Basics of Real-Time Scheduling

1
Basics of Real-Time Scheduling

• Jan Madsen
• Informatics and Mathematical Modelling
• Technical University of Denmark
• Richard Petersens Plads, Building 321
• DK2800 Lyngby, Denmark
• Jan_at_imm.dtu.dk

2
WAP application

• Tasks

- send requests to server
- translate data received from server
- render data on display
- operating the users control panel

• Why?

- logically different operations
- performed at different rates
Need structure and to perform multiple processes
3
Why use multiple processes?

• Processes introduce structure into the system
  design

• understand each process independently

• analyze communication between processes to
  determine system behavior

• Timing is easier to specify

• Communication between processes provide test
  points

• Suitable for team effort

Processes are a neat way to handle complex systems
4
What is a process?

• A process is a state and a procedure
• State
• Running
• Suspended
• Deactivated
• Procedure
• Sequence of separate steps
• A process may interact with its surroundings

5
Process execution
6
Process execution

• Periodic
• the period is the time between successive
  executions
• the process rate is the inverse of its period

• Aperiodic
• non-periodic process, typical an event handler

• Sporadic
• hard real-time aperiodic process
• usually, a minimum interarrival constraint is
  imposed

7
Example A software processes

• A process includes
• code
• memory which belongs to the process
• connections to other processes
• A process is a unique execution of a program
• In most embedded systems, processes never die!

8
Executing on a single CPU
9
Execution of multiple processes

• On a single CPU
• A Pentium processor

• On a distributed homogeneous system
• A network of Pentium processors

• On a distributed heterogeneous system
• A network of differing processors (CPUs and ASICs)

10
About this lecture

• Huge literature on scheduling theory
• Aim is to give an overview of problems and
  possible solutions to the multi-task scheduling
  problem related to RTOS
• Structure of lecture
• Some basic concepts and teminology
• Uni-processor scheduling
• Rate-monotonic scheduling
• Earlies-deadlines-first scheduling
• Multi-processor scheduling

11
PART I Scheduling Terminology and Basic Concepts
12
Design flow
Partitioning/clustering
P1
P2
Allocation
Mapping
P3
Scheduling
Communication
P1
P3
P2
13
Scheduling
14
Terminology

• Architecture

Application

• Application

• Scheduling

Architecture

• Constraints

15
Architecture

• Components
• Processors
• CPU, DSP, ASIC, ...
• Communication
• Bus, network, ...
• Devices
• sensor, actuator, display, ...
• Architecture may be fixed or (re-)configurable

os
device
a
b
c
mem
device
16
Application

• Task
• A module which can be envoked to perform a
  particular function
• A schedulable entity
• Characterized by its
• Timing constraints
• Precedence contraints
• Exclusion constraints
• Resource requirements

17
Timing constraints
1
18
Task execution

• Periodic
• the period is the time between successive
  executions
• Aperiodic
• non-periodic task, typical an event handler
• Sporadic
• hard real-time aperiodic task
• typical, a minimum interarrival constraint is
  imposed

19
Precedence constraints

• pi precede pj if pi must finish before pj begins
• process interrelate with each other to achieve
• synchronization
• exchange data

pi
pj
20
Exclusion constraints

• pi exclude pj if no execution of pj is allowed
  between the time that pi starts its computation
  and the time that pi complete its computation
• Prevent simultaneous access to shared resources
• data
• memory
• devices

pi
pj
21
Resource constraints

• R r1, r2, , rm
• Example
• R(pi) r1, r3 resource r1 and r3 is required
  by pi during its execution
• may be used to
• implement a critical section
• enforce PE assignment of processes

22
Scheduling approaches

• Classical scheduling theory
• Uniprocessor systems
• RMS
• EDF
• Distributed real-time scheduling
• Deterministic (static) scheduling
• Dynamic scheduling

23
Scheduling

• Allocation
• Determine number and type of processors/resources
• Assignment
• Binding tasks to processors
• Scheduling
• Determine execution order

24
Scheduling principles

• Off-line
• A scheduler performing its job before the
  scheduled system is put into operation
• On-line
• A scheduler performing its job at run-time, when
  the system is running
• Typically list-based
• When a task is released (ready) it is placed in a
  list
• Scheduler select which task from the list to
  execute next
• Selection based on some criteria scheduling
  policy

25
Scheduling areas

• High-level Synthesis
• fine grained, operation/instruction level
• schedule done off-line
• single time constraint
• Real-time operating systems (RTOS)
• coarse grained, programs/sub-programs
• scheduling done on-line
• multiple time constraints

26
SoC scheduling

• Similarity to RTOS
• Time constraints
• Context switching
• Task synchronization
• Task communication

• Difference to RTOS
• Large design space
• Variable granularity
• Wide range of metrics
• Time, power, cost, ...
• Flexibility, reliability, ...
• High degree of optimization
• Dedicated hardware

27
Preemption vs. Non-preemption

• Non-preemptive scheduling
• Preemptive scheduling

28
Metrics

• Minimize

Schedule length