demystifying the real time linux scheduling latency
play

Demystifying the Real-Time Linux Scheduling Latency Daniel Bristot - PowerPoint PPT Presentation

Demystifying the Real-Time Linux Scheduling Latency Daniel Bristot de Oliveira , Daniel Casini, Rmulo Silva de Oliveira and Tommaso Cucinotta Principal Software Engineer 1 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro


  1. Demystifying the Real-Time Linux Scheduling Latency Daniel Bristot de Oliveira , Daniel Casini, Rômulo Silva de Oliveira and Tommaso Cucinotta Principal Software Engineer 1 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  2. Introduction Real-Time Linux 2 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  3. Introduction “Real-Time” Linux 3 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  4. Why “real-time” Linux? Real-Time Linux vs Real-Time theory Experimental vs Analytical 4 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  5. Why “real-time” Linux? Real-Time Linux vs Real-Time theory Real-time analysis Based on the timing description of the system ● Capture all behaviors ● Precisely define the worst cases ● But depends on a precise definition of the ● system 5 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  6. Why “real-time” Linux? Real-Time Linux vs Real-Time theory Linux approach Linux was adapted to become a RTOS ● PREEMPT_RT: De facto standard ● Evaluated (mainly) with cyclictest ● Cyclictest: ● Practical: lightweight and out-of-the-box ○ It is a “black-box” test ○ No demonstration ○ Does not provide evidence of “root-cause” ○ 6 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  7. Why “Real-time” Linux? Why don’t we apply RT analysis on Linux? 7 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  8. Why “Real-time” Linux? Linux is complex - Lots of contexts - Lots of hacks - Lots of information - Fast pacing - ... Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  9. A way out. The PREEMPT_RT thread model It defines the specifications of threads synchronization: 9 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  10. Demystifying the Real-Time Linux Scheduling Latency Approach Formal specification Scheduling latency bound Measurement and analysis 10 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  11. Formal Specification From formal specification to synchronization rules Formally backed natural language arguments Generators ● Basic/Independent behavior ○ e.g., irq_disable/enable, scheduler call ○ Translated into a set of operations ● Specifications ● Relations among generators ○ e.g., necessary conditions to call the scheduler ○ Translated into a set of synchronization rules ● 11 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  12. Scheduling latency bound Scheduling latency definition From the first necessary condition to set need resched , to the the last The scheduling latency experienced by an arbitrary thread τ is: action after the scheduling, which is enabling preemption after the longest time elapsed between the time A in which any job of τ ● the return from becomes ready and with the highest priority , __schedule(). and the time F in which the scheduler returns and allows τ to execute ● its code . 12 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  13. Scheduling latency bound Interference and blocking The scheduling latency in this paper refers to the delay between the notification of a new highest priority thread, to The scheduling latency is caused by: point in which this thread starts running its own Blocking from the current (and so lower) priority ● code. thread; The highest priority thread Including scheduling. ● can belong to any scheduler: the analysis is scheduler independent. Interference from IRQs and NMI. ● 13 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  14. Blocking bound Blocking bound From the specification that bounds the block to a timeline 14 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  15. Blocking bound Timeline and cases All possible cases 15 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  16. Blocking bound Blocking variables D POID : preemption or interrupts disabled to ● In the model, the postpone the scheduler; preemption control is D PAIE : preemption and interrupts enabled, as a ● specialized into two transient state from poid to psd ; when scheduling different operations: to a new highest priority thread. postpone the scheduler D PSD : preemption disable to schedule; ● (the most known D ST : delay caused by the scheduling tail; the “non behavior) or to protect the ● execution of the return” point in which a new arrived task will have to __schedule() function from wait for the current scheduling operation to finish recursion. before scheduling. 16 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  17. Rtsl toolking Timeline and cases Variables in the the timeline 17 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  18. Interference bound Timeline and cases IRQ and NMI interference 18 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  19. Scheduling latency bound And the scheduling latency bounds to: The bound considers all possible cases. Note that the Latency L is present in both sides of the equation. So, L is bounded by the L = max(D ST , D POID ) + D PAIE + D PSD + I NMI (L) + I IRQ (L) least positive value fulfilling the equation (like on RTA). 19 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  20. Interrupts characterization Interrupts are workload dependent This topic was heavily Instead of proposing “the best” interrupt discussed at the Real-time ● Micro Conference (inside characterization, the rtsl reports the scheduling Linux Plumbers) in 2019, latency based on some well-known more info here: characterizations: No interrupt ○ Worst single interrupt ○ Single occurence of all interrupts ○ Sporadic ○ Sliding window (Author’s preferred) ○ Sliding window with oWCET ○ 20 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  21. rt_sched_latency toolkit A practical scheduling latency estimation tool Method and challenges Based on the latency bound ● The latency bound is based on the model ● The model is based on tracing of events ● but high frequency events ○ hundreds MB/sec/CPU ■ Challenges: ● To minimize the (runtime) overhead ○ Work out-of-the-box ○ 21 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  22. A toolkit rt_sched_latency (rtsl) Based on perf Works in two phases: - The record mode saves the trace data; - The report mode process the trace and does the analysis. 22 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  23. rt_sched_latency toolkit record phase Low overhead trace recording Filters the high frequency trace ● Doing in-kernel processing ○ For blocking variables: ● Reports only the discover of new max values ○ For IRQ and NMI: ● Reports one event for each occurrence ○ Discounts the interference: ● e.g., IRQ interference on a poid ○ 23 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  24. rt_sched_latency toolkit report phase Low overhead trace recording After the capture, analyzes the trace. ● All in user-space. ○ Most of the analysis is done in python ● Easy to extend ○ Two outputs: ● Textual: good for debug ○ Chart: good comparisons (and papers :-)) ○ Does a per-cpu scheduling latency analysis ● Using different IRQ/NMI characterization... ○ 24 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  25. rt_sched_latency toolkit rtsl report output Textual output 25 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

  26. rt_sched_latency toolkit rtsl report output Chart output 26 Demystifying The Real-Time Linux Scheduling Latency - 32 nd Euromicro Conference on Real-Time Systems - ECRTS’20

Recommend


More recommend