Embedded Real Time

Operating Systems
(ECE 367)
Dr. Rajiv Ranjan Singh
Professor
Department of Electronics and Communication
(ECE) Engineering
Lecture – 09-13

Real-Time Scheduling

2
 Outline (Lecture-09-13)

 Real-Time Tasks and its Scheduling

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 3
 Real-Time Scheduling
 Real-time tasks => Will have associated constraints

 Deadline constraint is the most common out of the three constraints

(Delay, Deadline and Duration)

 Real-time tasks => generated due to internal or external events.

 Internal Events => clock interrupt generated every 25 msec to poll the
temperature of a boiler.
 External events => pressing a switch

 A task when generated is said to have arrived or got released

 Appropriate Scheduling of Tasks helps in meeting the time constraints of

tasks.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 4
 Important Concepts and Terminologies
Task Instance
 Each time a task recurs => task instance
 First occurrence of a task => first instance
 Second occurrence of a task => second instance………and so on
 The jth instance of a task Ti is denoted as => Ti (j)
 Each instance of a task is associated with a deadline
 Task instances Processes

Relative Deadline vs Absolute Deadline

 Absolute Deadline => Absolute time value (from t = 0) by which the results
from task are expected => time interval between time t = 0 and deadline
instance of the corresponding task measured w.r.t. a clock.

 Relative Deadline => time interval between start (arrival) of the task and the
corresponding deadline.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Relative Deadline vs Absolute Deadline

Relative Deadline of task Ti(1) = d

Absolute Deadline of task Ti(1) = 𝚽 + 𝒅

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Response Time
 Time duration from the occurrence of the event generating the task to the time
task produces its results
 For Hard Real-Time Tasks need not be completed early as long as deadlines
are met
 For Soft Real-Time Tasks average response time is an important metric to
measure the performance of a scheduler.

Task Precedence
 A Task is said to precede another task if, the first task completes before the
second task starts.
 When a task Ti precedes another task Tj, then each instance of Ti precedes the
corresponding instance of Tj.
 If T1 precedes T2, then T1(1) precedes T2(1) and T1(2) precedes T2(2), and so on

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Task Precedence
 In the fig shown, T1 precedes T2, but we can not relate T1 with either T3 or T4.

Data Sharing
 Tasks often share data between each other => requires precedence
 A Task may precede another task even if, no data sharing happens.
 Data sharing may occur between concurrent tasks as well as overlapping tasks
to => data sharing can not be represented using precedence relation.
 In the fig shown, T2 uses results of T3 (shown using dashed line) but T2 and T3
may execute concurrently.
 T2 may even start executing first, after sometime it may receive some data from
T3 , and continue its execution, and so on.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Periodic Task
 A task that repeats after a certain fixed time interval (demarcated by clock
interrupts) => also referred as clock-driven tasks

 Period of a task = the fixed time interval after which the task repeats

 If Ti is a periodic task =>

 𝚽𝒊 = phase of Ti => time from t = 0 till the first instance of Ti (i.e. Ti (1) ).
 Second instance Ti (2) occurs at 𝚽𝒊 + 𝒑𝒊 .
 Third instance Ti (3) occurs at 𝚽𝒊 + 𝟐 ∗ 𝒑𝒊 .

 Formally, a periodic task can be represented by four tuple (𝚽𝒊 , 𝒑𝒊 , 𝒆𝒊 , 𝒅𝒊 )

 𝒑𝒊 => period of the task
 𝒆𝒊 => worst case execution time of the task
 𝒅𝒊 => relative deadline of the task

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Periodic Task
 Consider the track correction task found in a rocket control software
 track correction task starts at 2000ms after the launch of the rocket and recurs
every 50 ms
 Each instance of a task requires a processing time of 8 ms and its relative
deadline is 50 ms
 Ti (1) = (𝚽𝒊 , 𝒑𝒊 , 𝒆𝒊 , 𝒅𝒊 )

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Periodic Task
 Ti = (𝚽𝒊 , 𝒑𝒊 , 𝒆𝒊 , 𝒅𝒊 )
 𝚽𝒊 = phase of the task 2000 ms
 Ti can be formally represented as Ti = (2000 ms, 50ms, 8ms, 50ms)
 when 𝒑𝒊 = 𝒅𝒊 , the fourth tuple can be omitted, hence we can write
Ti = (2000 ms, 50ms, 8ms) => 𝒑𝒊 = 𝒅𝒊 = 50 ms

 When 𝚽𝒊 = 0 => it can also be omitted, when no confusion arises.

So Ti = (20 ms, 100ms) would indicate a new task with
 𝚽𝒊 = 0
 𝒑𝒊 = 100 ms
 𝒆𝒊 = 20 ms
 𝒅𝒊 = 100 ms

 If any confusion arises we normally write

So Ti = (𝒑𝒊 = 50ms, 𝒆𝒊 = 8ms, 𝒅𝒊 = 40ms)

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Periodic Task
 A vast majority of tasks in a real-time system are periodic in nature.
 Monitoring certain conditions
 Polling information from sensors at regular intervals => like in chemical
plants

 Periodic tasks can come into existence dynamically also.

 For e.g. the computation in Air Traffic Monitors, once a flight is detected, till
the flight exits the radar signal zone is an example of dynamically created
periodic tasks.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Sporadic Task
 Recurs at random instants
 Represented by Ti = (𝒆𝒊 , 𝒈𝒊 ,𝒅𝒊 )
 𝒆𝒊 => worst case execution time of an instance of a task
 𝒈𝒊 => minimum separation between two consecutive instances of the task
 𝒅𝒊 => relative deadline of the task

 𝒈𝒊 => once an instance of a sporadic task occurs, the second can not occur
before 𝒈𝒊 time units have elapsed.
 𝒈𝒊 restricts the rate at which the sporadic task can arise.
 First instance of a sporadic task Ti is denoted by Ti (1)
 Successive instances by Ti (2), Ti (3) etc.

 Many sporadic tasks are highly critical in nature => task handling fire in factory
 Highly critical => task handling fire in factory
 Moderately critical => I/O device interrupt or DMA interrupt etc.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Important Concepts and Terminologies
Aperiodic Task
 Similar to sporadic tasks in many ways
 Can arise at random instants
 𝒈𝒊 => minimum separation between two consecutive instances could be
‘0’
 Two or more instances can occur at the same time instant
 Deadline could be expressed either as average value or statistically
 Generally soft real-time tasks, as they can recur in quick succession =>
few deadline misses may be noticed as difficult to meet the deadlines of
all instances
 E.g. logging task in a distributed system (can be started different tasks
running at different nodes). Logging requests from different tasks may
arrive at the same time at the logger, or they may be spaced out in time
 Other examples – operator requests, keyboard presses, mouse
movements etc.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Task Scheduling
Real-Time Task Scheduling
 Refers to the order in which the various tasks are to be taken up for
execution by the operating system (OS).

 Every OS relies on one or more schedulers (characterized by

scheduling algorithms employed) to prepare the schedule of execution
of various tasks it needs to run.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Task Scheduling: Terminologies
Valid Schedule
 At most one task is assigned to a processor at time
 No task is scheduled before its arrival time
 Precedence and resources constraints of all tasks are satisfied

Feasible Schedule
 A valid schedule is called feasible, only if all tasks meet their respective time
constraints in the schedule

Proficient Scheduler
 A task scheduler sch1 is said to be more proficient than another scheduler sch2,
if sch1 can feasibly schedule all task sets that sch2 can feasibly schedule, but
not vice versa.
 A sch1 can feasibly schedule all task sets that sch2 can feasibly schedule and
vice versa, then sch1 and sch2 are called equally proficient schedulers.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Task Scheduling: Terminologies
Optimal Scheduler
 A task scheduler which can feasibly schedule any task set that can be feasibly
scheduled, by any other scheduler.
 An optimal scheduler can not schedule some task set, then no other scheduler
should be able to produce a feasible schedule for that task set.

Scheduling Points
 Scheduling Points of a scheduler are the points on time line at which the
scheduler makes decisions regarding which task is to be run next.
 A task scheduler need not run continuously, as it is activated by the OS only at
scheduling points to make the scheduling decisions as to which task is to be run
next.
 Clock-driven scheduler => defined at time instants marked by interrupts
generated by a periodic timer.
 Event-driven scheduler => determined by occurrence of certain events.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Task Scheduling: Terminologies
Preemptive Scheduler
 On arrival of a higher priority task, the scheduler suspends the lower priority task
that may be executing.
 A preempted lower priority task can resume its operation only when no higher
priority task is ready.

Utilization
 Processor utilization (or simply utilization) of a task is the average time for which
it executes per unit time interval.
𝒆
 For a periodic task Ti, the utilization is denoted as 𝒖𝒊 = 𝒑𝒊
𝒊
 𝒆𝒊 = execution time
 𝒑𝒊 = period of Ti,
 For a set 𝒆of periodic task {Ti}, the total utilization due to all tasks is given as
𝑼 = 𝒏𝒊=𝟏 𝒑𝒊
𝒊
 Objective of a good scheduling algorithm => feasibly schedule those task that have
task utilization = 1 (high utilization)
 On Uniprocessors tasks having utilization more than 1 can not be scheduled properly.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Task Scheduling: Terminologies
Jitter
 It is the deviation of a periodic task from its strict periodic behavior.

 Arrival Time Jitter => deviation of a the task from arriving at the precise
periodic time of arrival. Caused may be due to imprecise clocks or
factors like network congestions etc.

 Completion Time Jitter => may be caused by the specific scheduling

algorithm employed which takes up tasks as per convenience and the
load at an instant, rather than at some strict instance.

 Jitters are undesirable.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Scheduling Algorithms
Scheduling Algorithms

Clock Driven Event Driven Hybrid

Table Simple Rate Earliest Round

Cyclic
Driven Priority Monotonic Deadline Robin
Based Analysis First

Above Real-Time Task Scheduling Algorithms => based on how scheduling points are defined

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 20
 Scheduling Algorithms
Scheduling Algorithms
(based on type of task acceptance test )

Planning Best
based Effort

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 21
 Scheduling Algorithms
Scheduling Algorithms
(based on target platform on which tasks are to be run )

Uniprocessor Multiprocessor Distributed

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 22
 Clock Driven Scheduling
 Also known as offline schedulers =>
 a schedule is computed off-line and stored for use at runtime (i.e. fix the
schedule before the system starts to run)
 determines which job to execute when
 all the parameters of hard RT jobs are fixed and known.
 Scheduling decisions taken based on clock interrupt points
 Scheduling points determined by timer interrupts
 Advantage: Incur very little run time overhead

 Disadvantage: Can’t handle aperiodic and sporadic tasks

=> exact time of occurrence of tasks can not be predicted
=> static scheduler

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 23
 Clock Driven Scheduling
 Make scheduling decisions at regularly spaced time => frequently adopted
choice
 Use a hardware timer => makes scheduling decisions periodically
(scheduler is invoked by a timer)
 The timer is set to expire periodically without the intervention of the
scheduler.
 When the system is initialized, the scheduler selects and schedules the
job(s) that will execute until the next scheduling decision time and then
blocks itself waiting for the expiration of the timer.
 When the timer expires, the scheduler awakes and repeats these actions.
 static or cyclic
 predictable and deterministic
 multiple tables for different operation modes

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 24
 Table-Driven Scheduling
Usually precomputed which task would run when and store this
schedule in a table at the time the system is designed or
configured.
Application programmer selects his own schedule for the set of
tasks in the application and store the schedule in a schedule
table, used by the scheduler at run time.
Consider the table shown =>
How to decide size of the schedule table? => Important
Task Start Time in msec
If a set ST = {Ti} of n tasks is to be T1 0
scheduled, then the entries will be
given as LCM(p1, p2, p3,…pn) duration T2 3
in the schedule table T3 10
T4 12
T5 17

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 25
 Table-Driven Scheduling
Example 1: Consider the following example
T1 = (e1 = 5 ms, p1 = 20 ms)
T2 = (e2 = 20 ms , p2 = 100 ms)
T3 = (e3 = 30 ms, p3 = 250 ms)

Then, the schedule will repeat after every 500 ms. That is, LCM
of (20, 100, 250), with an assumption that Φi = 0.

LCM(p1, p2, p3,…pn) is called the major cycle of the set of tasks
ST.
A major cycle of a set of tasks is an interval of time
on the timeline such that in each major cycle, the
different tasks recur identically.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 26
 Table-Driven Scheduling
Theorem 1: The major cycle of a set of tasks ST = LCM(T1, T2,
T3,…Tn) is LCM(p1, p2, p3,…pn), even when the tasks have
arbitrary phasings.
Consider the figure shown below for a task Ti in the major cycle
First occurrence of Ti starts Φ time units from the start of major
cycle

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 27
 Table-Driven Scheduling
Theorem 1: Cont…
The major cycle ends x time units after the last (i.e. (k-1)th)
occurrence of Ti in the major cycle => must be same for each
major cycle
If M is size of each major cycle then for the task to repeat in
each major cycle => 𝑀 = 𝑘 − 1 𝑝𝑖 + Φ + 𝑥…..( I )
For Ti to have identical occurrence time in each major cycle, we
should have Φ + 𝑥 = 𝑝𝑖 , substituting in Eqn. ( I ), we get
𝑀 = 𝑘 − 1 𝑝𝑖 + 𝑝𝑖 = k*𝑝𝑖 => major cycle M contains an integral
multiple of 𝑝𝑖 , holds for each task in the major cycle.

Therefore, 𝑀 = LCM({p1, p2, p3,…pn}).

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 28
 Table-Driven Scheduling
Example 2: Consider the following example
We have to schedule the activities of a car control system. Let
C = e = worst case execution time
T = p (sampling) period
D = d = deadline
Speed measurement: C=4ms, T=20ms, D=20ms
ABS control: C=10ms,T=40ms, D=40ms
Fuel injection: C=40ms,T=80ms, D=80ms
Other software with soft deadlines e.g. audio, air condition etc.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 29
 Table-Driven Scheduling
Example 2: Cont…

LCM(20, 40, 80) => 80 ms => Shortest repeating cycle

All task instances are within the cycle

(ms)
Speed
ABS
Fuel Injection

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 30
 Cyclic Schedulers
Very popular in industry
Larger majority of all small embedded applications
manufactured presently => cyclic schedulers
Simple, efficient and easy to program
E.g. => Temperature Controller (used in computer-controlled air
conditioners) which periodically samples the temperature of a
room and maintains it at a preset value.
Repeats a precomputed schedule => to be stored only for one
major cycle
Each task in the task set (to be scheduled) repeats identically
in every major cycle
Major cycle is divided into one or more minor cycles
Each minor cycle is called a frame

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 31
 Cyclic Schedulers

Consider the example shown above

Major cycle has been divided into four minor cycles (frames)
Scheduling points occur at frame boundaries => task can start
executing only at the beginning of a frame
Frame boundaries defined through the interrupts generated by
a periodic timer
Assignment of tasks to frames stored in schedule table

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 32
 Cyclic Schedulers
Task Number Frame Number
T3 F1
Schedule table => T1 F2
T3 F3
T4 F2

Size of a frame should be chosen carefully and should satisfy

the following three constraints:
1. Minimum Context Switching:
 Imposed to minimize the no of context switches occurring during task
execution => task instance complete running within its assigned frame.
 A task might have to be suspended and restarted in a later frame when
a task doesn’t complete within its allocated time => would need some
processing overhead

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 33
 Cyclic Schedulers
1. Minimum Context Switching (Cont…):
 Select a frame size larger than the execution time of each task
 to avoid unnecessary task switches
 When a task starts at a frame boundary, it should be able to complete
within the same frame
The constraint is expressed as: max 𝑒𝑖 ≤𝐹
Where, 𝑒𝑖 = execution time of task 𝑇𝑖 and F = frame size
This constraint imposes a lower bound on frame size i.e. F should not be
smaller than max 𝑒𝑖 .

2. Minimization of Table Size:

 No of entries in the schedule table should be minimum in order to minimize the
storage requirement of the schedule table, as cyclic schedulers are used in
small embedded applications with very small storage capacity.
 Sol: minor cycle should squarely divide the major cycle => no fractional minor
cycles => major cycle contains integral no of minor cycles
𝑀 𝑀
 Represented as: 𝐹 = 𝐹 => If floor of M/F equals M/F, major cycle would
contain integral no of frames.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 34
 Cyclic Schedulers
3. Satisfaction of Task Deadline:
 Constraint: Between the arrival of a task and its deadline, there exist at
least one full frame (as shown in the figure below)

(a) (b)
 Let a task arrives after ∆𝑡 time units have passed since the last frame, Fig (b)
 The task can complete before its deadline iff 2𝐹 − ∆𝑡 ≤ 𝑑𝑖 or 2𝐹 ≤ 𝑑𝑖 + ∆𝑡
 min(∆𝑡) = minimum separation from the start of frame => worst case scenario
to miss a deadline by the task = gcd 𝐹, 𝑝𝑖 .
 The constraint is written as for every 𝑇𝑖 : 2𝐹 − gcd 𝐹, 𝑝𝑖 ≤ 𝑑𝑖 ; for one task

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 35
 Cyclic Schedulers
3. Satisfaction of Task Deadline (Cont…):
 The constraint defines an upper bound on frame size for a task .
 A task may miss its deadline if the frame size is any larger than the defined
upper-bound.
max gcd 𝐹,𝑝𝑖 +𝑑𝑖
 Hence for all tasks, the frame size must be smaller than 2

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 36
 Cyclic Schedulers
Example 3: A cyclic scheduler is to be used to run the following
set of periodic tasks on a uniprocessor where
T1 = (e1 = 1, p1 = 4) 1. max{ei} <= F

T2 = (e2 = 1, p2 = 5) F=2
T3 = (e3 = 1, p3 = 20)
2. Find Major Cycle = LCM (4, 5, 20, 20) =
T4 = (e4 = 2, p4 = 20) 20
Select an appropriate frame size
F can take values of
For F = 2 for all tasks T1 – T4 = Satisfied 2, 4, 5, 10, 20
For F = 4; Constraint is not satisfied
For F= 5; Constraint is not satisfied 2F –gcd(F, pi) <= di
For F = 10; Constraint is not satisfied
For F = 20; Constraint is not satisfied For F = 2 and Task T1
Hence only frame size of 2 is suitable for scheduling 2x2 – gcd(2,4) <= 4
4 – 2 =2 <= 4 => For p1 the selection is
satisfied….

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 37
 Cyclic Schedulers
Example 4: A cyclic scheduler is to be used to run the following
set of periodic tasks on a uniprocessor where
T1 = (e1 = 1, p1 = 4)
T2 = (e2 = 2, p2 = 5)
T3 = (e3 = 5, p3 = 20)

Determine a suitable frame size for the task set

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 38
 Cyclic Schedulers
Example 4: A cyclic scheduler is to be used to run the following
set of periodic tasks on a uniprocessor where
T1 = (e1 = 1, p1 = 4)
T2 = (e2 = 2, p2 = 5)
T3 = (e3 = 5, p3 = 20)

Determine a suitable frame size for the task set

For F = 5 or 10 or 20 and Tasks T1 = Not Satisfied => Split one task
Split
Constraint 1: F >= 2
T3.1 = (20, 1, 20)
=> Iterations (2, T1 or T2 or T3.1 or T3.2 or
T3.2 = (20, 2, 20)
T3.3) the condition is satisfied
T3.3 = (20, 2, 20)
Hence frame size of 2 is suitable for scheduling
Similarly, we can check for F = 4 (permissible
value of M is 2 and 4 only.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 39
 A Generalized Task Scheduler
Many practical applications consists of a mixture of several
periodic, aperiodic and sporadic tasks.

Since arrival times of aperiodic and sporadic tasks are

expressed statistically => they can’t be assigned to frames
without lowering the overall achievable utilization of the system

In a generalized scheduler

 A schedule is prepared for only periodic tasks (assigned to frames)
 aperiodic and sporadic tasks are scheduled in the slack times that may
be available in frames
 Slack time = time left in a frame after a periodic task completes its
execution
 Non-zero slack time exists iff => execution time < frame size for a task

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 40
 A Generalized Task Scheduler
Sporadic task is scheduled only if enough slack time available
for the arriving sporadic task before its deadline
Sporadic tasks are subjected to acceptance test on their
arrival => to check whether the task will be completed within
deadline (in the available slack times)
 If not possible => rejected and corresponding recovery routines are run
 Aperiodic tasks can be taken up for scheduling without
acceptance test
 since they do not have strict deadlines
 Best effort is made to complete them within the available slack times,
but no guarantee is assured for the completion time
Store the slack times in a table during acceptance test for
schedulability
Accept both sporadic and aperiodic tasks without acceptance
test => make best effort to meet respective deadlines

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 41
 Generalized Scheduler Pseudo Code
cyclic-scheduler () {
current-task T = Schedule-Table [k];
k = k + 1;
k = k mod N; // N = total no of tasks in the schedule table
dispatch-current-task(T);
schedule-sporadic-tasks( ); // Current task T completed early, sporadic task
; // can be taken up
schedule-aperiodic-tasks( ); // At the end of the frame, the running task is
; // preempted, if not complete
Idle ( ); // No task to run, idle
} Assumptions:
Precomputed schedule for a periodic task is stored in a schedule table.
Sporadic task those passed the acceptance test has been kept for
scheduling, if possible

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 42
 Cyclic vs Table-driven Schedulers
Cyclic Table-driven
Set a periodic timer only once at the A timer has to be set every time a task starts
application initialization time, which give an to run. A call to a timer is made every few
interrupt exactly at every frame boundary milliseconds which represents a overhead.
Since size of the frame that needs to be If overhead of setting a timer is ignored
chosen should be at least as long as the size then, table-driven is more efficient.
of the largest execution time of a task in the
task set. This is a source of inefficiency since
this results in processor time being wasted
in case of those tasks whose execution times
are smaller than the chosen frame size.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 43
 Clock-driven Schedulers
Example 5: Consider a real-time system which consists of three
tasks as shown in the table. If the tasks are to be scheduled
using a table driven scheduler, what is the length of time for
which the schedules have to be stored in the precomputed
schedule table of the scheduler?

Task Phase Execution Relative Period

(msec) Time Deadline (msec)
(msec) (msec)
T1 20 10 20 20
T2 40 10 50 50
T3 70 20 80 80

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 44
 Cyclic Schedulers
Example 6: A cyclic real-time scheduler is to be used to
schedule three periodic tasks as shown in the table. Suggest a
suitable frame size that can be used. Show all intermediate
steps in your calculations.

Task Phase Execution Relative Period

(msec) Time Deadline (msec)
(msec) (msec)
T1 0 20 100 100
T2 0 20 80 80
T3 0 30 150 150

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 45
 Cyclic Schedulers
Example 7: Consider the following three set of independent
real-time periodic tasks shown in the table below. Suppose a
cyclic scheduler is to be used to schedule the task set. What is
the major cycle of task set? Suggest a suitable frame size and
provide a feasible schedule (task to frame assignment for a
major cycle) for the task set.
Task Phase Processing Period Deadline
(msec) Time (msec) (msec)
(msec)
T1 20 25 150 100
T2 40 10 50 30
T3 60 50 200 150

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 46
 Hybrid Schedulers
Clock-driven Schedulers => scheduling points defined through
clock interrupts.

Event-driven Schedulers => scheduling points defined by

events such as arrival and completion of tasks.

Hybrid Schedulers => scheduling points defined both through

clock interrupts and the event occurrences.
Time-Sliced Round Robin Scheduling

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 47
 Time-Sliced Round Robin Schedulers
This is a preemptive scheduling methods.
The ready tasks are held in circular queue and are taken up
one after another in a sequence from the queue.
Once a task is taken up, it runs for a certain fixed interval of
time called its time slice.
Each process is given a time slice to run in
 When the time slice expires, the next process preempts the current
process, and runs for its time slice, and so on
The preempted process is placed at the end of the queue
 If a task does not complete within its allocated time slice, it is inserted
back into the circular queue.
Implemented with
 A ready queue
 A regular timer interrupt

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 48
 Time-Sliced Round Robin Schedulers

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 49
 Time-Sliced Round Robin Schedulers

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 50
 Time-Sliced Round Robin Schedulers

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 51
 Time-Sliced Round Robin Schedulers
 Pros
 Fair, easy to implement
 Con
 Assumes everybody is equal
 Issue: What should the time slice be?
 Too short
 Waste a lot of time switching between processes
 Example: time slice of 4 ms with 1 ms context switch = 20% round robin
overhead
 Too long
 System is not responsive
 Example: time slice of 100ms
 If 10 people hit “enter” key simultaneously, the last guy to run will only see
progress after 1 second.
 Degenerates into First Come First Serve (FCFS) if time slice longer than
burst length

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 52
 Event-Driven Schedulers
Event Driven

Simple Rate Earliest

Priority Monotonic Deadline
Based Analysis First
 Cyclic schedulers are very efficient
 However its difficult to determine a suitable frame size and a feasible
schedule when no of tasks increases.
 Additionally, some processing time is wasted in every frame (frame
size is larger than all tasks execution time).
 Event-driven schedulers overcome these shortcomings
 Can handle aperiodic and sporadic tasks proficiently
 However, they deploy complex scheduling algorithms => less efficient

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 53
 Event-Driven Schedulers
 Event-driven schedulers are less suitable for embedded
applications, as they should be of
 Small size
 Low cost
 Consume less power
 Can handle aperiodic and sporadic tasks proficiently

 Event-driven schedulers are used in all moderate and large-

sized applications having many tasks

 Event-driven schedulers are normally preemptive

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 54
 Foreground-Background Schedulers
 It is the simplest priority-driven preemptive scheduler
 Real-Time tasks run as foreground tasks
 Foreground tasks => highest priority tasks run at every
scheduling points
 Background tasks run at lowest priority => when no
foreground task running

 Consider a certain real-time system, where ‘n’ foreground

tasks are there denoted as T1, T2, T3,……,Tn.
 Let TB be the only background task
 Let eB be the processing time requirement of TB.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 55
 Foreground-Background Schedulers
 When any foreground task is executing, the background task
waits.
 The average CPU utilization due to the foreground task Ti is
𝑒
equal to 𝑖 , since ei amount of processing is required over
𝑝𝑖
every pi period.
 All foreground tasks together would result in CPU utilization
𝑒
of 𝑛𝑖=1 𝑖
𝑝𝑖
 Therefore, the average time available for execution of the
𝑒
background tasks in every unit of time is 1 - 𝑛𝑖=1 𝑖
𝑝𝑖
 Hence, the completion time for the background task is given
𝑒𝐵
by 𝑐𝑡𝐵 =
1 − 𝑛𝑖=1𝑝𝑒𝑖
𝑖

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 56
 Foreground-Background Schedulers
Example 8: Consider a real-time system in which tasks are
scheduled using foreground-background scheduling. There is
only one foreground task
𝑻𝒇 = (𝚽𝒇 = 𝟎, 𝒑𝒇 = 𝟏𝟎𝟎 𝒎𝒔𝒆𝒄, 𝒆𝒇 = 𝟓𝟎 𝒎𝒔𝒆𝒄, 𝒅𝒇 = 𝟏𝟎𝟎 𝒎𝒔𝒆𝒄),
and the background task be 𝑻𝑩 = (𝒆𝑩 = 𝟏𝟎𝟎𝟎 𝒎𝒔𝒆𝒄). Compute
the completion time for the background task.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 57
 Foreground-Background Schedulers
Example 8: Consider a real-time system in which tasks are
scheduled using foreground-background scheduling. There is
only one foreground task
𝑻𝒇 = (𝚽𝒇 = 𝟎, 𝒑𝒇 = 𝟏𝟎𝟎 𝒎𝒔𝒆𝒄, 𝒆𝒇 = 𝟓𝟎 𝒎𝒔𝒆𝒄, 𝒅𝒇 = 𝟏𝟎𝟎 𝒎𝒔𝒆𝒄),
and the background task be 𝑻𝑩 = (𝒆𝑩 = 𝟏𝟎𝟎𝟎 𝒎𝒔𝒆𝒄). Compute
the completion time for the background task.
Solution:
𝑒𝐵 1000
𝑐𝑡𝐵 = 𝑛 𝑒𝑖
= 50 𝑚𝑠𝑒𝑐 = 2000 𝑚𝑠𝑒𝑐
1 − 𝑖=1𝑝 1−100
𝑖

So background task 𝑻𝑩 would take 2000 msec to complete.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 58
 Foreground-Background Schedulers
Example 9: In a simple priority-driven preemptive scheduler, two
periodic tasks T1 and T2 and a background tasks are scheduled.
The periodic tasks T1 has the highest priority and executes once
every 20 msec and requires 10 msec of execution time each
time. T2 requires 20 msec of processing every 50 msec. T3 is a
background task and requires 100 msec to complete. Assuming
that all the tasks start at time 0, determine the time at which T3
will complete.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 59
 Foreground-Background Schedulers
Example 9: Given
T1 = (e1 = 10 ms, p1 = 20 ms)
T2 = (e2 = 20 ms, p2 = 50 ms)
T3 = (e3 = 100 ms) => background task
Solution:
2 𝑒𝑖 10 20 90
The total utilization due to foreground task 𝑖=1 𝑝 = + =
𝑖 20 50 100
That is, fraction of time remaining for the background task to
𝑒 90 10 1
execute is given by 1 − 𝑛𝑖=1 𝑖 = 1 - = =
𝑝𝑖 100 100 10
=> background task gets 1 ms every 10 ms .
10
=> So background task 𝑻𝟑 would take 100 ∗ = 1000 ms to
1
complete.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 60
 Foreground-Background Schedulers
Example 10: Assume that in example 8 an overhead of 1 ms on
account of every context switch is to be taken into account.
Compute the completion time of TB.

The very first time the foreground task runs (at t = 0), it incurs a
context switch overhead of 1 ms. In every preempt cycle it incurs
one context switch.
The background task runs after every instance of the foreground
tasks and incurs one context switch too => the execution time of
every foreground task is increased by two context switch times
i.e. from 50 ms to 52 ms.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 61
 Foreground-Background Schedulers
Example 10:
Therefore the completion time of TB.

𝑒𝐵 1000
𝑐𝑡𝐵 = 𝑛 𝑒𝑖
= 𝑚𝑠𝑒𝑐 = 2083.4 𝑚𝑠𝑒𝑐
1− 𝑖=1𝑝 1−
52
100
𝑖

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 62
 Priorities
 Each Process (or thread) is associated with a priority
 Provides basic mechanism to influence a scheduler
decision:
 Scheduler will always chooses a thread of higher priority over lower
priority
 Priorities can be defined internally or externally
 Internal: e.g. I/O bound or CPU bound
 External: e.g. based on importance to the user
CPU-Bound process I/O-Bound process
Spends most of its computing Spend most of its time waiting for I/O to complete
 Small bursts of CPU to process I/O and request
next I/O
Time to completion largely determined Time to completion largely determined by I/O request
by received CPU time time

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 63
 Priorities: Example
 Consider 5 Jobs as shown
 Job number equals priority
 Priority 1 > priority 5
 Release and execution
times as shown
 Priority-driven preemptively
scheduled

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 64
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 65
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 66
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 67
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 68
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 69
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 70
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 71
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 72
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 73
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 74
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 75
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 76
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 77
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 78
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 79
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 80
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 81
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 82
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 83
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 84
 Priorities: Example

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 85
 Priorities: Example

 Usually implemented by multiple priority queues, with round

robin on each queue
 Con
 Low priorities can starve
 Need to adapt priorities periodically
 Based on ageing or execution history

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 86
 Online Scheduling

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Schedulability Test
Test to determine whether a feasible schedule exists
Sufficient Test
 If test is passed, then tasks are definitely schedulable
 If test is not passed, tasks may be schedulable, but not necessarily
Necessary Test
 If test is passed, tasks may be schedulable, but not necessarily
 If test is not passed, tasks are definitely not schedulable
Exact Test (= Necessary + Sufficient)
 The task set is schedulable if and only if it passes the test.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Earliest Deadline First (EDF) Scheduling
In Earliest Deadline First (EDF) Scheduling, at every
scheduling point the task with shortest deadline is taken up
for scheduling.
A task is schedulable under EDF if and only if it satisfies the
condition that the total processor utilization due to the task
set is less than 1.
For a set of periodic tasks {T1, T2, T3, …., Tn}, the EDF
schedulability criterion is given as:
𝑛 𝑒𝑖 𝑛
𝑖=1 =
𝑝𝑖 𝑖=1 𝑢𝑖 ≤ 1 => pi of each task same as di (assumption)

Where ui is the average utilization due to task Ti

And ‘n’ is the total number of tasks in the task set
Note: The above expression is both necessary and sufficient
for a set of tasks to be EDF schedulable.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 Earliest Deadline First (EDF) Scheduling
EDF is optimal scheduling algorithm => if a set of tasks is
unschedulable under EDF, then no other scheduling
algorithm can feasibly schedule this task set.
In practical problems the period pi of a task may be different
than its deadline di => schedulability test needs to be
changed.
If pi > di => each task needs ei amount of computing time
every min(pi, di) duration of
𝑒
time, therefore the expression
could be written as 𝑛𝑖=1 𝑖
≤1
min(𝑝𝑖 , 𝑑𝑖 )
If pi < di => it is possible that a set of task is EDF
schedulable, even when the task set fails to meet the
expression as above. For this case, the above condition is
not a necessary condition but only a sufficient condition for a
task to be EDF schedulable.

11 May 2021 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh
 EDF Scheduling
Example 11: Given three periodic real-time tasks to be
scheduled using a EDF on a uniprocessor.
T1 = (e1 = 10 ms, p1 = 20 ms)
T2 = (e2 = 5 ms, p2 = 50 ms)
T3 = (e3 = 10 ms, p3 = 35 ms)
Determine whether the task set is schedulable

Solution:
The total utilization due to foreground task
3 𝑒𝑖 10 5 10
𝑖=1 𝑝 = + + =0.89 < 1 => the task set is EDF schedulable
𝑖 20 50 35

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 91
 Minimum Laxity First (MLF) Scheduling
Variant of EDF
At every scheduling point, a laxity value is computed for every
task in the system, and the task having minimum laxity is
executed first.
Laxity of a task measures the amount of time that would
remain if the task is taken up for execution next.
Unlike EDF, MLF considers the execution time of a task =>
measure of flexibility available for scheduling

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 92
 Implementation of EDF
Implementation 1:
Maintain all tasks in a queue that are ready for execution
Insert any freshly arriving task at the end of queue
Every node in the queue would contain the absolute deadline
of the task.
Scan the entire queue from beginning at the every preemption
point to determine the task having the shortest deadline
However, this implementation is inefficient because:
 Each task insertion would be achieved in O(1) or constant time
 But the task selection (to run next) and its deletion would require O(n)
time, where ‘n’ is the no of tasks in the queue.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 93
 Implementation of EDF
Implementation 2:
Maintain all ready tasks in a sorted priority queue (can be
efficiently implemented using a heap data structure)
In priority queue, tasks are always kept sorted according to the
proximity of their deadline.
Insert a record of any freshly arriving task into the heap in
O(log2n) time, where ‘n’ is the no of tasks in the priority queue.
At every scheduling point, the next task to be run can be found
at the top of the heap.
When a task is taken up for scheduling, it needs to be
removed from priority queue which could be achieved in O(1)
time.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 94
 Implementation of EDF
Implementation 3:
Assume that the no of distinct deadlines that the tasks can
have in an application are restricted.
Whenever task arrives, its absolute deadline is computed from
its release time and its relative deadline.
A separate FIFO queue is maintained for each distinct relative
deadline that tasks can have.
The scheduler inserts a newly arrived task at the end of the
corresponding relative deadline queue => tasks are ordered
according to their absolute deadlines.
To find a task with the earliest absolute deadlines => scheduler
only searches the threads of all FIFO queues.
If Q = no of priority queues maintained, then the order of
searching as well as time to insert a task would be O(1).

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 95
 Shortcomings of EDF
Transient Overload Problems:
Transient overload occurs when some tasks take more time to
complete than what was planned originally at the design time.
 Enter an infinite loop
 Encounter an unusual condition
 Enter a rarely used branch due to some abnormal input values
A task overshooting its completion time can cause some
task(s) to miss their deadlines.
Usually very difficult to predict which task might miss
deadlines.
Only prediction possible => the task(s) that would run
immediately after the tasks causing the transient overload
would get delayed and might miss its (their) deadlines.
EDF can’t guarantee about deadline misses => big drawback

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 96
 Shortcomings of EDF
Resource Sharing Problems:
While EDF is used for scheduling real-time tasks => high
overheads might have to be incurred to support resource
sharing among the tasks without making tasks to miss their
deadlines
Efficient Implementation Problems:
It is difficult to restrict the no of tasks with distinct deadlines to a
reasonable number.
 The efficient implementation that achieves O(1) overhead assumes that
the no of relative deadlines is restricted => sometimes unacceptable.
 The priority queue implementation [which uses O(log2n)] also creates
high runtime overhead, since most real-time tasks are periodic with
small periods and strict deadlines.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 97
 Next Lecture
 RMA Scheduling

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 98
 Bibliography
 K.C. Wang, “Embedded and Real-Time Operating Systems,” Springer
International Publishing, 2017.
 Rajib Mall, “Real Times Systems: Theory and Practice,” Pearson Education.
 Jonathan W. Valvano, “Embedded Systems: Real-Time Operating Systems for
Arm® Cortex™-M Microcontrollers,” CreateSpace Independent Publishing
Platform.
 Wolf, Wayne, Computers as Components – Principles of Embedded Computing
System Design, Second Edition, Morgan-Kaufmann, 2008.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh 99
.

Thank You!
©These lecture slides have been prepared for the benefits of the registered students of the course
Embedded Real Time Operating Systems (ECE 367) of Presidency University, Bengaluru. Anybody
who may wish to use these slides for educational purpose could use it without any explicit permission. Any
kind of redistribution, alteration, and commercial use of this material is strictly prohibited.

08 February 2020 Embedded Real Time Operating Systems (ECE 367) @Dr. Rajiv Ranjan Singh