Tim iming analysis of f AVB Ethernet network usin ing the Forward - - PowerPoint PPT Presentation

▶
tim iming analysis of f avb ethernet network usin ing the
SMART_READER_LITE
LIVE PREVIEW

Tim iming analysis of f AVB Ethernet network usin ing the Forward - - PowerPoint PPT Presentation

Feb 04, 2024 129 likes •568 views

Tim iming analysis of f AVB Ethernet network usin ing the Forward end-to to-end Dela lay Analysis Nassima BENAMMAR Frdric Ridouard Henri Bauer Pascal Richard Oc October 12, 12, 2018 2018 1 Context A UTOMOTIVE S YSTEM : ECU S ECU

slide-1
SLIDE 1

Tim iming analysis of f AVB Ethernet network usin ing the Forward end-to to-end Dela lay Analysis

Nassima BENAMMAR

Frédéric Ridouard Henri Bauer Pascal Richard

Oc October 12, 12, 2018 2018

1

slide-2
SLIDE 2

AUTOMOTIVE SYSTEM: ECUS

Context

2

ECU : Electronic Control Unit

slide-3
SLIDE 3

EXPLOSION OF AUTOMOTIVE SYSTEMS

Context

3

Passive autonomous driving Limited driver substitution Complete autonomous capability 2012 2017 2016 2018 2019 2023 100% autonomous cars for public http://www.caam.org.cn/english/

Critical real time systems

slide-4
SLIDE 4

4

DISTRIBUTED ARCHITECTURE

Context

slide-5
SLIDE 5

5

REAL TIME NETWORKS IN DISTRIBUTED ARCHITECTURE

Context

 Deterministic ;  Security, fault tolerance ;  Bandwidth;  Easy integration;  Decrease wiring.

slide-6
SLIDE 6

AUTOMOTIVE NETWORKING

Context

6

Heterogeneous networking Trends to Ethernet based standards Audio Video Bridging Switched Ethernet (AVB) Time Sensitive Networking (TSN)

Complexity

slide-7
SLIDE 7

7

AUDIO VIDEO BRIDGING SWITCHED ETHERNET (AVB)

Context

Credit Based Shaper (CBS)

SR Class A SR Class B Best Effort FPFIFO

  • Synchronization ;
  • Quality of service ;
  • SR : Stream reservation.
slide-8
SLIDE 8

REAL TIME SYSTEMS ANALYSIS

Problematic

8

Traffic shaper End-to-end delay Real time constraints

slide-9
SLIDE 9

Deadline

TIMING ANALYSIS

Problematic

9

Minimum delay Maximum delay by simulation Exact worst case delay Maximum jitter

Distribution of end-to-end delay

Mo Model checkin ing

Worst case delay computed by timing analysis

Tim iming anal analysis base based on

  • n

mathematics

slide-10
SLIDE 10

Motivations

10

FIFO FP/FIFO CBS NC TA CPA FA

AVB

  • Network Calculus (NC)
  • Trajectory approach (TA)
  • Compositionnal Performance Analysis

(CPA)

  • Forward end-to-end delay Analysis (FA)

Existing Our contribution

slide-11
SLIDE 11

Summary ry

11

  • FA METHOD
  • FA EXTENSION TO CBS ALGORITHM
  • EXPERIMENTATIONS
  • CONCLUSION
slide-12
SLIDE 12

FA method

12

slide-13
SLIDE 13

13

NETWORK MODEL OF FA

FA method FA extension to CBS algorithm Experimentation Conclusion node 1 node 2 node 4

Node = multiplexing point

node 3

slide-14
SLIDE 14

14

END-TO-END DELAY ELEMENTS

. . . . . . . . . . . .

Source node Variable delay L L L: technological latency Variable delay Variable delay FA method FA extension to CBS algorithm Experimentation Conclusion

slide-15
SLIDE 15

15

PROPERTIES A flow 𝑤𝑗 is defined by :

  • 𝐷𝑗

ℎ : maximum transmission time of a frame from 𝑤𝑗 in the node ℎ.

  • 𝑈

𝑗 : period of 𝑤𝑗.

  • 𝑄𝑏𝑢ℎ𝑗 : path of 𝑤𝑗.
  • 𝑄𝑗: priority level of 𝑤𝑗.
  • 𝑡𝑞𝑗

ℎ: set of flows 𝑤𝑘 ∈ Γℎ où 𝑄𝑗 = 𝑄 𝑘.

  • ℎ𝑞𝑗

ℎ: set of flows 𝑤𝑘 ∈ Γℎ où 𝑄𝑗 < 𝑄 𝑘.

  • 𝑚𝑞𝑗

ℎ: set of flows 𝑤𝑘 ∈ Γℎ où 𝑄𝑗 > 𝑄 𝑘.

….

FPFIFO priority 1 priority 2 priority n Fixed Priority First In First Out (FPFIFO) FA method FA extension to CBS algorithm Experimentation Conclusion

slide-16
SLIDE 16

16

ANALYSIS OF A FRAME WITH FA

𝑔

𝑗

Generation instant time of 𝑔

𝑗

𝑔𝑗𝑠𝑡𝑢𝑗 ℎ + 1 ℎ

Worst case waiting time Worst case arrival instant

L

Worst case arrival instant

L: technological latency

𝑋

𝑗 ℎ 𝑢 = 𝑋𝑀𝑄𝑗 ℎ 𝑢 + 𝑋𝑇𝑄𝑗 ℎ 𝑢 + 𝑋𝐼𝑄𝑗 ℎ 𝑢

𝑋

𝑗 ℎ 𝑢

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-17
SLIDE 17

17

WORST CASE WAITING TIME FOR A FRAME IN A NODE DURING A TIME INTERVAL

𝑔

𝑗

𝑢 𝑔

𝑗

𝑋ℎ 𝑢 = ෍

𝑤𝑘∈Γℎ

𝑠𝑐𝑔

𝑘 ℎ(𝑢)

ℎ

𝑋ℎ 𝑢 − 𝑢 − 𝐷𝑗

ℎ

  • 𝑠𝑐𝑔

𝑘 ℎ 𝑢 : transmission time of all frames generated from 𝑤𝑘 in the node ℎ during 𝑢.

FA method FA extension to CBS algorithm Experimentation Conclusion Assumption : one level of priority (FIFO)

slide-18
SLIDE 18

18 18

MAXIMUM INTERFERENCE GENERATED IN A NODE DURING 0, 𝑢

ℎ − 1 ℎ 𝑢 𝑢

ℎ

𝑢

….

Burst of frames

L L

Given a frame 𝑔

𝑗 generated from 𝑤𝑗 in the node ℎ incoming from an imput port 𝐽𝑄 1 ℎ during 𝑢.

𝑔

𝑗

𝑔

𝑗

𝑔

𝑗

𝑋

𝑗 ℎ 𝑢 = 𝑋𝑀𝑄𝑗 ℎ 𝑢 + 𝑋𝑇𝑄𝑗 ℎ 𝑢 + 𝑋𝐼𝑄𝑗 ℎ 𝑢 FA method FA extension to CBS algorithm Experimentation Conclusion

slide-19
SLIDE 19

19

WORST CASE WAITING TIME IN A NODE

While 𝑋

𝑗 ℎ 𝑢 ≥ t

ℎ 𝑢 𝑋ℎ(𝑢)

𝑛𝑏𝑦𝑢≥0(𝑋

𝑗 ℎ 𝑢 − 𝑢)

Maximum delay incurred in the node ℎ =

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-20
SLIDE 20

FA extension to CBS algorithm

20

slide-21
SLIDE 21

21

SRB SRA BE SRA SRA SRA SRB SRA SRA

Credit B Credit A

𝑢 𝑢 𝑢

CBS ALGORITHM

BE

𝑋

𝑗 ℎ 𝑢 = 𝑋𝑀𝑄𝑗 ℎ 𝑢 + 𝑋𝑇𝑄𝑗 ℎ 𝑢 + 𝑋𝐼𝑄𝑗 ℎ 𝑢 +𝑋𝑂𝐽𝑀𝑗 ℎ 𝑢

A flow 𝑤𝑗 belongs to a SR class 𝑌𝑗

𝑢1 𝑢2 𝑢3 𝑢5 𝑢4 FA method FA extension to CBS algorithm Experimentation Conclusion

slide-22
SLIDE 22

22

IMPACT OF SAME PRIORITY FLOWS 𝑋𝑇𝑄𝑗

ℎ 𝑢 = ෍ 𝑦=1 𝑜

min(𝑠𝑐𝑔

𝑗,𝑦 ℎ 𝑢 , 𝑀𝑗𝑜𝑙𝑆𝑏𝑢𝑓𝑗,𝑦 ℎ 𝑢 , 𝑇ℎ𝑏𝑞𝑓𝑠𝑆𝑏𝑢𝑓𝑗,𝑦 ℎ (𝑢))

𝑢 𝐷𝑠𝑓𝑒𝑗𝑢𝑌𝑗 𝑛𝑏𝑦𝑤𝑘∈𝑌𝑗𝐷

𝑘

𝐷𝑠𝑒𝑌𝑗,𝑛𝑏𝑦

ℎ

  • 𝜏𝑌𝑗

ℎ,−

(i) (iii) (ii)

Busy period of 𝐼𝑄

𝑌𝑗 ℎ

𝐷𝑠𝑓𝑒𝑗𝑢𝑌𝑗 𝐷𝑠𝑒𝑌𝑗,𝑛𝑏𝑦

ℎ

𝜏𝑌𝑗

ℎ,+

𝑛𝑏𝑦𝑤𝑘∈𝑀𝑄𝑌𝑗

ℎ 𝐷

𝑘

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-23
SLIDE 23

23

IMPACT OF HIGH PRIORITY FLOWS

ℎ 𝑢 𝑔

𝑗

Given a frame 𝑔

𝑗 generated from 𝑤𝑗 belonging to an SR Class 𝑌𝑗 in the node ℎ during 𝑢.

Serialized part Non serialized part 𝑋

𝑗 ℎ(𝑢)

𝑔

𝑗

𝑒𝑓𝑞𝑗

ℎ(𝑢)

ℎ

𝑢

….

Burst of frames

𝑢 𝐷𝑠𝑓𝑒𝑗𝑢𝑌𝑗 𝑛𝑏𝑦𝑤𝑘∈𝑌𝑗𝐷

𝑘

𝐷𝑠𝑒𝑌𝑗,𝑛𝑏𝑦

ℎ

  • 𝜏𝑌𝑗

ℎ,−

(i) (iii) (ii)

𝑋𝐼𝑄

𝑗 ℎ 𝑢 = 𝑛𝑗𝑜( ෍ 𝑤𝑘∈ℎ𝑞𝑗

ℎ

𝑠𝑐𝑔

𝑘 ℎ

𝑒𝑓𝑞𝑗

ℎ 𝑢

, ෍

𝑍∈𝐼𝑄𝑌𝑗

ℎ

𝐶𝑝𝑠𝑒𝑍

ℎ(𝑒𝑓𝑞𝑗 ℎ 𝑢 + 𝑛𝑏𝑦𝑤𝑘∈𝑇𝐼𝑄𝑍

ℎ𝐷

𝑘 ))

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-24
SLIDE 24

24

TAKING INTO ACCOUNT RESPLENISHING TIME OF THE CREDIT OF 𝑌𝑗

𝐷𝑠𝑓𝑒𝑗𝑢𝑌𝑗 𝐷

𝑘

𝜏− 𝜏+

𝑋𝑇𝑄𝑗

ℎ 𝑢

1 + 𝜏− 𝜏+ 𝑋𝑇𝑄𝑗

ℎ 𝑢 + 𝑋𝑂𝐽𝑀𝑗 ℎ 𝑢

𝑢 𝑔

𝑘

  • 𝜏−

𝜏+ 𝐷

𝑘

𝑔

𝑙

𝐷𝑙 𝜏− 𝜏+ 𝐷𝑙

𝑋𝐼𝑄𝑗

ℎ 𝑢

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-25
SLIDE 25

Experimentations

25

slide-26
SLIDE 26

26

INDUSTRIAL CASE

Head_Unit RSE Amplifier CPU DA_CAM Gateway TV MM_Disk 𝒘𝒋 𝑼𝒋(𝝂𝒕) 𝑫𝒋(𝝂𝒕) Class 𝑺𝒋(𝝂𝒕) Source Destination 𝒘𝟐 5000 5,12 A 2762,72 Gateway CPU 𝒘𝟑 5000 5,12 A 2994 Gateway RSE 𝒘𝟒 125 32,64 A 188,96 DA_CAM Heand_Unit 𝒘𝟓 280 115,68 B 631,9 MM_Disk RSE 𝒘𝟔 1400 115,68 B 618,67 MM_Disk Amplifier 𝒘𝟕 560 115,68 B 395,68 TV Head_Unit

  • Technological latency= 8𝜈𝑡.
  • Service rate= 100 Mbit/s.

[Li et al., Henia et al., Steinbach et al.] FA method FA extension to CBS algorithm Experimentations Conclusion

slide-27
SLIDE 27

27

COMPARISON WITH CPA

Ctrl 𝑇𝑥𝑗𝑢𝑑ℎ1 𝑇𝑥𝑗𝑢𝑑ℎ16 I/O I/O Ctrl 𝑇𝑥𝑗𝑢𝑑ℎ1 I/O I/O I/O I/O I/O I/O 𝑇𝑥𝑗𝑢𝑑ℎ4 I/O I/O I/O I/O I/O I/O Linear topology Clustered topology

  • Technological latency = 0,33𝜈𝑡.
  • Service rate = 1 Gbit/s.

[Diemer et al.] FA method FA extension to CBS algorithm Experimentations Conclusion

slide-28
SLIDE 28

28

RESULTS

Linear topology

Flow Worst case delay(𝜈𝑡)

FA method FA extension to CBS algorithm Experimentations Conclusion

slide-29
SLIDE 29

Conclusion

29

slide-30
SLIDE 30

30

CONCLUSION

  • Tight bounds when taking into account :
  • the serialization effect ;
  • the shaper on the maximum interference from high and same

priority flows ;

  • and the resplenishing time of the credit without counting

twice the interference of high priority flows.

  • Interesting results compared to CPA.

FA method FA extension to CBS algorithm Experimentation Conclusion

slide-31
SLIDE 31

31

PERSPECTIVES

  • Extension of FA method to the servicing policy of TSN standard

(Time Aware Shaper).

FA method FA extension to CBS algorithm Experimentation Conclusion Credit Based Shaper (CBS)

Classe SR A Classe SR B Flux BE Time aware shaper Critical traffic

TDMA

slide-32
SLIDE 32

NASSIMA.BENAMMAR@ENSMA.FR

Questions?

32

Thank you for your attention

slide-33
SLIDE 33

COMPLEXITY OF AUTOMOTIVE SYSTEMS : DISTRIBUTED ARCHITECTURE

33

https://www.st.com/en/solutions-reference-designs/automotive-solutions.html

slide-34
SLIDE 34

References

34

  • Bauer, H., et al. « Worst-Case Backlog Evaluation of Avionics Switched Ethernet Networks with the Trajectory

Approach ». 2012 24th Euromicro Conference on Real-Time Systems, 2012, p. 78-87. IEEE Xplore, doi:10.1109/ECRTS.2012.12.

  • Bauer, Henri. Analyse pire cas de flux hétérogènes dans un réseau embarqué avion. 4 octobre 2011. oatao.univ-

toulouse.fr, http://ethesis.inp-toulouse.fr/archive/00001615/.

  • --. « Applying Trajectory Approach with Static Priority Queuing for Improving the Use of Available AFDX

Resources ». Real-Time Systems, vol. 48, no 1, janvier 2012, p. 101-33. link.springer.com, doi:10.1007/s11241-011- 9142-9.

  • Benammar, N., et al. « Forward end-to-end delay analysis extension for FP/FIFO policy in AFDX networks ». 2017

22nd IEEE International Conference on Emerging Technologies and Factory Automation (ETFA), 2017, p. 1-8. IEEE Xplore, doi:10.1109/ETFA.2017.8247606.

  • --. « Forward End-to-End Delay for AFDX Networks ». IEEE Transactions on Industrial Informatics, vol. 14, no 3,

mars 2018, p. 858-65. IEEE Xplore, doi:10.1109/TII.2017.2720799.

  • Benammar, Nassima, et al. « Tighter Buffer Dimensioning in AFDX Networks ». SIGBED Rev., vol. 13, no 4,

novembre 2016, p. 37–42. ACM Digital Library, doi:10.1145/3015037.3015043.

  • Bordoloi, U. D., et al. « Schedulability analysis of Ethernet AVB switches ». 2014 IEEE 20th International Conference
  • n Embedded and Real-Time Computing Systems and Applications, 2014, p. 1-10. IEEE Xplore,

doi:10.1109/RTCSA.2014.6910530.

  • Boudec, Jean-Yves Le, et Patrick Thiran. Network Calculus: A Theory of Deterministic Queuing Systems for the
  • Internet. Springer-Verlag, 2001. www.springer.com, //www.springer.com/us/book/9783540421849.
slide-35
SLIDE 35

References

35

  • Bouillard, Anne, et Aurore Junier. « Worst-case Delay Bounds with Fixed Priorities Using Network Calculus ».

Proceedings of the 5th International ICST Conference on Performance Evaluation Methodologies and Tools, ICST (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering), 2011, p. 381–390. ACM Digital Library, http://dl.acm.org/citation.cfm?id=2151688.2151731.

  • Diemer, J., et al. « Formal worst-case timing analysis of Ethernet topologies with strict-priority and AVB

switching ». 7th IEEE International Symposium on Industrial Embedded Systems (SIES’12), 2012, p. 1-10. IEEE Xplore, doi:10.1109/SIES.2012.6356564.

  • Henia, R., et al. « System level performance analysis - the SymTA/S approach ». IEE Proceedings - Computers and

Digital Techniques, vol. 152, no 2, mars 2005, p. 148-66. IEEE Xplore, doi:10.1049/ip-cdt:20045088.

  • Kemayo, Georges Arnaud. Evaluation et validation des systèmes distribués avioniques. ISAE-ENSMA Ecole

Nationale Supérieure de Mécanique et d’Aérotechique - Poitiers, 23 septembre 2014. tel.archives-ouvertes.fr, https://tel.archives-ouvertes.fr/tel-01127020/document.

  • Li, Xiaoting, et Laurent George. « Deterministic Delay Analysis of AVB Switched Ethernet Networks Using an

Extended Trajectory Approach ». Real-Time Systems, vol. 53, no 1, janvier 2017, p. 121-86. link.springer.com, doi:10.1007/s11241-016-9260-5.

  • Steinbach, T., et al. « Tomorrow’s In-Car Interconnect? A Competitive Evaluation of IEEE 802.1 AVB and Time-

Triggered Ethernet (AS6802) ». 2012 IEEE Vehicular Technology Conference (VTC Fall), 2012, p. 1-5. IEEE Xplore, doi:10.1109/VTCFall.2012.6398932.

slide-36
SLIDE 36

LA FONCTION REQUEST BOUND FUNCTION

Annexe

36

Un flux sporadique non préemptif 𝑤𝑗 est défini par :

  • 𝐷𝑗 , 𝑈𝑗, 𝑄𝑏𝑢ℎ𝑗
  • 𝑇𝑛𝑏𝑦𝑗

ℎ: le délai de traversée maximum pour une trame de 𝑤𝑗 à partir de son node source

pour atteindre le node ℎ.

  • 𝑇𝑛𝑗𝑜𝑗

ℎ: le délai de traversée minimum pour une trame de 𝑤𝑗 à partir de son node source

pour atteindre le node ℎ. Le nombre de trames maximum générées par 𝑤𝑗 dans le node ℎ durant [𝑢0, 𝑢1] avec (𝑢0 + t = 𝑢1) ≤ 1 + 𝑢1 − 𝑢0 𝑈i

≤

1 + 𝑢 + (𝑇𝑛𝑏𝑦𝑗

ℎ − 𝑇𝑛𝑗𝑜𝑗 ℎ)

𝑈i Car (𝑢0 − 𝑇𝑛𝑏𝑦𝑗

ℎ) + t = 𝑢1 − 𝑇𝑛𝑗𝑜𝑗 ℎ

𝑆𝐶𝐺

𝑗 ℎ(𝑢) =

1 + 𝑢 + (𝑇𝑛𝑏𝑦𝑗

ℎ − 𝑇𝑛𝑗𝑜𝑗 ℎ)

𝑈i 𝐷𝑗

slide-37
SLIDE 37

AUDIO VIDEO BRIDGING SWITCHED ETHERNET (AVB)

37

AVB VB Li Listener Non Non AVB VB ES Non Non AVB VB ES AVB VB Li Listener AVB VB Talk alker AVB VB bri bridge Non Non AVB VB bri bridge Dom Domain ine AVB VB Dom Domain ine AVB VB Phyis isic ical link LAN LAN bo boun undary AVB VB

  • Synchronization ;
  • Quality of service ;
  • SR : Stream reservation.
slide-38
SLIDE 38

38

MAXIMUM INTERFERENCE GENERATED FROM A FLOW

𝑔𝑗𝑠𝑡𝑢𝑘

𝑔

1

𝑔

2

𝑔

3

ℎ 𝑔𝑗𝑠𝑡𝑢𝑘, ℎ ∈ 𝑄𝑏𝑢ℎ𝑘

𝑔

1

𝑔

2

𝑔

3

𝑈

𝑘

𝑈

𝑘

𝑤𝑘 ∈ Γℎ 𝑔𝑗𝑠𝑡𝑢𝑘 ℎ ℎ …. 𝑢 𝑈

𝑘

𝑈

𝑘

𝑈

𝑘

𝑔𝑗𝑠𝑡𝑢𝑘 Supposition : one level of priority (FIFO)

slide-39
SLIDE 39

39

SERIALISATION EFFECT

ℎ

𝑔 𝑔′

ℎ

𝐽𝑄

𝑜 ℎ

𝐽𝑄

1 ℎ

𝐽𝑄

𝑦 ℎ

. . . . . .

𝑋ℎ 𝑢 = ෍

𝑦=1 𝑜

𝑋

𝑦 ℎ(𝑢)

Assumption : one level of priority (FIFO)

slide-40
SLIDE 40

40

MAXIMUM INTERFERENCE INCOMING FROM ONE INPUT PORT

𝑛𝑏𝑦 𝐷

𝑘

ℎ Maximum workload during 𝑢

𝑢 𝑢

L L ℎ − 1

𝑋

𝑦 ℎ 𝑢 ≤ ෍ 𝑤𝑘∈Γ𝑦

ℎ

𝑠𝑐𝑔

𝑘 ℎ(𝑢)

𝑋

𝑦 ℎ 𝑢 ≤ 𝑀𝑗𝑜𝑙𝑆𝑏𝑢𝑓𝑗,𝑦 ℎ (𝑢)

𝑀𝑗𝑜𝑙𝑆𝑏𝑢𝑓𝑗,𝑦

ℎ (𝑢)

Supposition : one level of priority (FIFO)

slide-41
SLIDE 41

TESTING TIME

Annexe

41

slide-42
SLIDE 42

42

DIFFÉRENCE DE RÉSULTATS POUR DEUX SCÉNARIOS ÉQUIVALENTS

ℎ ℎ − 1

𝑤1, 𝑤2,𝑤3 𝑤4 𝑤1 𝑤2 𝑤3 𝑤4 𝐷𝑗 10 10 40 10 𝑄𝑗 1 2 3 2 ℎ − 1 ℎ

𝑢 = 40

𝑔 𝑔

10 20 30 50 60 70 80 90

Temps d’attente= 40 Temps d’attente= 60

ℎ − 1 ℎ

𝑢 = 40

𝑔 𝑔

10 20 30 50 60 70 80 90 110 100

Annexe