Multiprocessor and RealTime Scheduling

1
Multiprocessor and Real-Time Scheduling

• Fred Kuhns

2
Classifications of Multiprocessor

• Loosely coupled multiprocessor
• each processor has its own memory and I/O
  channels Clusters
• Functionally specialized processors
• such as I/O processor
• controlled by a master processor
• Tightly coupled multiprocessing
• processors share main memory
• controlled by operating system

3
Synchronization Granularity

• Independent Parallelism - No explicit
  synchronization
• Course and Very Coarse Parallelism - similar to
  running many processes on one processor
• good when infrequent interaction among processes
• Medium Parallelism - Parallel processing or
  multitasking within a single application
• requires relatively frequent synchronization
• Fine Parallelism
• Parallelism inherent in a single instruction

4
Scheduling Design Issues

• Assignment of processes to processors
• Use of multiprogramming on individual processors
• Actual dispatching of a process

5
Processor Assignment

• Processor Pool - Treat processors as a pooled
  resource and assign process to processors
• Static assignment
• Dynamic assignment
• Dispatch queue arrangements
• Per Processor Queue - Dedicate short-term queue
  for each processor
• Global queue - Schedule to any available
  processor

6
Basic Design Architectures

• Peer architecture
• Operating system can execute on any processor
• Each processor does self-scheduling
• Complicates the operating system
• Make sure two processors do not choose the same
  process
• Master/slave architecture
• Key kernel functions always run on a particular
  processor
• Master is responsible for scheduling
• Slave sends service request to the master
• Disadvantages
• Failure of master brings down whole system
• Master can become a performance bottleneck

7
Threads

• Executes separate from the rest of the process
• An application can be a set of threads that
  cooperate and execute concurrently in the same
  address space
• Threads running concurrently on separate
  processors may yield dramatic gain in performance

8
Multiprocessor Thread Scheduling

• Load sharing
• global (thread) Ready Queue
• Dedicated processor assignment
• threads assigned to specific processor
• Dynamic scheduling
• number of threads altered during execution
• Gang scheduling
• a set of related threads co-scheduled on set of
  processors at same time

9
Load Sharing

• Load is distributed evenly across the processors
• Assures no processor is idle
• No centralized scheduler required
• Use global queues

10
Disadvantages of Load Sharing

• Central queue needs mutual exclusion
• may be a bottleneck when more than one processor
  looks for work at the same time
• Preemptive threads are unlikely to resume
  execution on the same processor
• cache use is less efficient
• unlikely that threads of one program will gain
  access to the processors simultaneously

11
Gang Scheduling

• Simultaneous scheduling of threads that make up a
  single process
• Useful for applications where performance
  severely degrades when any part of the
  application is not running
• Threads often need to synchronize with each other

12
Dedicated Processor Assignment

• When application is scheduled, its threads are
  assigned to a processor
• Some processors may be idle
• Avoids process switching

13
Dynamic Scheduling

• Number of threads in a process are altered
  dynamically by the application
• Operating system adjusts load to improve use
• assign idle processors
• new arrivals may be assigned to a processor that
  is used by a job currently using more than one
  processor
• hold request until processor is available
• new arrivals will be given a processor before
  existing running applications

14
What is a Real-Time System?

• Real-time systems have been defined as "those
  systems in which the correctness of the system
  depends not only on the logical result of the
  computation, but also on the time at which the
  results are produced"
• J. Stankovic, "Misconceptions About Real-Time
  Computing," IEEE Computer, 21(10), October 1988.

15
Real-Time Characteristics

• Real-time systems often are comprised of a
  controlling system, controlled system and
  environment.
• Controlling system acquires information about
  environment using sensors and controls the
  environment with actuators.
• Timing constraints derived from physical impact
  of controlling systems activities. Hard and soft
  constraints.
• Periodic Tasks Time-driven recurring at regular
  intervals.
• Aperiodic Event-driven.

16
Typical Real-Time System
Controlled System
sensor
Controlling System
Environment
sensor
sensor
sensor
actuator
actuator
actuator
actuator
17
Some Definitions

• Timing constraint constraint imposed on timing
  behavior of a job hard or soft.
• Release Time Instant of time job becomes
  available for execution. If all jobs are
  released when the system begins execution, then
  there is said to be no release time
• Deadline Instant of time a job's execution is
  required to be completed. If deadline is
  infinity, then job has no deadline. Absolute
  deadline is equal to release time plus relative
  deadline
• Response time Length of time from release time
  to instant job completes.

18
Hard versus Soft

• Hard failure to meet constraint is a fatal
  fault. Validated system always meets timing
  constraints.
• Deterministic constraints
• Probabilistic constraints
• Constraints in terms of some usefulness function.
• Soft late completion is undesirable but
  generally not fatal. No validation or only
  demonstration job meets some statistical
  constraint. Occasional missed deadlines or
  aborted execution is usually considered
  tolerable. Often specified in probabilistic terms

19
Real-Time Systems

• Control of laboratory experiments
• Process control plants
• Robotics
• Air traffic control
• Telecommunications
• Games

20
Schedules and Scheduling

• Jobs scheduled and allocated resources based on a
  set of scheduling algorithms and access control
  protocols.
• Scheduler Module implementing scheduling
  algorithms
• Schedule assignment of all jobs to available
  processors, produced by scheduler.
• Valid schedule
• every processor assigned to at most one job at a
  time
• every job assigned to at most one processor at a
  time
• no job scheduled before its release time
• Total amount of processor time assigned to every
  job is equal to its maximum or actual execution
  time

21
Definitions

• Feasible schedule Every job starts at or after
  release time and completes by deadline
• Schedulable set of jobs schedulable according to
  an algorithm if the it always produces a feasible
  schedule.
• Optimal Scheduling algorithm optimal if it
  always produces a feasible schedule if such a
  schedule exists
• Tardiness Zero if completion time lt deadline,
  otherwise gt 0 (complete - deadline).
• Lateness difference between completion time and
  deadline, can be negative if early.

22
Performance Measures

• Miss rate percentage of jobs executed but
  completed late
• Loss rate percentage of jobs discarded
• Invalid rate sum of miss and loss rate.
• makespan If all jobs have same release time and
  deadline, then makespan response time of the
  last job to execute.
• Max or average response times
• Max or average tardiness/lateness

23
Real-Time Scheduling

• Considerations
• schedulability analysis performed
• static or dynamic schedulability analysis
• analysis result includes schedule
• Scheduling Paradigms
• static table-driven - static analysis and
  schedule
• static priority driven - static analysis, no
  (explicit) schedule, priorities.
• dynamic planning based - feasibility check at
  run-time, schedule produced
• dynamic best effort - No feasibility check,
  attempts to meet deadlines but no guarantee.

24
RTOS

• Primary function areas
• Process management and synchronization
• Memory management
• Interprocess communication
• I/O
• Categories of RTOS
• small proprietary
• RT extensions to commercial timesharing systems
• research operating systems

25
Small Proprietary OSes

• Typically, kernel provides
• fast context switch, small size, quick interrupt
  response time (predictable?), minimize interrupt
  disable times, memory partitions (no virtual
  memory, why?), special sequential files (why?)
• Timing
• bounded execution time of most primitives
• real-time clock
• priority scheduling mechanism
• special alarms and timeout

26
RT Extensions to Commercial OSes

• Advanced SW development environments
• Problems
• optimized for average case
• assigns resources on demand
• ignore application specific information
• independent CPU scheduling and resource allocation

27
Research Operating System

• Develop extended or new process models
• RT synchronization mechanisms
• facilitate timing analysis
• fault tolerance
• multiprocessor or distributed support
• real-time micro-kernel

28
Clock Driven

• Scheduling time instants When scheduling
  decisions are made
• clock-driven - instants predefined offline
• table-driven - table precomputed offline

Scheduling time instants
29
Weighted Round Roben

• Round Robin Scheduling - time-sharing systems
• Weighted Round Robin associate weight with each
  job.
• weight w number of time slices
• adjusting weight allocates CPU shares

30
Priority Driven

• Work-conserving The highest priority, runnable
  job is always dispatched to available CPU
• resources are never idle while runnable jobs
• Priority list, preemptions and other rules
  describe algorithm completely
• Example algorithms EDF, FIFO, LIFO etc.
• preemption versus non-preemption

31
Periodic Task Parameters
Utilization U C / T
Execution Time C
Idle
Task P
Processing
Processing
Task Period T
32
Deadline Scheduling

• Real-time applications are not concerned with
  speed but with completing tasks
• Scheduling tasks with the earliest deadline
  minimize the fraction of tasks that miss their
  deadlines
• includes new tasks and amount of time needed for
  existing tasks

33
Deadline Scheduling

• Information used
• Ready time
• Starting deadline or Completion deadline
• Processing time
• not used, given or OS measures an exponential
  average
• Resource requirements
• Priority
• Relative importance of task
• Subtask scheduler
• mandatory (hard deadline) or optional subtask

34
Deadline Scheduling

• For a specific preemption strategy and deadline
  (starting or completion), scheduling task with
  earliest deadline minimizes missed deadlines.
• Starting deadlines - non-preemptive
• Ending deadlines - preemptive
• Utilization for periodic tasks
• U (C1/T1) (Cn/Tn) ? 1

35
Scheduling of Real-Time Tasks
0
10
20
30
40
50
60
70
80
90
100
110
120
A
B
D
E
C
Arrival times
Requirements
Starting deadline
A
D
B
C
E
A
B
D
E
C
Arrival times
Service
Earliest deadline
A
C
E
D
C
A
B
E
(missed)
Starting deadline
D
36
Scheduling of Real-Time Tasks
0
10
20
30
40
50
60
70
80
90
100
110
120
A
B
D
E
C
Arrival times
Requirements
Starting deadline
A
D
B
C
E
A
B
D
E
C
Arrival times
Earliest deadline with unforced idle times
Service
A
B
C
E
D
C
A
B
E
Starting deadline
D
37
Scheduling of Real-Time Tasks
0
10
20
30
40
50
60
70
80
90
100
110
120
A
B
D
E
C
Arrival times
Requirements
Starting deadline
A
D
B
C
E
A
B
D
E
C
Arrival times
Service
First-come first-served (FCFS)
A
C
D
C
A
B
E (missed)
(missed)
Starting deadline
D
38
Rate Monotonic Scheduling (RMS)

• Parameters
• input Period T and Execution Time C
• derive Utilization U C/T
• Priority increases with decreasing period
• For n tasks, require
• (C1/T1) (Cn/Tn) ? n(21/n - 1)
• Note EDF U lt 1
• as n ? ?, n(21/n - 1) ? ln2 0.693

39
Rate Monotonic Bounds
N n(21/n - 1) 1 1.0 2 0.828 3
0.779 4 0.756 5 0.743 6 0.734 ? 0.693
40
UNIX Scheduling - Solaris

• Set of 160 priority levels divided into three
  priority classes
• Basic kernel is not preemptive

Scheduling Sequence
Priority Class
Global Value
159
first
.
.
.
Real-time
.
100
99
.
Kernel
.
60
59
.
.
.
Time-shared
.
last
0
41
Linux Scheduling

• Classes
• SCHED_FIFO First-in-first-out RT threads
• SCHED_RR Round-robin RT threads
• SCHED_OTHER non-RT threads