Clock-Driven Scheduling

Real-Time and Embedded Systems (M)

Lecture 4
 Lecture Outline

• Assumptions and notation for clock-driven scheduling

• Handling periodic jobs
– Static, clock-driven schedules and the cyclic executive
• Handling aperiodic jobs
– Slack stealing
• Handling sporadic jobs
• Advantages and disadvantages of clock driven scheduling
Copyright © 2006 University of Glasgow

Material corresponds to chapter 5 of Liu’s book

 Assumptions

• Clock-driven scheduling applicable to deterministic systems

• A restricted periodic task model:
– The parameters of all periodic tasks are known a priori
– For each mode of operation, system has a fixed number, n, periodic tasks
• For task Ti each job Ji,k is ready for execution at its release time ri,k and is
released pi units of time after the previous job in Ti such that ri,k = ri,k-1 + pi
• Variations in the inter-release times of jobs in a periodic task are negligible

– Aperiodic jobs may exist

• Assume that the system maintains a single queue for aperiodic jobs
• Whenever the processor is available for aperiodic jobs, the job at the head of
this queue is executed
Copyright © 2006 University of Glasgow

– There are no sporadic jobs

• Recall: sporadic jobs have hard deadlines, aperiodic jobs do not
 Notation
• The 4-tuple Ti = (φi, pi, ei, Di) refers to a periodic task Ti with
phase φi, period pi, execution time ei, and relative deadline Di
– Default phase of Ti is φi = 0, default relative deadline is the period Di = pi
Omit elements of the tuple that have default values
– Examples:
T1 = (1, 10, 3, 6) ⇒ φ1 = 1 p1 = 10 e1 = 3 D1 = 6
J1,1 released at 1, deadline 7
J1,2 released at 11, deadline 17
…

T2 = (10, 3, 6) ⇒ φ2 = 0 p2 = 10 e2 = 3 D2 = 6
J1,1 released at 0, deadline 6
J1,2 released at 10, deadline 16
…
Copyright © 2006 University of Glasgow

T3 = (10, 3) ⇒ φ3 = 0 p3 = 10 e3 = 3 D3 = 10
J1,1 released at 0, deadline 10
J1,2 released at 10, deadline 20
…
 Static, Clock-driven Cyclic Scheduler

• Since the parameters of all jobs with hard deadlines are known
can construct a static cyclic schedule in advance
– Processor time allocated to a job equals its maximum execution time
– Scheduler dispatches jobs according to the static schedule, repeating each
hyperperiod
– Static schedule guarantees that each job completes by its deadline
• No job overruns ⇒ all deadlines are met

• Schedule calculated off-line ⇒ can use complex algorithms

– Run-time of the scheduling algorithm irrelevant
– Can search for a schedule that optimizes some characteristic of the system
• e.g. a schedule where the idle periods are nearly periodic; accommodating
aperiodic jobs
Copyright © 2006 University of Glasgow

– The book gives an example, but scheduling algorithm unimportant as long

as resulting schedule meets deadline
 Example Cyclic Schedule

• Consider a system with 4 independent periodic tasks:

– T1 = ( 4, 1.0) Execution time

– T2 = ( 5, 1.8) [Phase and deadline take default values]

Period
– T3 = (20, 1.0)
– T4 = (20, 2.0)
• Hyper-period H = 20 (least common multiple of 4, 5, 20, 20)
• Can construct an arbitrary static schedule to meet all deadlines:

Feasible intervals

Execution times T1 T3 T2 T1 T4 T2 T1 T2 T1 T1 T2 T1

0 4 8 12 16 20
Copyright © 2006 University of Glasgow

Slack time, can be used

to execute aperiodic jobs.
 Implementing a Cyclic Scheduler

• Store pre-computed schedule as a table k tk T(tk)

0 0 T1
• The system creates all the tasks that are to be 1 1 T3
executed: 2 2 T2
– Allocates sufficient memory for the code and data 3 3.8 I
of every task 4 4 T1

– Brings the code executed by the task into memory 5 5 I

6 6 T4
• Scheduler sets the hardware timer to interrupt 7 8 T2
at the first decision time, tk=0 8 9.8 T1

• On receipt of an interrupt at tk: 9 10.8 I

10 12 T2
– Scheduler sets the timer interrupt to expire at tk+1 11 13.8 T1
– If T(tk) = I and aperiodic job waiting, start 12 14.8 I
aperiodic job 13 17 T1
Copyright © 2006 University of Glasgow

– Otherwise, start next job in task T(tk) executing 14 17 I

15 18 T2
16 19.8 I
 Implementing a Cyclic Scheduler
Input: stored schedule (tk, T(tk)) for k = 0, 1, n – 1.
Task SCHEDULER:
set the next decision point i = 0 and table entry k = 0;
set the timer to expire at tk;
do forever:
accept timer interrupt;
if an aperiodic job is executing, preempt the job;
current task T = T(tk);
increment i by 1;
compute the next table entry k = i mod n;
set the timer to expire at [i / n] * H + tk;
if the current task T is I,
let the job at the head of the aperiodic queue execute;
else
let the task T execute;
Copyright © 2006 University of Glasgow

sleep;
end do.
End SCHEDULER.
 Structured Cyclic Schedules

• Arbitrary table-driven cyclic schedules flexible, but inefficient

– Relies on accurate timer interrupts, based on execution times of tasks
– High scheduling overhead
• Easier to implement if structure imposed:
– Make scheduling decisions at periodic intervals (frames) of length f
– Execute a fixed list of jobs with each frame, disallowing pre-emption
except at frame boundaries
– Require phase of each periodic task to be a non-negative integer multiple
of the frame size
• The first job of every task is released at the beginning of a frame
• φ = k⋅f where k is a non-negative integer
• Gives two benefits:
– Scheduler can easily check for overruns and missed deadlines at the end of
Copyright © 2006 University of Glasgow

each frame
– Can use a periodic clock interrupt, rather than programmable timer
 Frame Size Constraints

• How to choose frame length?

– To avoid preemption, want jobs to start and complete execution within a
single frame:
f ≥ max(e1, e2, …, en) (Eq.1)
– To minimize the number of entries in the cyclic schedule, the hyper-period
should be an integer multiple of the frame size (⇒ f divides evenly into the
period of at least one task):
∃ i : mod(pi, f ) = 0 (Eq.2)
– To allow scheduler to check that jobs complete by their deadline, should be
at least one frame boundary between release time of a job and its deadline:
2*f – gcd(pi, f ) ≤ Di for i = 1, 2, …, n (Eq.3)
(see book for derivation of Eq.3)
Copyright © 2006 University of Glasgow

• All 3 constraints should be satisfied

 Frame Size Constraints – Example

• Review the system of independent periodic tasks from our earlier

example:
T1 = (4, 1.0) T2 = (5, 1.8) Hyper-period H = lcm(4, 5, 20, 20) = 20
T3 = (20, 1.0) T4 = (20, 2.0)
• Constraints:
Eq.1 ⇒f ≥ max(1, 1.8, 1, 2) ≥ 2
Eq.2 ⇒f ∈ { 2, 4, 5, 10, 20 }
Only value that satisfies
Eq.3 ⇒2f - gcd( 4, f ) ≤ 4 (T1) all constraints is f = 2
2f - gcd( 5, f ) ≤ 5 (T2)
2f - gcd(20, f ) ≤ 20 (T3, T4)

Feasible intervals
Copyright © 2006 University of Glasgow

Execution times T1 T3 T2 T1 T4 T2 T1 T2 T1 T1 T2 T1

0 2 4 6 8 10 12 14 16 18 20
 Job Slices

• Sometimes, a system cannot meet all three frame size constraints

simultaneously
• Can often solve by partitioning a job with large execution time
into slices (sub-jobs) with shorter execution times/deadlines
– Gives the effect of preempting the large job, so allow other jobs to run
– Sometimes need to partition jobs into more slices than required by the
frame size constraints, to yield a feasible schedule

• Example:
– Consider a system with T1 = (4, 1), T2 = (5, 2, 7), T3 = (20, 5)
– Cannot satisfy constraints: Eq.1 ⇒ f ≥ 5 but Eq.3 ⇒ f ≤ 4
– Solve by splitting T3 into T3,1 = (20, 1), T3,2 = (20, 3) and T3,3 = (20,1)
Copyright © 2006 University of Glasgow

– Result can be scheduled with f = 4

 Building a Structured Cyclic Schedule

• To construct a cyclic schedule, we need to make three kinds of

design decisions:
– Choose a frame size based on constraints
– Partition jobs into slices
– Place slices in frames
• These decisions cannot be taken independently:
– Ideally want as few slices as possible, but may be forced to use more to get
a feasible schedule
Copyright © 2006 University of Glasgow
 Implementation: A Cyclic Executive

• Modify previous table-driven cyclic scheduler to be frame based,

schedule all types of jobs in a multi-threaded system
• Table that drives the scheduler has F entries, where F = H/f
– Each corresponding entry L(k) lists the names of the job slices that are
scheduled to execute in frame k; called a scheduling block
– Each job slice implemented by a procedure, to be called in turn
• Cyclic executive executed by the clock interrupt that signals the
start of a frame:
– Determines the appropriate scheduling block for this frame
– Executes the jobs in the scheduling block in order
– Starts job at head of the aperiodic job queue running for remainder of
frame
Copyright © 2006 University of Glasgow

• Less overhead than pure table driven cyclic scheduler, since only
interrupted on frame boundaries, rather than on each job
 Scheduling Aperiodic Jobs

• Thus far, aperiodic jobs are scheduled in the background after all
jobs with hard deadlines scheduled in each frame have completed
– Delays execution of aperiodic jobs in preference to periodic jobs
– However, note that there is often no advantage to completing a hard real-
time job early, and since an aperiodic job is released due to an event, the
sooner such a job completes, the more responsive the system
• Hence, minimizing response times for aperiodic jobs is typically a
design goal of real-time schedulers
Copyright © 2006 University of Glasgow
 Scheduling Aperiodic Jobs: Slack Stealing

• Periodic jobs scheduled in frames that end before their deadline;

there may be some slack time in the frame after the periodic job
completes
• Since we know the execution time of periodic jobs, can move the
slack time to the start of the frame, running the periodic jobs ‘just
in time’ to meet their deadline
• Execute aperiodic jobs in the slack time, ahead of periodic jobs
– The cyclic executive keeps track of the slack left in each frame as the
aperiodic jobs execute, preempts them to start the periodic jobs when there
is no more slack
– As long as there is slack remaining in a frame, the cyclic executive returns
to examine the aperiodic job queue after each slice completes
Copyright © 2006 University of Glasgow

• Reduces response time for aperiodic jobs, but requires accurate

timers
 Scheduling Sporadic Jobs

• We assumed there were no sporadic jobs – what if this is relaxed?

• Sporadic jobs have hard deadlines, release and execution times
that are not known a priori
– Hence, a clock-driven scheduler cannot guarantee a priori that sporadic
jobs complete in time
• However, scheduler can determine if a sporadic job is schedulable
when it arrives
– Perform an acceptance test to check whether the newly released sporadic
job can be feasibly scheduled with all the jobs in the system at that time
– If there is sufficient slack time in the frames before the new job’s deadline,
the new sporadic job is accepted; otherwise, it is rejected
• Can be determined that a new sporadic job cannot be handled as soon as that
job is released; earliest possible rejection
Copyright © 2006 University of Glasgow

– If more than one sporadic job arrives at once, they should be queued for
acceptance in EDF order
 Practical Considerations

• Handling overruns:
– Jobs are scheduled based on maximum execution time, but failures might
cause overrun
– A robust system will handle this by either: 1) killing the job and starting an
error recovery task; or 2) preempting the job and scheduling the remainder
as an aperiodic job
• Depends on usefulness of late results, dependencies between jobs, etc.
• Mode changes:
– A cyclic scheduler needs to know all parameters of real-time jobs a priori
– Switching between modes of operation implies reconfiguring the scheduler
and bringing in the code/data for the new jobs
– This can take a long time: schedule the reconfiguration job as an aperiodic
or sporadic task to ensure other deadlines met during mode change
Copyright © 2006 University of Glasgow

• Multiple processors:
– Can be handled, but off-line scheduling table generation more complex
 Clock-driven Scheduling: Advantages

• Conceptual simplicity
– Ability to consider complex dependencies, communication delays, and
resource contention among jobs when constructing the static schedule,
guaranteeing absence of deadlocks and unpredictable delays
– Entire schedule is captured in a static table
– Different operating modes can be represented by different tables
– No concurrency control or synchronization required
– If completion time jitter requirements exist, can be captured in the schedule
• When workload is mostly periodic and the schedule is cyclic,
timing constraints can be checked and enforced at each frame
boundary
• Choice of frame size can minimize context switching and
Copyright © 2006 University of Glasgow

communication overheads
• Relatively easy to validate, test and certify
 Clock-driven Scheduling: Disadvantages

• Inflexible
– Pre-compilation of knowledge into scheduling tables means that if
anything changes materially, have to redo the table generation
– Best suited for systems which are rarely modified once built
• Other disadvantages:
– Release times of all jobs must be fixed
– All possible combinations of periodic tasks that can execute at the same
time must be known a priori, so that the combined schedule can be pre-
computed
– The treatment of aperiodic jobs is very primitive
• Unlikely to yield acceptable response times if a significant amount of soft real-
time computation exists
Copyright © 2006 University of Glasgow
 Summary

We have discussed:
• Static, clock-driven schedules and the cyclic executive
• Handling aperiodic jobs
– Slack stealing
• Handling sporadic jobs
• Advantages and disadvantages of clock driven scheduling

Clock-driven scheduling applicable to static systems, small

number of aperiodic jobs; the next lecture begins our study
of priority scheduling for more dynamic environments
Copyright © 2006 University of Glasgow