270 likes | 378 Views
G53SRP: Introduction to Real Time Scheduling. Chris Greenhalgh School of Computer Science. Contents. Specifying timing constraints Specifying the system Current system example Tasks, values, time lines General strategy Fixed Priority Scheduling & DMPO Further issues
E N D
G53SRP: Introduction to Real Time Scheduling Chris Greenhalgh School of Computer Science
Contents • Specifying timing constraints • Specifying the system • Current system example • Tasks, values, time lines • General strategy • Fixed Priority Scheduling & DMPO • Further issues • Modes, fault tolerance, resource contention • Book: Wellings 7.5-7.7; Burns & Wellings 13.1-13.3
Real time… • A real-time system is any information processing system which has to respond to externally generated input stimuli within a finite and specified period • the correctness depends not only on the logical result but also the time it was delivered
Specifying timing constraints: temporal scopes • For a section of code (“process”, “task”, “scope”, Java Runnable), specify: • When it should (nominally) start • When it must be completed by class Foo implements Runnable { public void run() { … } }
Minimum delay Maximum delay a Blocking (sleeping, waiting, I/O), other task(s) running Maximum elapsed time b c Units of execution i.e. actually on CPU Deadline Release Time Maximum execution time = a + b +c
Specifying the system • Decompose into a fixed set of tasks • For each task: • Specify when it should be started • E.g. periodic (repeated, on a timer) or triggered by some other/external event (such as an interrupt) • Specify then deadline within which it should finish • Timing requirements derived from specification, e.g. system stability requirements • Arrange for the system to run those tasks at the right times and in the right order
T P S Concurrency Example (review) ADC Thermocouples Pressure Transducer • Overall objective is to keep the temperature and pressure of some chemical process within well-defined limits ADC Switch Heater Screen DAC Pump/Valve
Example system processes (or “tasks”, …) • P1 - temperature control task • (say) Run once every 100ms • Complete within 50ms • P2 – pressure control task • Run once every 150ms • Complete within 100ms
Process Release Time Process Completion Time Deadline Possible task execution 0 50 100 150 200 250 Example system abstract time-line Process P1 P2 Time (ms)
More questions… • How long will each task actually take execute? • In what order should the tasks be executed? • How many processors does the system have? • If more than one, on which processor should each task be run? • Will the deadlines be met?
More questions…example system • How long will each task actually take execute? • (say) P1 40ms, P2 60ms • In what order should the tasks be executed? • (say) P1 before P2 • How many processors does the system have? • (say) 1 • Will the deadlines be met? • See next slide…
0 50 100 150 200 250 Example system concrete time-line Process Release Time Process All deadlines met Process Completion Time Deadline met Process Completion Time Deadline missed P1 Actual task execution 60 50 10 P2 Task waiting/ pre-empted Time (ms)
Example system Gantt Chart CPU-1 P1 P2 P1 P2 P1 P2 0 50 100 150 200 250 Time (ms)
More questions…example system (v.2) • How long will each task actually take execute? • (say) P1 40ms, P2 70ms • In what order should the tasks be executed? • (say) P1 before P2 • How many processors does the system have? • (say) 1 • Will the deadlines be met? • See next slide…
0 50 100 150 200 250 Example system (v.2) concrete time-line Process Release Time Process P2 deadlines missed Process Completion Time Deadline met Process Completion Time Deadline missed P1 Actual task execution 60 10 50 20 P2 Task waiting/ pre-empted Time (ms)
More questions…example system (v.3) • How long will each task actually take execute? • (say) P1 40ms, P2 60ms • In what order should the tasks be executed? • (say) P2 before P1 • How many processors does the system have? • (say) 1 • Will the deadlines be met? • See next slide…
0 50 100 150 200 250 Example system (v.3) concrete time-line Process Release Time Process First P1 deadline missed Process Completion Time Deadline met Process Completion Time Deadline missed P1 Actual task execution 60 60 P2 Task waiting/ pre-empted Time (ms)
General answers… (1) • How long will each task actually take execute? • i.e. worst-case execution time • Determine from static code analysis and/or controlled testing
General answers… (2) • In what order should the tasks be executed? 1. Decide a general strategy, (e.g.) fixed priority scheduling • Assign a priority to each process in advance and always run the highest-priority runnable process • (Want a strategy that is flexible but deterministic and analysable – and supported)
2. Assign priorities to each process, (e.g.) Shortest Deadline First • Also known as Deadline Monotonic Priority Ordering (DPMO) • Can be shown to be optimal, i.e. if any strategy works then this will • Example system: P1 deadline 40 < P2 deadline 60
General answers… (3) • How many processors does the system have? If more than one, on which processor should each task be run? • Depends on system design • Typically fix tasks to processors for predictability
General answers… (4) • Will the deadlines be met? • Can be tackled through static analysis of whole system • E.g. response time analysis - see later notes • Note: equivalent to example time-line with all processes released at time zero • N.B. this is what we were after: predictable timing behaviour
Further issues: modes • Many systems have different “modes” of operation • e.g. start-up, active, shut-down • Each may have its own • Specified process set • Timing parameters • Priorities • Scheduling analysis • Plus changes between modes must be checked
Further issues: fault tolerance • What if the running system does… • Miss a deadline? • Use more processor time than it should? • System may • Detect missed deadlines or cost overrun • Interrupt or suspect that process • Call an error handler, to take contingency actions
Further issues: resource contention • If processes • share resources (e.g. common data or devices) • and require exclusive access (i.e. locks) then this must be taken into account when • setting process priorities • perhaps dynamically – see later notes • analysing the system’s schedulability • may introduce additional process blocking
Summary • A real-time system is comprised of tasks or processes with specified timing constraints • When/how often it should happen • When it should finish by (deadline) • How long it would take in isolation (cost) • The real-time system schedules these tasks • e.g. using fixed priority scheduling and deadline-monotonic priority ordering
Summary (2) • On- or off-line analysis of the system can then determine whether deadlines should be met or not • Missed deadlines or cost over-runs may be detected, allowing the system to respond to failures (fault tolerance) • Scheduling and analysis may be complicated by (e.g.) system modes or contention for shared resources