Real-Time

Scheduling

Chenyang Lu
CSE 467S Embedded Compu5ng Systems
Readings
Single-Processor Scheduling: Hard Real-Time Computing Systems, by G.
Buttazzo.

q Chapter 4 Periodic Task Scheduling

q Chapter 5 (5.1-5.4) Fixed Priority Servers

q Chapter 7 (7.1-7.3) Resource Access Protocols



Optional further readings

q A Practitioner's Handbook for Real-Time Analysis: Guide to Rate Monotonic
Analysis for Real-Time Systems, by Klein et al.

q Deadline Scheduling for Real-Time Systems: EDF and Related Algorithms, by
Stankovic et al.

Real-Time Scheduling
What are the optimal scheduling algorithms?

How to assign priorities to processes?

Can a system meet all deadlines?

Benet of Scheduling Analysis
Schedulability analysis reduces development time by 50%!

Reduce wasted implementation/testing rounds

Analysis time << testing

More reduction expected for more complex systems

Quick exploration of design space!

VEST (UVA) Baseline (Boeing)

Design one processor 40 Design one processor 25
Implementation one processor 75
Scheduling analysis - MUF 1 Timing test 30
Design - two processors 25 Design - two processors 90
Implementation two processors 105
Scheduling analysis - DM/Offset 1 Timing test 20
Implementation 105
Total composition time 172 Total composition time 345

J.A. Stankovic, et al., VEST: An Aspect-Based Composi5on Tool for Real-Time Systems, RTAS 2003.
Consequence of Deadline Miss
Hard deadline

q System fails if missed.

q Goal: guarantee no deadline miss.

Soft deadline

q User may notice, but system does not fail.

q Goal: meet most deadlines most of the time.

 Comparison
General-purpose systems

q Fairnessto all tasks (no starvation)

q Optimize throughput

q Optimize average performance

Embedded systems

q Meet all deadlines.

q Fairness or throughput is not important

q Hard real-time: worry about worst case performance

Chenyang Lu
6

Terminology
Task

q Map to a process or thread

q May be released multiple times

Job: an instance of a task

Periodic task

q Ideal: inter-arrival time = period

q General: inter-arrival time >= period

Aperiodic task

q Inter-arrival time does not have a lower bound

Chenyang Lu
7

Timing Parameters

Task Ti

q Period Pi

q Worst-case execution time Ci

q Relative deadline Di

Job Jik

q Release time: time when a job is ready

q Response time Ri = finish time release time

q Absolute deadline = release time + Di

A job misses its deadline if

q Response time Ri > Di

q Finish time > absolute deadline

Chenyang Lu
8

Example
P1 = D1 = 5, C1 = 2; P2 = D2 = 7, C2 = 4.

Chenyang Lu
9

 Metrics
A task set is schedulable if all jobs meet their deadlines.

Optimal scheduling algorithm

q If a task set is not schedulable under the optimal algorithm, it is not
schedulable under any other algorithms.

Overhead: Time required for scheduling.

Chenyang Lu
10

Scheduling
Single Processor
OpCmal Scheduling Algorithms
Rate Monotonic (RM)

q Higher rate (1/period) Higher priority

q Optimal preemptive static priority scheduling algorithm

Earliest Deadline First (EDF)

q Earlier absolute deadline Higher priority

q Optimal preemptive dynamic priority scheduling algorithm

Chenyang Lu
12

Example

P1 = D1 = 5, C1 = 2; P2 = D2 = 7, C2 = 4.

Chenyang Lu
13

 AssumpCons
Single processor.

All tasks are periodic.

Zero context switch time.

Relative deadline = period.

No priority inversion.

RM and EDF have been extended to relax assumptions.

Chenyang Lu
14

Schedulable UClizaCon Bound

Utilization of a processor:

n
Ci
U =
i =1 Pi

n: number of tasks on the processor.

Utilization bound Ub: All tasks are guaranteed to be
schedulable if U Ub.

No scheduling algorithm can schedule a task set if U>1

Ub 1

An algorithm is optimal if its Ub = 1

Chenyang Lu
15

RM UClizaCon Bound
Ub(n) = n(21/n-1)

q n: number of tasks

q Ub(2) = 0.828

q Ub(n) Ub() = ln2 = 0.693

U Ub(n) is a sufficient condition, but not necessary.

Ub = 1 if all task periods are harmonic

q Periods are multiples of each other

q e.g., 1,10,100

Chenyang Lu
16

ProperCes of RM
RM may not guarantee schedulability even when CPU is not
fully utilized.

Low overhead: when the task set is fixed, the priority of a task
never changes.

Easy to implement on POSIX APIs.

Chenyang Lu
17

EDF UClizaCon Bound
Ub = 1

U 1: sufficient and necessary condition for schedulability.

Guarantees schedulability if CPU is not over-utilized.

Higher overhead than RM: task priority may change online.

Chenyang Lu
18

AssumpCons
Single processor.

All tasks are periodic.

Zero context switch time.

Relative deadline = period.

No priority inversion.

What if relative deadline < period?

Chenyang Lu
19

 OpCmal Scheduling Algorithms
RelaCve Deadline < Period

Deadline Monotonic (DM)

q Shorter relative deadline Higher priority

q Optimal preemptive static priority scheduling

Earliest Deadline First (EDF)

q Earlier absolute deadline Higher priority

q Optimal preemptive dynamic priority scheduling algorithm

Chenyang Lu
20

DM Analysis

Sufficient but pessimistic test

n
Ci 1/ n
n(2 -1)
i =1 Di

Sufficient and necessary test: response time analysis

Chenyang Lu
21

Response Time Analysis

Works for any xed-priority preemp5ve scheduling algorithm.

Cri5cal instant
results in a tasks longest response 5me.
when all higher-priority tasks are released at the same 5me.
Worst-case response 5me
Tasks are ordered by priority; T1 has highest priority

i 1
Ri
Ri = Ci + C j
j =1 Pj

Chenyang Lu
22

Response Time Analysis
Tasks are ordered by priority;
T1 has the highest priority.

for (each task Tj) {
I = 0; R = 0;
while (I + Cj > R) {
R = I + Cj;
if (R > Dj) return UNSCHEDULABLE;
j-1 R
I = k=1 P Ck;
k
}
}
return SCHEDULABLE;

Chenyang Lu
23

Example

P1 = D1 = 5, C1 = 2; P2 = D2 = 7, C2 = 4.

Chenyang Lu
24

EDF: Processor Demand Analysis
To start, assume Di = Pi

Processor demand in interval [0, L]: total time needed for
completing all jobs with deadlines no later than L.

n
L
CP (0, L) = Ci
i =1 Pi

Chenyang Lu
25

Schedulable CondiCon
Theorem: A set of periodic tasks is schedulable by EDF if
and only if for all L 0:

n
L
L Ci
i =1 Pi

There is enough time to meet processor demand at every

time instant.

Chenyang Lu
26

Busy Period Bp
End at the first time instant L when all the released jobs are
completed

W(L): Total execution time of all tasks released by L.



n
L
W ( L) = Ci
i =1 Pi

B p = min{L | W ( L) = L}

Chenyang Lu
27

 ProperCes of Busy Period
CPU is fully utilized during a busy period.

The end of a busy period coincides with the beginning
of an idle time or the release of a periodic job.

Chenyang Lu
28

Schedulable CondiCon
All tasks are schedulable if and only if

n
L
L Ci
i =1 Pi

at all job release times before min(Bp, H)

Chenyang Lu
29

 Compute Busy Period
busy_period
{
H = lcm(P1,,Pn); /* least common
multiple */
L = Ci;
L' = W(L);
while (L' != L and L' <= H) {
L = L';
L' = W(L);
}
if (L' <= H)
Bp = L;
else
Bp = INFINITY;
}

Chenyang Lu
30

Processor Demand Test: Di < Pi
A set of periodic tasks with deadlines no more than than
periods is schedulable by EDF if and only if

L Di
n
L D, L + 1 Ci

Pi
i =1

where D = {Di,k | Di,k = kPi+Di, Di,k min(Bp, H), 1in, k0}.

Note: only need to test all deadlines before min(Bp,H).

Chenyang Lu
31

Schedulability Test Revisited

D = P
D < P

Static Priority
RM
DM

Utilization bound
Response time

Response time



Dynamic Priority
EDF
EDF

Utilization bound
Processor demand

Chenyang Lu
32

AssumpCons
Single processor.

All tasks are periodic.

Zero context switch time.

Relative deadline = period.

No priority inversion.

Chenyang Lu
33

QuesCons
What causes priority inversion?

How to reduce priority inversion?

How to analyze schedulability?

Chenyang Lu
34

Priority Inversion
A low-priority task blocks a high-priority task.

Sources of priority inversion

q Access shared resources guarded by semaphores.

q Access non-preemptive subsystems, e.g., storage, networks.

Chenyang Lu
35

Semaphores
OS primitive for controlling access to shared variables.

q Get access to semaphore S with wait(S).

q Execute critical section to access shared variable.

q Release semaphore with signal(S).

Mutex: at most one process can hold a mutex.

wait(mutex_info_bus);
Write data to info bus;
signal(mutex_info_bus);

Chenyang Lu
36

What happened to Pathnder?
But a few days into the mission, not long after Pathfinder
started gathering meteorological data, the spacecraft began
experiencing total system resets, each resulting in losses of
data

Real-World (Out of This World) Story: Priority

inversion almost ruined the path nder mission
on MARS! hYp://[Link][.com/~mbj/

Chenyang Lu
37

 Priority Inversion

critical section

T1 blocked!

1 1 1

4 4 4 4
0 2 4 6 8 10 12 14 16 18 20 22

Chenyang Lu
38

 Unbounded Priority Inversion

critical section
T1 blocked by T4,T2,T3!

1 1 1

4 4 4 4 4
0 2 4 6 8 10 12 14 16 18 20 22

Chenyang Lu
39

 SoluCon

The low-priority task inherits the priority of the blocked

high-priority task.

critical section
T1 only blocked by T4

1 1 1
Inherit Return to
priority 1! 3
priority 4!
2

4 4 4 4
0 2 4 6 8 10 12 14 16 18 20 22
Chenyang Lu
40

Priority Inheritance Protocol (PIP)

When task Ti is blocked on a semaphore held by Tk

q If prio(Tk) is lower than prio(Ti), prio(Ti) Tk

When Tk releases a semaphore

q If Tk no longer blocks any tasks, it returns to its normal priority.

q If Tk still blocks other tasks, it inherits the highest priority of the
remaining tasks that it is blocking.

Priority Inheritance is transitive

q T2 blocks T1 and inherits prio(T1)

q T3 blocks T2 and inherits prio(T1)

Chenyang Lu
41

 How was Path Finder saved?
When created, a VxWorks mutex object accepts a boolean parameter that
indicates if priority inheritance should be performed by the mutex.

q The mutex in question had been initialized with the parameter FALSE.

VxWorks contains a C interpreter intended to allow developers to type in C

expressions/functions to be executed on the fly during system debugging.

The initialization parameter for the mutex was stored in global variables,
whose addresses were in symbol tables also included in the launch software,
and available to the C interpreter.

A C program was uploaded to the spacecraft, which when interpreted,

changed these variables from FALSE to TRUE.

No more system resets occurred.

Chenyang Lu
42

Bounded Number of Blocking
Assumptions of analysis

q Fixed priority scheduling

q All semaphores are binary

q All critical sections are properly nested

Task Ti can be blocked by at most min(m,n) times

q m: number of distinct semaphores that can be used to block Ti

q n: number of lower-priority tasks that can block Ti

Chenyang Lu
43

Extended RMS UClizaCon Bound
A set of periodic tasks can be scheduled by RMS/PIP if

i
Ck Bi 1/ i
i, 1 i n, + i ( 2 1)
k =1 Pk Pi
Tasks are ordered by priorities (T1 has the highest priority).

Bi: the maximum amount of time when task Ti can be blocked
by a lower-priority task.

Chenyang Lu
44

 Extended Response Time Analysis

Consider the effect of blocking on response time:

i 1 Ri
Ri = Ci + Bi + C j
j =1 Pj

The analysis becomes sufficient but not necessary.

Chenyang Lu
45

Priority Ceiling
C(Sk): Priority ceiling of a semaphore Sk

q Highest priority among tasks requesting Sk.

A critical section guarded by Sk may block task Ti only if C(Sk)

is higher than prio(Ti)

Chenyang Lu
46

 Compute Bi
Assumption: no nested critical sections.

/* potential blocking by other tasks */
B1=0; B2=0;
for each Tj with priority lower than Ti {
b1 = longest critical section in Tj that can block
Ti
B1 = B1 + b1
}

/* potential blocking by semaphores */
for each semaphore Sk that can block Ti {
b2 = longest critical section guarded by Sk among
lower priority tasks
B2 = B2 + b2
}
return min(B1, B2)

Chenyang Lu
47

Priority Ceiling Protocol

Priority ceiling of the processor: The highest priority ceiling

of all semaphores currently held.

A task can acquire a resource only if

q the resource is free, AND

q it has a higher priority than the priority ceiling of the system.

A task is blocked by at most one critical section.

Higher run-time overhead than PIP.

Chenyang Lu
48

AssumpCons
Single processor.

All tasks are periodic.

Zero context switch time.

Relative deadline = period.

No priority inversion.

Chenyang Lu
49

Hybrid Task Set
Periodic tasks + aperiodic tasks

Problem: arrival times of aperiodic tasks are unknown

Sporadic task with a hard deadline

q Inter-arrival time must be lower bounded

q Schedulability analysis: treated as a periodic task with period =
minimum inter-arrival time can be very pessimistic.

Aperiodic task with a soft deadline

q Possibly unbounded inter-arrival time

q Maintainhard guarantees on periodic tasks

q Reduce response time of aperiodic tasks

Chenyang Lu
50

Background Scheduling
Handle aperiodic requests with the lowest-priority task

Advantages

q Simple

q Aperiodic tasks usually has no impact on periodic tasks.

Disadvantage

q Aperiodic tasks have very long response times when the utilization of
periodic tasks is high.

Acceptable only if

q System is not busy

q Aperiodic tasks can tolerate long delays

Chenyang Lu
51

Polling Server
A periodic task (server) serves aperiodic requests.

q Period: Ps

q Capacity: Cs

Released periodically at period Ps

Serves any pending aperiodic requests

Suspends itself until the end of the period if

q it has used up its capacity, or

q no aperiodic request is pending

Capacity is replenished to Cs at the beginning of the next period

Chenyang Lu
52

Example: Polling Server

Chenyang Lu
53

Schedulability
Polling server has the same impact on periodic tasks as a
periodic task.

q n tasks with m servers: Up + Us Ub(n+m)

Disadvantage: If an aperiodic request misses the server, it

has to wait till the next period. long response time.

Can have multiple servers (with different periods) for different

classes of aperiodic requests

Chenyang Lu
54

Deferrable Server (DS)
Preserve unused capacity till the end of the current period
shorter response to aperiodic requests.

Impact on periodic tasks differs from a periodic task.

Chenyang Lu
55

Example: Deferrable Server

Chenyang Lu
56

RM UClizaCon Bound with DS
Under RMS

U s + 2
U b = U s + ln
2U s + 1
As n :

U + 2 1/ n
U b = U s + n s 1
2U s + 1

When Us = 0.186, min Ub = 0.652

U s + 2
System is schedulable if
U p ln
2U s + 1
Chenyang Lu
57

DS: Middleware ImplementaCon
First DS implementation on top of priority-based OS (e.g., Linux, POSIX)

Server thread processes aperiodic events (2nd highest priority)

Budget manager thread (highest priority) manages the budget and controls the
execution of server thread

Replenish Timer
High Priority
Budget
Manager
Thread
ACE Timer Queue
Aperiodic Events Budget Exhausted Timer
Server
Thread
Kokyu Dispatching Queue
Periodic Events
Dispatching
Thread
Kokyu Dispatching Queue
Periodic Events
Dispatching
Thread
Low Priority Kokyu Dispatching Queue Y. Zhang, C. Lu, C. Gill, P. Lardieri, G. Thaker, Middleware Support
for Aperiodic Tasks in Distributed Real-Time Systems, RTAS'07.

Chenyang Lu
58

AssumpCons
Single processor.

All tasks are periodic.

Zero context switch time.

Relative deadline = period.

No priority inversion.

Chenyang Lu
59

Context Switch Time
RTOS usually has low context switch overhead.

Context switches can still cause overruns in a tight schedule.

q Leave margin in your schedule.

Techniques exist to reduce number of context switches by

avoiding certain preemptions.

Other forms of overhead: cache, thread migration, interrupt

handling, bus contention, thread synchronization

Chenyang Lu
60

Fix an Unschedulable System
Reduce task execution times.

Reduce blocking factors.

Get a faster processor.

Replace software components with hardware.

Multi-processor and distributed systems.

Chenyang Lu
61

Final
1-2:30 April 21st

Open book/note

Scope: Operating Systems, Real-Time Scheduling

62

Final Demo
April 23rd, 1pm-2:30pm

20 min per team

Set up and test your demo in advance

All expected to attend the whole session

Return devices to Rahav

Itll be fun! J

63

Project Report
Submit report and materials by 11:59pm April 30th.

Email to Rahav

Report

q Organization: See conference papers in the reading list.

q 6 pages, double column, 10 pts fonts.

q Use templates on the class web page.

Other materials

q Slides of your final presentation

q Source code

q Documents: README, INSTALL, HOW-to-RUN

q Video (Youtube is welcome!)

64

Suggested Report Outline
Abstract

Introduction

Goals

Design: Hardware and Software

Implementation

Experiments

Related Work

Lessons Learned

Conclusion and Future Work

65

Peer Review
For fairness in project evaluation.

Email me individually by 11:59pm, April 30th

q Estimated percentage of contribution from each team member.

q Brief justification.

66