Design Optimization of Multi-Cluster Embedded Systems for Real-Time Applications

1. Design Optimization of Multi-Cluster Embedded Systemsfor Real-Time Applications Paul Pop, Petru Eles, Zebo Peng, Viaceslav IzosimovEmbedded Systems Lab (ESLAB)Linköping University, Sweden Magnus Hellring, Olof BridalDept. of Electronics and SoftwareVolvo Technology Corp., Göteborg, Sweden

2. Distributed Heterogeneous System NoCs ... ... ... ... ... Factory Systems Heterogeneous NetworksMulti-Cluster Systems Automotive Electronics Heterogeneous Networks

3. ... • Applications distributed over the heterogeneous networks • Reduce costs:use resources efficiently • Requirements:close to sensors/actuators Gateway ... • Applications distributed over heterogeneous networks are difficult to... • Analyze (guaranteeing timing constraints) • Design(partitioning, mapping, bus access optimization) [DATE’03] This paper! Distributed Safety-Critical Applications

4. Outline • Motivation • System architecture and application model • Scheduling for multi-clusters [DATE’03] • Design optimization problems • Partitioning • Mapping • Bus access optimization • Optimization strategy • Experimental results • Contributions and Message

5. ... Time-triggered cluster • Static cyclic scheduling • Time-triggered protocol Gateway ... Event-triggered cluster • Fixed priority preemptive scheduling • Controller area network protocol Time Triggered Protocol (TTP) • Bus access scheme: time-division multiple-access (TDMA) • Schedule table located in each TTP controller: message descriptor list (MEDL) Controller Area Network (CAN) • Priority bus, collision avoidance • Highest priority message wins the contention • Priorities encoded in the frame identifier S0 S1 S2 SG S0 S1 S2 SG Slot TDMA Round Cycle of two rounds Hardware Architecture

6. ... ... m3 m1 m1 m2 m3 Software Architecture Gateway Time-triggered cluster NG N2 N1 CPU CPU CPU CPU CAN Controller CAN Controller P1 P4 OutCAN OutTTP OutN2 T T P2 P3 MBI MBI TTP Controller TTP Controller Event-triggered cluster SG S1 SG S1 Round2

7. Multi-Cluster Scheduling [DATE’03] Application, Partitioning,Mapping,Architecture • MultiClusterScheduling algorithm • Schedulability analysis: communication delays through the gateway • Scheduling: cannot be addressed separately for each cluster TT Bus Configuration Priorities ResponseTimes ResponseTimeAnalysis StaticScheduling Offsets Multi-Cluster Scheduling ScheduleTables Responsetimes

8. ... ... • Output • Design implementation such that the application is schedulable • Partitioning for each un-partitioned process • Mapping for each un-mapped process • Priorities for ET messages • TDMA slot sequence and sizes for the TT bus • Priorities for ET processes • Schedule table for TT messages Partitioning andmapping Communicationinfrastructure Application: set of process graphs Architecture: Multi-cluster Schedulinginformation Problem Formulation • Input • System architecture • Application • Partial partitioning and mapping, based on the designer’s experience

9. Deadline for P5 Deadline for P6 Preemption not allowed N1 N1 N1 P1 P1 P1 Met (faster) N2 (faster) N2 (faster) N2 P4 P4 P3 P3 P4 P3 P6 P6 P6 Met Met Met (slower) N3 (slower) N3 (slower) N3 P2 P2 P2 Preempted P4 Missed P5 P5 P4 Met P5 Preempted N1 N2 N3 P1 70 X X P2 X X 40 P3 X 50 X P4 X 70 90 P5 X X 40 P6 X 40 X Motivational Example #1 In which cluster to place process P4? CAN P1 P2 TTC ETC N1 N2 N3 P3 P4 TTP P5 P6

10. N3 P3 Missed Deadline N2 P2 N1 P1 m1 m2 TTP S2 S3 SG S2 S3 S1 SG S3 S1 SG NG Met N3 P3 CAN N2 N4 N1 P1 m1 m2 TTP S2 S2 SG S2 S3 S1 S2 S3 S1 SG NG T T m1 CAN m2 N4 N4 P2 Motivational Example #2 Where to map process P2? CAN TTC ETC N2 N3 N4 N1 N1 N2 N3 N4 m1 m2 NG P1 20 X X X P1 P2 P3 P2 X 40 X 50 P3 X X 20 X TTP

11. Missed N1 P1 P4 Round 4 TTP S1 SG m1 SG m2 SG S1 m3 S1 m4 S1 SG NG T T T T Deadline m1 m3 m2 m4 CAN N2 P2 P3 Missed N1 P1 P4 TTP S1 SG m1 SG m2 SG S1 m3 m4 S1 SG NG T T T T m1 m2 m3 m4 CAN N2 P2 P3 P1 N1 N2 m1 m2 P1 20 X Met N1 P1 P4 P2 P3 P2 X 40 P3 X 20 m3 m4 P4 40 X TTP SG m1 m2 SG S1 m4 S1 m3 S1 SG P4 Round 1 NG T T T m2 m1 m4 m3 CAN N2 P3 P2 Motivational Example #3 What are the priorities on ETC? Which slot should come first on the TTC? CAN TTC ETC N1 N2 TTP

12. Optimization Strategy • Multi-Cluster Configuration • Initial Partitioninig and Mapping • Determines an initial partitioning and mapping • List scheduling-based greedy approach • Priority of ready processes: critical path • Partitioning and Mapping Heuristic • Iteratively improves on the initial partitioning and mapping • Intelligent design transformations that improve schedulability • Based on feedback from MultiClusterScheduling • Bus Access Optimization • Determines the slot sequence and lengths on the TTC, message priorities on the ETC • Greedy optimization heuristic • Straightforward solution • Partitioning and mapping that balances the utilization of processors and buses • Could be produced by a designer without optimzation tools

13. 100% 90% 80% 70% 60% 50% Bus AccessOptimization 40% 30% 20% Partitioning andMapping Heuristic 10% 0% Initial Partitioningand Mapping 50 100 150 200 250 Straightforwardsolution Experimental Results Can we increase the number of schedulable applications? Percentage schedulable applications [%] Number of processes

14. Multi-Cluster Configuration • Case study • Vehicle cruise controller • Distributed over theTT and ET clusters Partitioning and Mapping Heuristic Initial Partitioningand Mapping Bus Access Optimization Experimental Results, Cont. How time-consuming is our optimization strategy? 5:20 5:00 4:40 4:20 4:00 3:40 3:20 3:00 Average execution time [hours:minutes] 2:40 2:20 2:00 1:40 1:20 1:00 0:40 0:20 0:00 50 100 150 200 250 Number of processes

15. Analysis and optimization methods areneeded for the efficient implementation of applications distributed over interconnected heterogeneous networks. Contributions and Message • Contributions • Addressed design problems characteristic to multi-clusters • Partitioning • Mapping • Bus Access Optimization • Proposed heuristics for design optimization