Task Migration for FaultTolerance in MixedCriticality Embedded Systems
Task Migration for Fault-Tolerance in Mixed-Criticality Embedded Systems Prabhat Kumar Saraswat Paul Pop Jan Madsen Workshop on Adaptive and Reconfigurable Embedded Systems (at ESWeek’ 09) October 11, 2009, Grenoble, France
Problem Formulation ? ? ? Mapping? ? ? Soft Task Hard Task Utilization? • Given: Implementation and fault occurrence • Determine: Mapping and Utilization • Such that: – Deadlines for all hard real-time tasks are satisfied – Graceful degradation for soft tasks Online task migration and utilization allocation algorithm 2 DTU Informatics, Technical University of Denmark 26/02/2021
Example Embedded Applications ABS Requirements • Timing (Antilock – Breaking) Hard – Soft Engine Control • Safety Requirements – Permanent Faults Steering Wheel Faults – Transient – No fault tolerance Audio Same platform Climate Control • Economic Pressures • Multicore Power Seat Sun Roof Transmission • Control Example – Automotive Applications Hard Constraints 3 DTU Informatics, Technical University of Denmark Driver’s Info. panel FAULTS! Soft Constraints 26/02/2021
Outline • Application Model • Platform Model • Example • Task Migration and Bandwidth Allocation (TMBA) • Experimental results • Conclusions 4 DTU Informatics, Technical University of Denmark 26/02/2021
Application Model Hard real-time tasks WCET Deadline Period Safety-criticality Soft real-time tasks Permanent faults Transient faults Probability PDF Soft deadline Execution Time Periodic 5 DTU Informatics, Technical University of Denmark 26/02/2021
Platform Model A Without checkpointing τ1 Execution Segment τ1τ1 1 τ1 2 With checkpointing and fault recovery Checkpointing Overhead Error Detection Overhead Recovery Overhead 6 DTU Informatics, Technical University of Denmark 26/02/2021
Constant Bandwidth Server – Each soft task is assigned a CBS with parameters: • Qi – maximum server budget (bandwidth) Soft • Ti – server period (equal to the period of the soft task) Hard – A soft task is allowed to execute for only Qi units of time every period Ti Processor Util. – Probability of meeting the deadline (Qo. S) depends on Qi 7 DTU Informatics, Technical University of Denmark 26/02/2021
CBS Example [Abeni 98] Hard WCET=2 Period=3 3 2+7+7+7 2+7+71 2+7 2 Soft Requests CBS Bandwidth = 2 Period = 7 2 8 4 6 8 DTU Informatics, Technical University of Denmark 10 12 14 16 18 20 22 26/02/2021
Stochastic Analysis Example How does Q affects the Qo. S? (Probability of meeting the deadline for soft tasks) Important to choose right Q! 9 DTU Informatics, Technical University of Denmark 26/02/2021
Example PE 2 PE 1 τ3 8 20 13 40 τ10 30(48) 230 Offline Solution Qo. S : 72. 21 % τ7 31(39) 12(39) τ8 31(62) 19(62) 150 PE 3 τ9 12(46) 200 300 DTU Informatics, Technical University of Denmark τ9 τ4 150 190 τ2 99. 54% Qo. S 10 τ5 72. 21% Offline τ1 150 29(34) Initial τ6 29(33) 18(33) PE 3 Fails! 15 35 τ10 33(46) 200 8 25 35(48) 230 τi τi Q (Deadline) Period WCET Period 26/02/2021
Example PE 2 PE 1 200 8 20 40 τ10 20(48) 230 Time: Proposed Solution Qo. S : 70. 58 % Proposed <<Offline 11 τ7 31(39) 14(39) τ8 31(62) 23(62) PE 3 τ9 150 τ4 150 190 300 τ2 15 35 99. 54% DTU Informatics, Technical University of Denmark 72. 21% 70. 58% Proposed 13 τ5 Offline τ3 τ9 29(34) 11(34) Initial τ1 150 17(46) Qo. S τ6 29(33) 13(33) PE 3 Fails! τ10 33(46) 200 8 25 35(48) 230 τi τi Q (Deadline) Period WCET Period 26/02/2021
Greedy based Task Migration and Bandwidth Allocation (TMBA) Iteration • Greedy • Hard tasks considered first • Tasks ordered according to their Utilizations • CBS parameters are adjusted proportionally to their means. System Qo. S Tryingτ 4 on PE 1 X Tryingτ4 on PE 1 84. 11 Tryingτ9 on PE 1 78. 54 Tryingτ9 on PE 2 56. 32 Tryingτ10 on PE 1 70. 58 Tryingτ10 on PE 2 τ6 τ10 35(48) τ9 33(46) τ4 230 200 8 τ10 (0. 15) τ9 (0. 16) τ4 (0. 32) τ1 τ3 25 12 DTU Informatics, Technical University of Denmark 13(33) 29(33) 150 8 20 13 40 Can’t be mapped 59. 20 PE 2 PE 1 Failed Processor Decision τ9 (0. 08) τ10 20(48) 230 ττ6 (0. 18) τ9 17(46) τ1 200 τ10 (0. 09) 6 (0. 08) (0. 4) τ3 τ55 11(34) 29(34) τ77 14(39) 31(39) τ88 23(62) 31(62) τ4 τ(0. 32) 5 (0. 19) 150 190 300 τ4 ττ22 8 25 15 15 35 35 (0. 32) 26/02/2021 ττ7 (0. 16) 5 (0. 07) ττ87 (0. 10) 8 (0. 07) τ23 (0. 4)
Experimental Results Case Study – Portable media player Qo. S reported by TMBA : 73. 42 % Optimal Qo. S : 74. 19 % • Qo. S resulted by TMBA is quite close to the offline. (difference of only 0. 66%) • TMBA runs in polynomial time • Hard deadlines were satisfied for all cases 13 DTU Informatics, Technical University of Denmark 26/02/2021
Conclusion • A greedy-based online heuristic is proposed for migration of safetycritical tasks to tolerate permanent faults on a mixed hard/soft real-time system. • Better design choices can be made by taking stochastic execution times of soft tasks into consideration. • Proposed heuristic provides very good quality solutions. 14 DTU Informatics, Technical University of Denmark 26/02/2021
Thanks Questions? 15 DTU Informatics, Technical University of Denmark 26/02/2021
- Slides: 15