Evaluation of admissible CAN bus load with weak synchronization - - PowerPoint PPT Presentation

Context and goal Experimental protocol Experimental results Conclusion

Evaluation of admissible CAN bus load with weak synchronization mechanism

Hugo Daigmorte, Marc Boyer

ONERA – The French aerospace lab International Conference on Real-Time Networks and Systems (RTNS 2017)

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 1 / 25

Context and goal Experimental protocol Experimental results Conclusion

Table of Contents

1

Context and goal CAN bus with offsets CAN identifier Sporadic flows and errors

2

Experimental protocol Breakdown utilisation Configuration Pattern

3

Experimental results Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

4

Conclusion

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 2 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Table of Contents

1

Context and goal CAN bus with offsets

Global clock Local clocks Bounded phases

CAN identifier Sporadic flows and errors

Sporadic flows Errors

2

Experimental protocol

3

Experimental results

4

Conclusion

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 3 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Context and goal

Context Real-time networked system (CAN bus) Periodic flows with Offsets

reduces contentions ⇒ reduces delays requires synchronization

RTNS 2016 new offest-based mechanism reduces delays

• nly periodic flows

Open questions quantitative gain evaluation? in practice: sporadic, errors?

CAN Node n°N CAN Node n°2 CAN Node n°1

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 4 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Flow model

Periodic flow N nodes Flow Fi: ji (Source node), Pi (Period),Si (maximal frame Size),Oi (Offset) Each node j has a clock: cj(t) Sending k-th frame of flow Fi when: cj(t) = Oi + kPi

P O

F,1

P

F,2 F,3

c(t)

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 5 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Global clock (aka TDMA)

∀j, j′ : cj(t) ≈ cj′(t) Advantage: efficient global schedule ⇒ no contention Drawback: perfect synchronization (HW/SW cost)

A,1 B,1 A,2 C,3 C,1 C,2 N1 N2 BUS A,1 B,1 A,2 A,1 B,1 A,2 C,3 C,1 C,2

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 6 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Local clocks

Advantages: efficient schedule ⇒ no contention intra-nodes efficient schedule ⇒ workload spread over time

A,1 B,1 A,2 C,1 C,2 N1 N2 BUS C,3 A,1 B,1 A,2 C,2 C,1 C,3 A,1 B,1 C,1 N1 N2 BUS B,1 A,2 C,2 A,1 B,1 A,2 C,1 C,2

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 7 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Bounded phases

∀j, j′ : cj(t) − cj′(t) ≤ Φj,j′ Objectives: Bounded phases: trade off between global and local clocks affordable synchronization reduces delays with regard to local clocks

A,1 B,1 A,2 C,3 C,1 C,2 N1 N2 A,1 B,1 A,2 C,3 C,1 C,2 N1 N2

X X

|c1(t) − c2(t)| ≤ x

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 8 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

CAN identifier encodes both label and priority

S O F Identifier Control Data Cyclic redundancy check A C K E O F R T R

Each message contains an identifier unique to the whole system: used as priority to solved bus access contentions

⇒ CAN bus used a non preemptive static priority policy

used as label

⇒ data semantics (engine speed, fuel pressure, etc.) ⇒ filter the message at the reception

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 9 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

CAN, priorities and offset

Priorities greatly influence the bus latency: Efficient priorities assignment ⇒ load close to 100% Industrial context ⇒ some or all labels (priorities) are constrained by design constraints: reusability: try to maximize the reuse of components standard: ARINC 825, SAE J1939 Offsets can be used in complement or independently.

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 10 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Sporadic flows and errors

Sporadic flows A frame is sent as soon as a specific event occurs Minimum update time MUT between two frames

MUT

F,1

MUT

F,2 F,3

c(t)

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 11 / 25

Context and goal Experimental protocol Experimental results Conclusion CAN bus with offsets CAN identifier Sporadic flows and errors

Sporadic flows and errors

Errors are a random phenomenon ⇒ cannot be forecast ⇒ Hypotheses are made at design Common error model Nerror, the burst errors,maximal number of errors that could

• ccur back-to-back

Terror, the residual error interval Maximal number of transmission errors during the duration d: Nerror +

• d

Terror

• − 1
• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 12 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

Table of Contents

1

Context and goal

Global clock Local clocks Bounded phases Sporadic flows Errors

2

Experimental protocol Breakdown utilisation Configuration Pattern

3

Experimental results

4

Conclusion

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 13 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

Response time depends on several parameters: periods, priority, size, offsets, link speed, nb of nodes, etc, Ideal objective:

compare bounded phases wrt no offsets and local clocks independently of other parameters

Realistic but significant measurement:

single configuration criteria: breakdown utilisation large number of random configurations ⇒ breakdown utilisation distribution

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 14 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

Breakdown utilisation

Breakdown utilisation The maximal load to guarantee that all deadlines are respected Example Breakdown utilisation Data sent on the network: 150 kbit/s The link speed in order to respect all deadlines must be at least: 300kbit/s ⇒ The breakdown utilisation for this configuration is 50%.

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 15 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

10 20 30 40 50 60 70 80 90 100 Breakdown Utilisation 50 100 150 200 250 300 350 400 450 500 Number of configuration

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 16 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

10 20 30 40 50 60 70 80 90 100 Breakdown Utilisation 50 100 150 200 250 300 350 400 450 500 Number of configuration

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 16 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

10 20 30 40 50 60 70 80 90 100 Breakdown Utilisation 50 100 150 200 250 300 350 400 450 500 Number of configuration

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 16 / 25

Context and goal Experimental protocol Experimental results Conclusion Breakdown utilisation Configuration Pattern

Configuration pattern under study Flows characteristics Sender: uniform choice between 16 nodes Periodic flows Period: uniform choice in {20, 25, 40, 50, 100, 200}ms Payload: 8 bytes Deadline: equal to their period, i.e. implicit deadlines Nb of flows: such that 150 + ε kbit/s sent on the bus Offsets chosen using the SOPA algorithm (RTaW-Pegase). 5,000 configurations generated. Breakdown utilisation for:

no offsets local clocks bounded phases (1ms)

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 17 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

Table of Contents

1

Context and goal

Global clock Local clocks Bounded phases Sporadic flows Errors

2

Experimental protocol

3

Experimental results Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 18 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

Phases bounded by 0.5-1-2ms

40 50 60 70 80 90 100 Breakdown Utilisation 200 400 600 800 1000 Number of configuration

Bounded Phases: 0.5 ms Bounded Phases: 1 ms Bounded Phases: 2 ms No offsets Local Clocks

Random priorities: No offsets: 52% Local clocks: 58% Bounded phases 2ms: 77% Bounded phases 1ms: 83% Bounded phases 0.5ms: 87%

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 19 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

Combining priorities and offsets

40 50 60 70 80 90 100 Breakdown Utilisation 200 400 600 800 1000 Number of configuration

No offsets: implicit deadline Bounded Phases 1ms: half implicit deadline No offsets: half implicit deadline Local Clocks: half implicit deadline

Monotonic assignment, implicit deadlines: No offsets: 95% Monotonic assignment, half implicit deadlines: No offsets: 73% Local clocks: 75% Bounded phases: 92%

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 20 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

No errors, 20% sporadic

40 50 60 70 80 90 100 Breakdown Utilisation 200 400 600 800 1000 Number of configuration

Bounded Phases 1ms, no sporadic Local clocks, no sporadic Bounded Phases 1ms, 20% sporadic Local clocks, 20% sporadic No offsets

No errors, no sporadic: No offsets: 52% Local clocks: 58% Bounded phases: 83% No errors, 20% sporadic: No offsets: 52% Local clocks: 57.5% Bounded phases: 73%

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 21 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

Errors, No sporadic

40 50 60 70 80 90 100 Breakdown Utilisation 200 400 600 800 1000 Number of configuration

Bounded Phases 1ms, no errors Local clocks, no errors No offsets, no errors Bounded Phases 1ms, with errors Local clocks, with errors No offsets, with errors

No errors, no sporadic: No offsets: 52% Local clocks: 58% Bounded phases: 83% Nerror=2,Terror=100ms: No offsets: 50% Local clocks: 55% Bounded phases: 77%

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 22 / 25

Context and goal Experimental protocol Experimental results Conclusion Phases bounded by 0.5-1-2ms Combining priorities and offsets No errors, 20% sporadic Errors, No sporadic Errors, 20% sporadic

Errors, 20% sporadic

40 50 60 70 80 90 100 Breakdown Utilisation 200 400 600 800 1000 Number of configuration

Bounded Phases 1ms, no errors and no sporadic Local clocks, no errors and no sporadic No offsets, no errors and no sporadic Bounded Phases 1ms, errors and 20% sporadic Local clocks, errors and 20% sporadic No offsets, errors and 20% sporadic

No errors, no sporadic: No offsets: 52% Local clocks: 58% Bounded phases: 83% Errors, 20% sporadic: No offsets: 50% Local clocks: 55% Bounded phases: 69%

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 23 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Offsets reduce contention and delays global clock has HW/SW cost

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 24 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Offsets reduce contention and delays global clock has HW/SW cost ⇒ Is there a benefit even with a weak inter-nodes synchronization?

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 24 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Offsets reduce contention and delays global clock has HW/SW cost ⇒ Is there a benefit even with a weak inter-nodes synchronization? RTNS 2016 Bounded phases: reduces delays

• nly periodic flows
• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 24 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Offsets reduce contention and delays global clock has HW/SW cost ⇒ Is there a benefit even with a weak inter-nodes synchronization? RTNS 2016 Bounded phases: reduces delays

• nly periodic flows

2017 contribution: breakdown utilisation distribution no offsets, local clocks, bounded phases purely periodic, mixing periodic and sporadic, with errors

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 24 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Results on 5,000 configurations:

very important gain with only periodic ⇒ 25% wrt local clocks complementary with priorities assignment compatible with sporadic flows ⇒ 15% wrt local clocks with 20% sporadic significant impact of errors

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 25 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Results on 5,000 configurations:

very important gain with only periodic ⇒ 25% wrt local clocks complementary with priorities assignment compatible with sporadic flows ⇒ 15% wrt local clocks with 20% sporadic significant impact of errors

Further work

• ffset assignment when some offsets are already set

combine the priorities assignment and the offsets assignment

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 25 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Results on 5,000 configurations:

very important gain with only periodic ⇒ 25% wrt local clocks complementary with priorities assignment compatible with sporadic flows ⇒ 15% wrt local clocks with 20% sporadic significant impact of errors

Further work

• ffset assignment when some offsets are already set

combine the priorities assignment and the offsets assignment

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 25 / 25

Context and goal Experimental protocol Experimental results Conclusion

Conclusion

Results on 5,000 configurations:

very important gain with only periodic ⇒ 25% wrt local clocks complementary with priorities assignment compatible with sporadic flows ⇒ 15% wrt local clocks with 20% sporadic significant impact of errors

Further work

• ffset assignment when some offsets are already set

combine the priorities assignment and the offsets assignment

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 25 / 25

Context and goal Experimental protocol Experimental results Conclusion

Local clocks and computation time

Two methods for bounds in the case local clocks: transaction based [1] network calculus based (contribution) Only one plotted in this presentation (the best calculable one).

Transaction based Network calculus based Accuracy Exact (up to one frame length) Upper bound On test cases same breakdown ± 3% Computation time mn up to h mn On test cases 7mn-2h10 1-1.5mn Sporadic No (can be adapted) Yes

Yomsi, P. M., Bertrand, D., Navet, N., & Davis, R. I. (2012, May). Controller area network (can): Response time analysis with offsets. In Factory Communication Systems (WFCS), 2012 9th IEEE International Workshop on (pp. 43-52). IEEE.

• H. Daigmorte, M. Boyer

Admissible CAN bus load with weak synchronization 25 / 25