Analysis of a Time-driven Chain of Dependent Components

1. Analysis of a Time-driven Chain of Dependent Components M.A. Weffers-Albu, J.J. Lukkien, P.D.V. v.d. Stok Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

2. Contents • Goal & Approach • Analyzed Systems • Characterization of behavior • QoS Requirement. First solution • Stable Phase Theory • Practical Applications • Future work Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

3. Goal & Approach • The goal of our work is the prediction and optimization of performance parameters to provide guaranteed and optimized Quality of Service (QoS) for real-time streaming applications. • Approach: • provide a characterization of streaming applications execution to determine performance parameters and provide insight into best design practices for optimising these attributes. • Performance parameters: • Response Time of tasks and chain (RT), • Resource utilization (RU) for CPU, memory. • Number of Context Switches (NCS) Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

4. … Physical Platform Media Processing Applications Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

5. Media Processing Applications … Physical Platform Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

6. Media Processing Applications … Physical Platform Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

7. Get Full Packet Put Full Packet …   fqi-1? fqi! Physical Platform eqi-1! eqi?   Put Empty Packet Get Empty Packet Component Processing code(ci)  fq1 fq2 fqN-1 C2 CN … C1 eq1 eq2 eqN-1 Media Processing Applications Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

8. arbitrary interleavings channel consistent traces schedule consistent traces priority consistent traces Characterization of chain behaviour • Express behavior as: • traces, • time assignment (schedule) associated with each trace Impose predicates until obtain trace and schedule that specify the system behavior. Unique trace ρ, eager scheduleeager Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

9. QoS Requirement QoS Requirement – CN executes always strictly at rate TN. First step solution - rate of production higher than rate of consumption for packets in fqN-1: PRkN-1. TN, k  N First step solution non-optimal, pessimistic – implies sum of max computation times of components actions must be smaller than TN. Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

10. … C1 CN Stable Phase Theorem Stable Phase Theorem - Provided that PRkN-1. TN , k  N, the pipeline system assumes a repetitive, periodic behavior after a finite initial phase.The complete behavior is characterized the unique trace: ρ= tinit (inc(i) fqN-1? eqN? cN eqN-1! tL fqN ! d(i *TN)) ω. tinit – trace recording the initial phaseof the system execution. tstable– stable phase: (inc(i) fqN-1? eqN? cN eqN-1! tL fqN ! d(i *TN)) ω tL – subtrace recording the interleaved execution of C1..CN-1. Execution in cascade of sub-chain - all backward queues empty. Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

11. TN … idle time idle time idle time tinit Stable Phase Theorem Stable Phase Theorem - Provided that PRkN-1. TN , k  N, the pipeline system assumes a repetitive, periodic behavior after a finite initial phase.The complete behavior is characterized the unique trace: ρ= tinit (inc(i) fqN-1? eqN? cN eqN-1! tL fqN ! d(i *TN)) ω. tinit – trace recording the initial phaseof the system execution. tstable– stable phase: (inc(i) fqN-1? eqN? cN eqN-1! tL fqN ! d(i *TN)) ω tL – subtrace recording the interleaved execution of C1..CN-1. Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

12. Practical applications • Given the computation times of components actions, ρ andeagercan be calculated at design time. • Hence NCS, task and chain RT, RU for CPU and memory can also be calculated. • CN has the same influence as a minimum priority data-driven component. • Minimum Queue Capacity: 1 • Chain RT cannot be optimized • Minimum NCS at Stable Phase: • P(C1)<…<P(CN-1) Cap(fqi) = 2 for all i, 1  i < N−1. Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven

13. On-going work • Optimal solution for satisfying QoS requirement • Study chains containing a time driven component at each end (realistic surveillance application) • Study chains that include video/audio decoding components, execution depends on input stream(realistic video/audio decoding chains). Alina Weffers-Albu, m.a.albu@tue.nl TU/e Computer Science, System Architecture and Networking Philips Research Laboratories Eindhoven