AsingleUDPpacketwithaminimum payloadsizeof20byteswillcover95%oftheincarcontrolmessagelength.

≈ 77%of FlexRay messages have amessage length ≤8byte

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

REVERSEENGINEERINGOF CAN,FLEXRAY 2 [*]

(2)CycleTimes–Timebetweentwoconsecutivemessagesforcyclebasedmessages

P[X≤t] P[X≤t]

CDF:Probabilityofhavingcycletime(CDF) CDF:Probabilityofhavinginterarrivaltime(CDF)

Seite7 [*]WorkwaspresentedattheNets4Cars2011Conference,Oberpfaffenhofen,Germany

t:cycleTime[ms] t:Timebetweentwoconsecutivemessages[ms]

<cyclebasedmessages> <eventbasedmessages>

P[X P[X

ECUsusingcyclemessageswithlowcycletimesarepreferredbyFlexRay,while highcycletimesaremainlyusedbytheCANbus. Incaseofeventbasedmessages, CANbussystemarepreferredtouse.

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

RESEARCHWORKANDMETHODOLOGY

Threeessentialaspectsareconsidered inourwork: No. Ethernet Types Methodology Simulationbased Evaluation Prototyping Evaluation SwitchedEthernetwithout Prio.(‚Legacy‘) 1 Finished Finished

Seite8

DothedifferentEthernettypessupportrealtime communicationsandfulfillQoSrequirementsintermsof bandwidth andendtoenddelaybyagiventopology andapplications?

SwitchedEthernet withPrio 2 IEEE802.1 AudioVideoBridging (AVB) 3 Finished Finished Ongoing Ongoing

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

RESEARCHWORKANDMETHODOLOGY

Threeessentialaspectsareconsideredinourwork: No. Ethernet Types Methodology Simulationbased Evaluation Prototyping Evaluation SwitchedEthernetwithout Prio.(‚Legacy‘) 1 Finished Finished

Seite9

SwitchedEthernet withPrio 2 Finished Finished

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

INTRODUCTION:OMNET++

NetworkSimulationTool componentbased,modularandopenarchitecturediscreteeventnetworksimulator. Specificapplicationareasareimplementedbyvarioussimulationmodelsandframeworks, mostofthemopensource. OMNeT++withtheINETFramework

• FrameworkforwiredandwirelessTCP/IPbasedsimulations

(containsmostofthestandardprotocolsfromOSILayerL1– L7)

• somelimitationsformypurposes

Seite10

1.NoPrioritizationmechanism: datatrafficisnotclassifiedbydifferentpriorities. allapplicationsareconsideredasbesteffort theswitchesuseonlyasingleoutputqueueand aFirstInFirstOut(FIFO)scheduler 2. DataTrafficbasedonstatisticalmodels: packetsizeandsendingratearesetbystatistical distributionfunctions Prioritization as defined inIEEE802.1Q (MACLayerPrio.) Loading external Trace data Frameworkmodified [*]

[*]WorkwasrealizedincooperationwiththeBMWdevelopmentpart

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

SWITCHEDETHERNETBASEDINCARNETWORK[*]

Goaloftheanalysis: Influenceofthe (1)Topology (2) Prioritizationmechanisms (3)Linkload

Seite11

Starbased (Topology1) Daisychainbased (Topology2) Treebased (Topology3)

[*]Workwaspresentedatthe48thDesignAutomationConference(DAC2011)Conference,SanDiego HyungTaek Lim,BMWForschungundTechnik,30.11.2011

INCARAPPLICATIONS:TRAFFICCHARACTERISTICS

Traffic Type UDP PacketLength [Byte] Sending Rate[ms] Bandwidth [Mbit/s] Prio Max. EndtoEnd Delay[ms] Control 18 uniform (10,100) <1 3 10[1,2] Driver Assistance CAM 1472 0.5 24 2 45[2] Navigation 1400 0.7 16 1 100

Seite12

Navigation 1400 0.7 16 1 100 MMVideo 1400 0.28 40 150 MM Audio 1400 1.4 8 150 TVVideo 1400 uniform (0.56,1.12) 10– 20 150 TVAudio 1400 2.33 4.8 150

[1]R.Steffen,R.Bogenberger,M.Rahmani,J.Hillebrand,W.Hintermaier,andA.Winckler,,The FirstAnnualInternationalSymposiumonVehicularComputingSystems,Dublin,July 2008. [2]M.Rahmani,R.Steen,K.Tappayuthpijarn,G.Giordano,R.Bogenberger,and E.Steinbach, ! !,The4thInternationalTelecommunication NetworkingWorkShop onQoS inMultiserviceIPNetworks(QoSIP2008),Venice,Italy,Feb 2008.

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

PERFORMANCEEVALUATION[*]

SystemModel

• Assumptions

1) EthernetLinkBandwidth:100Mbit/s Ethernetisbasedonthe100BaseTXstandard 2) QualityofService(QoS)withPrioritizationmechanismasspecified inIEEE802.1Q(VLANtag) 4QueuesperoutputPort:

Seite13

4QueuesperoutputPort: Prio3(highestPriority):StrictPriorityScheduler Prio2..Prio1:WeightedFairQueuing(WFQ) 3) MACTransmissionQueueSize:100Frames 4) SwitchProcessingTime:3Ts[1]

• Metrics

EndtoEnddelay

[1] M. Rahmani, R. Steen, K. Tappayuthpijarn, G. Giordano, R. Bogenberger, and E. Steinbach, Performance Analysis of Different Network Topologies for In-Vehicle Audio and Video Communication, The 4th International Telecommunication Networking WorkShop on QoS in Multiservice IP Networks(QoS-IP 2008), Venice, Italy, Feb 2008.

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

RESULT:PERFORMANCEANALYSIS

(1)CDF:EndtoEnddelay Serviceconstraints: CTRL:Delay≤10ms CAM:Delay≤45ms

Top1(‚starbased‘) Top2(‚daisychain‘) Top3(‚Treebased‘)

Seite14

Top2(‚daisychain‘)

Prioritization reducestheendto enddelayofthehighestdata classtoapprox.40% IndependentoftheusedTopology, Prioritization reducestheendtoend delayofdriverassistancecamera datatoapprox.lessthan10%

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

SUMMARY:PERFORMANCEANALYSIS

• ataMACLayerasdefinedintheIEEE802.1p/qstandardcan

considerablyimprovetheperformanceintermsofthe

• Applicationconstraintsoftheincarapplicationsarefulfilled
• The hasthebestperformanceintermsoftheminimum

endtoenddelay.

Seite15

• ArethereanymechanismsatLayer2tosupportadeterministicbehavior of

applicationsinaswitchedEthernetnetwork? framesshouldarrivedatthedestinationwithinacertaintime highsynchronizationaccuracy lowjitter IEEE802.1Audio/VideoBridging(AVB)standard

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

IEEE802.1ETHERNETAUDIO/VIDEOBRIDGING (AVB)

IEEE802.1AVB: Specificationsthatwillallowtimesynchronizedlowlatencystreamingservices andQoS through802networks(SwitchedEthernet,WLAN) MechanismsonLayer2(MACLayer) TheStandardguarantees

• Maximumlatency of2msover7Hops(ClassA)or50ms(ClassB)
• Synchronizationaccuracyoflessthan1usover7Hops

Seite16

• Synchronizationaccuracyoflessthan1usover7Hops

consistsofseveralsubstandardswhich arecurrentlyratifiedandpublished AVBProtocols

• Max.75%ofBandwidthforAVB
• Min.25%ofBandwidthforlegacy

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

BACKGROUNDINFORMATION:ETHERNETAVB1

IEEE802.1AS– TimeSynchronizationProtocol

SynchronizationofdistributednodesinaswitchedEthernetnetwork toachievetwogoals: 1. CommonTimeBasis/ReferenceClock

• Synchronization ofdistributed,networked ECUs

(Audio/Video:LipSync)

• Coordination multipleECUs

Seite17

2. Meets jitter requirements

• Guarantees timely execution
• Administrationfree protocol

SyncProcessisexecutedbytwosteps: (1) Selectionofthebestmasterclock(BMC)inanetwork (2) StartofthesynchronizationbytheBMC(grandmaster) Measurementonlybetweentwoadjacentsystems ClockdriftsbetweenAVBsystemsandthegrandmasteraredetermined

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

BACKGROUNDINFORMATION:ETHERNETAVB2

IEEE802.1Qat– StreamReservationProtocol

Signalingprotocoltoreservetherequiredbandwidthforaspecificstream(AVB Stream)overthenetwork Signalingprocessisexecutedbythreesteps: 1. StreamAnnouncementbyasourcenode(‘Talker’) 2. StreamRegistrationbyasinknode(‘Listener’) 3. StreamDeregistration

• Initiationbyasourcenode
• Initiationbyasinknode

Seite18

• Initiationbyasinknode

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

BACKGROUNDINFORMATION:ETHERNETAVB3

IEEE802.1Qav– QueuingandForwardingRules

Mechanisms for switches to guarantee timesensitivedatatransmission interms of ," and requirements Based onthe IEEE802.1Qstandard which allows aseperation of the network traffic into differentclasses by proritization mechanisms Following two mechanisms are specifed: 1) Mappingofthe IEEE802.1Qpriority values to AVB

Seite19

1) Mappingofthe IEEE802.1Qpriority values to AVB AVBFrames (ClassA;ClassB) AVBframeshavealwaysthe highestpriorityvalue

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

BACKGROUNDINFORMATION:ETHERNETAVB4

IEEE802.1Qav– QueuingandForwardingRules

2) QueuingAlgorithms for AVB and nonAVBframe TwodifferentschedulingalgorithmstotransmitAVBandnonAVB (legacyEthernet)framesinanetwork a) StrictPriorityAlgorithm forLegacyEthernet b) CreditBasedShaperAlgorithm forAVB

• EachAVBClass(AorB)hascertaincredits

Seite20

• EachAVBClass(AorB)hascertaincredits
• ATransmissionisonlyallowedwhenacreditis>=0
• Foreachtransmissionthecreditisdecreasedatarateof#

Otherwise: creditisincreasedatarateof# TransmissionoflegacyEthernetframes

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

RESEARCHWORK

Threeessentialaspectsareconsideredinourwork: No. Ethernet Types Methodology Simulationbased Evaluation Prototyping Evaluation

Seite21

3 IEEE802.1 AudioVideoBridging (AVB) Ongoing Ongoing

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

EVALUATIONOF ETHERNETAVB[*]

ResearchQuestions:

1. Isthe IEEE802.1EthernetAVBstandard able to guarantee the ," and requirement of streaming data even inhigh load situations ? 2. Docontrol datatransmitted with legacy Ethernetframes meet the application constraints in terms of the $ and ?

SystemModel

Control

• Topology:daisychain with 3switches inanetwork

Seite22 [*]Workissubmittedatthe75thIEEEVTC2012Spring Conference,Yokohama,Japan

DA_CAM MM Disc HU Legend: ECU(EndNode) Switch 100Mbit/sLink Switch2 Switch3 Switch1 RSE/ Navi. CT1 CT2 Amp TV CTRL PU

• Applications

1)DriverAssistanceCamera Data(AVB:ClassA) 2)TVStreamingData(AVB:ClassB) 3)MMStreamingData(AVB:ClassB) 4)Control Data 5)NavigationData 6)BulkData

• Metrics:

MaximumLatency Jitter FrameLossRate

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

EVALUATIONOF ETHERNETAVB– SYSTEMMODEL

TrafficCharacteristics Assumptions

• EthernetLink:100Mbit/s
• Static clock drifts ofthe ECUswithout BMCA

Grandmaster(node with best clock):HeadUnit Each nodes have static clock drifts

• Six DifferentTrafficClasses

Six Queuesperoutput port (2AVB+4LegacyEthernet)

Seite23

TrafficCharacteristics

TrafficType Ethernet Payload [Byte] Sending Rate [ms] Bandwidth [Mbit/s] Prio DriverAssistance CAM 390 0.125 27.6 ClassA (Prio5) TVStreaming 700 0.250 23.7 ClassB (Prio4) MMStreaming 700 0.125 47.5 Control 46 uniform (10,100) 70 e3 ..700e3 3 Bulk 1428 0..0.168 [0,25,50,70] 1 Navigation 1000 5 1.667 2

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

EVALUATIONOF ETHERNETAVB– RESULT

Metric ControlData (Control→ CTRL_PU) AVB ClassA (DA_CAM→HU) Add.Load 0% 25% 50% 70% 0% 25% 50% 70% Max Latency [ms]

0.095 0.434 0.391 22.40 2.0003 2.0004 2.0003 2.0003

MaxJitter [ms]

0.059 0.372 0.285 12.39 0.0003 0.0004 0.0003 0.0004

Frame Loss Rate

0.584

ControlandAVBdatadependingonthebackgroundLoad

Seite24

Rate

AdditionalBackgroundLoad[%] Max.Latency of AVBData Max.EndtoEnddelay[ms] AdditionalBackgroundLoad[%] Max.LatencyofControlData Max.EndtoEnddelay[ms]

58%for our requirement Requirement given by Toyota 100us[*]

[*]RequirementsforAutomotiveAVBSystemProfiles,WhitePaper,Toyota,2011 HyungTaek Lim,BMWForschungundTechnik,30.11.2011

EVALUATIONOF ETHERNETAVB– SUMMARY

ThesimulationbasedperformanceEvaluationoftheAVBprotocolsshows: (1)ThelatencyandframelossofAVBstreamingdataareindependent of thenetworkload. AllapplicationsmodeledasAVBClassA/Bframeshavethelatencyless

• requalthanspecifiedvalues(2msforClassA;50msforClassB)

(2)TheperformanceofcontroldatainanAVBnetworkdepends strongly

• nthenetworkload.

Seite25

Inordertoguaranteethehardlatencyrequirementof100us,theadditional backgroundloadshouldbelessorequalthan15%.

• TheIEEE802.1EthernetAVBstandardimprovesdefinitelytheperformancesof

multimediaandapplicationsfordriverassistancepurposes. Adeterministicbehavioroftheseapplicationsareachieved.

• Inordertofulfillthehardrealtimerequirementsofcontroldataadditional

schedulingandprioritizationmechanismsarerequired.

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

PUBLICATIONS– 2011

– HyungTaekLim,KayWeckemann,DanielHerrscher: ##%&. SpringerLectureNotesonComputerScience,Proc.ofthe3rdInternationalWorkshoponCommunicationTechnologiesforVehicles, March2011,Oberpfaffenhofen,Germany. – HyungTaekLim,LarsVölker,DanielHerrscher: '(% .Proc.ofDAC2011, The48thDesignAutomationConference(DAC)2011,June2011,SanDiego,USA. – KayWeckemann;HyungTaekLim,DanielHerrscher: %$ ). Proc.oftheFifthInternationalConferenceonCOMmunication SystemsoftWAre andmiddlewaRE (COMSWARE),July2011,Verona, Italy.

Published

Seite26

– HyungTaekLim,BenjaminKrebs,LarsVölker,PeterZahrer: (%, Proc.ofthe36thIEEEConferenceonLocalComputerNetworks(LCN)*October2011,Bonn,Germany. – HyungTaekLim,DanielHerrscher,LarsVölker,MartinJohannesWaltl: IEEE802.1ASTimeSynchronizationinaswitchedEthernetbasedInCarNetwork, Proc.ofthe3rdIEEEVehicularNetworkingConference(VNC)2011,November2011,Amsterdam,TheNetherlands. – Hyung-Taek Lim, Daniel Herrscher, Firas Chaari IEEE 802.1 Ethernet Audio/Video Bridging in an In-Car Network,

• Proc. of the 75th IEEE Vehicular Technology Conference: VTC2012-Spring, Yokohama, Japan.
• Hyung-Taek Lim, Daniel Herrscher, Martin Johannes Waltl, Firas Chaari

Performance Analysis of the IEEE 802.1 Ethernet Audio/Video Bridging Standard, SimuTools 2012, The 5th International ICST Conference on Simulation Tools and Techniques, Sirmione- Desenzano, Italy

Submitted

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

Questions ? Any Comments?

Seite27

Any Comments?

thankyou.

Researchand Technology.

HyungTaek Lim,BMWForschungundTechnik,30.11.2011

INNOVATIONENGESTALTEN STEIGENSIEEIN

Seite28

• MöglichkeitenzurMitarbeitbeiBMWForschungundTechnik
• Praktikum(nachBedarf)
• Diplom/Masterarbeit(6Monate)
• Dissertation(3Jahre)
• ArbeitvorOrtbeiBMW,wissenschaftlicheBetreuungdurchUni
• BevorzugtInformatiker,ElektrotechnikeroderverwandteStudiengänge

EMail: hyungtaek.lim@bmw.de

Thank You

HyungTaek Lim,BMWForschungundTechnik,05.12.2011 Seite29

You

Researchand Technology.