Advanced Operating

System
Professor Mangal Sain
 Lecture 3

Thread and CPU Scheduling

Lecture 3 – Part 1

Threads
MOTIVATION

 Most modern applications are multithreaded

 Threads run within application
 Multiple tasks with the application can be
implemented by separate threads
 Update display
 Fetch data
 Spell checking
 Answer a network request
 Process creation is heavy-weight while
thread creation is light-weight
 Can simplify code, increase efficiency
 Kernels are generally multithreaded
MULTITHREADED SERVER ARCHITECTURE
 BENEFITS

 Responsiveness – may allow continued

execution if part of process is blocked,
especially important for user interfaces
 Resource Sharing – threads share resources
of process, easier than shared memory or
message passing
 Economy – cheaper than process creation,
thread switching lower overhead than context
switching
 Scalability – process can take advantage of
multiprocessor architectures
MULTICORE PROGRAMMING

 Multicore or multiprocessor systems putting

pressure on programmers, challenges include:
 Dividing activities
 Balance
 Data splitting
 Data dependency
 Testing and debugging
 Parallelism implies a system can perform more
than one task simultaneously
 Concurrency supports more than one task
making progress
 Single processor / core, scheduler providing concurrency
 MULTICORE PROGRAMMING (CONT.)

 Types of parallelism
 Data parallelism – distributes subsets of the
same data across multiple cores, same operation
on each
 Task parallelism – distributing threads across
cores, each thread performing unique operation
 As # of threads grows, so does architectural
support for threading
 CPUs have cores as well as hardware threads
 Consider Oracle SPARC T4 with 8 cores, and 8
hardware threads per core
CONCURRENCY VS. PARALLELISM
Concurrent execution on single-core system:

Parallelism on a multi-core system:

SINGLE AND MULTITHREADED PROCESSES
 USER THREADS AND KERNEL THREADS
 User threads - management done by user-level threads library
 Three primary thread libraries:
 POSIX Pthreads
 Windows threads
 Java threads
 Kernel threads - Supported by the Kernel
 Examples – virtually all general purpose operating systems,
including:
 Windows
 Solaris
 Linux
 Tru64 UNIX
 Mac OS X
 MULTITHREADING MODELS

 Many-to-One

 One-to-One

 Many-to-Many
MANY-TO-ONE

 Many user-level threads

mapped to single kernel thread
 One thread blocking causes all
to block
 Multiple threads may not run in
parallel on muticore system
because only one may be in
kernel at a time
 Few systems currently use this
model
 Examples:
 Solaris Green Threads
 GNU Portable Threads
ONE-TO-ONE
 Each user-level thread maps to
kernel thread
 Creating a user-level thread creates
a kernel thread
 More concurrency than many-to-one

 Number of threads per process

sometimes restricted due to overhead
 Examples
 Windows
 Linux
 Solaris 9 and later
MANY-TO-MANY MODEL
 Allows many user level
threads to be mapped to
many kernel threads
 Allows the operating
system to create a sufficient
number of kernel threads
 Solaris prior to version 9

 Windows with the

ThreadFiber package
TWO-LEVEL MODEL

 Similar to M:M, except that it allows a

user thread to be bound to kernel thread
 Examples
 IRIX
 HP-UX
 Tru64 UNIX
 Solaris 8 and earlier
THREAD LIBRARIES
 Thread library provides programmer with
API for creating and managing threads
 Two primary ways of implementing
 Library entirely in user space
 Kernel-level library supported by the OS
PTHREADS

 May be provided either as user-level or kernel-

level
 A POSIX standard (IEEE 1003.1c) API for thread
creation and synchronization
 Specification, not implementation

 API specifies behavior of the thread library,

implementation is up to development of the library
 Common in UNIX operating systems (Solaris,
Linux, Mac OS X)
JAVA THREADS

 Java threads are managed by the JVM

 Typically implemented using the threads model
provided by underlying OS
 Java threads may be created by:

 Extending Thread class

 Implementing the Runnable interface
JAVA MULTITHREADED PROGRAM
JAVA MULTITHREADED PROGRAM (CONT.)
IMPLICIT THREADING

 Growing in popularity as numbers of threads

increase, program correctness more difficult with
explicit threads
 Creation and management of threads done by
compilers and run-time libraries rather than
programmers
 Three methods explored
 Thread Pools
 OpenMP
 Grand Central Dispatch
 Other methods include Microsoft Threading
Building Blocks (TBB),
java.util.concurrent package
Lecture 3 – Part 2

Threads
THREAD POOLS
 Create a number of threads in a pool where
they await work
 Advantages:
 Usually slightly faster to service a request with
an existing thread than create a new thread
 Allows the number of threads in the
application(s) to be bound to the size of the pool
 Separating task to be performed from mechanics
of creating task allows different strategies for
running task
 i.e.Tasks could be scheduled to run periodically
OPENMP
 Set of compiler directives and an API for
C, C++, FORTRAN
 Provides support for parallel
programming in shared-memory
environments
 Identifies parallel regions – blocks of
code that can run in parallel

#pragma omp parallel

Create as many threads as there are cores
#pragma omp parallel for
for(i=0;i<N;i++) {
c[i] = a[i] + b[i];
}
Run for loop in parallel
GRAND CENTRAL DISPATCH
 Apple technology for Mac OS X and iOS operating systems
 Extensions to C, C++ languages, API, and run-time library

 Allows identification of parallel sections

 Manages most of the details of threading

 Block is in “^{ }” - ˆ{ printf("I am a block"); }

 Blocks placed in dispatch queue

 Assigned to available thread in thread pool when removed from
queue
GRAND CENTRAL DISPATCH

 Two types of dispatch queues:

 serial – blocks removed in FIFO order, queue is per
process, called main queue
 Programmers can create additional serial queues within
program
 concurrent – removed in FIFO order but several may be
removed at a time
 Three system wide queues with priorities low, default, high
THREADING ISSUES

 Semantics of fork() and exec() system calls

 Signal handling
 Synchronous and asynchronous
 Thread cancellation of target thread
 Asynchronous or deferred
 Thread-local storage
 Scheduler Activations
 SEMANTICS OF FORK() AND EXEC()

 Does fork()duplicate only the calling

thread or all threads?
 Some UNIXes have two versions of fork
 exec() usually works as normal – replace
the running process including all threads
 SIGNAL HANDLING

n Signals are used in UNIX systems to notify a

process that a particular event has occurred.
n A signal handler is used to process signals
1. Signal is generated by particular event
2. Signal is delivered to a process
3. Signal is handled by one of two signal handlers:
1. default
2. user-defined
n Every signal has default handler that kernel
runs when handling signal
l User-defined signal handler can override
default
l For single-threaded, signal delivered to process
SIGNAL HANDLING (CONT.)
n Where should a signal be delivered for
multi-threaded?
l Deliver the signal to the thread to which the
signal applies
l Deliver the signal to every thread in the
process
l Deliver the signal to certain threads in the
process
l Assign a specific thread to receive all signals
for the process
 THREAD CANCELLATION
 Terminating a thread before it has finished
 Thread to be canceled is target thread

 Two general approaches:

 Asynchronous cancellation terminates the target
thread immediately
 Deferred cancellation allows the target thread to
periodically check if it should be cancelled
 Pthread code to create and cancel a thread:
THREAD CANCELLATION (CONT.)
 Invoking thread cancellation requests cancellation,
but actual cancellation depends on thread state

 If thread has cancellation disabled, cancellation

remains pending until thread enables it
 Default type is deferred
 Cancellation only occurs when thread reaches
cancellation point
 I.e. pthread_testcancel()
 Then cleanup handler is invoked
 On Linux systems, thread cancellation is handled
through signals
THREAD-LOCAL STORAGE

 Thread-local storage (TLS) allows each

thread to have its own copy of data
 Useful when you do not have control over the
thread creation process (i.e., when using a
thread pool)
 Different from local variables
 Local variables visible only during single
function invocation
 TLS visible across function invocations
 Similar to static data
 TLS is unique to each thread
SCHEDULER ACTIVATIONS
 Both M:M and Two-level models require
communication to maintain the appropriate
number of kernel threads allocated to the
application
 Typically use an intermediate data
structure between user and kernel threads
– lightweight process (LWP)
 Appears to be a virtual processor on which
process can schedule user thread to run
 Each LWP attached to kernel thread
 Scheduler activations provide upcalls - a
communication mechanism from the kernel
to the upcall handler in the thread library
 This communication allows an application
to maintain the correct number kernel
threads
 OPERATING SYSTEM EXAMPLES

 Windows Threads
 Linux Threads
WINDOWS THREADS

 Windows implements the Windows API – primary

API for Win 98, Win NT, Win 2000, Win XP, and Win
7
 Implements the one-to-one mapping, kernel-level

 Each thread contains

 A thread id
 Register set representing state of processor
 Separate user and kernel stacks for when thread runs in
user mode or kernel mode
 Private data storage area used by run-time libraries and
dynamic link libraries (DLLs)
 The register set, stacks, and private storage area are
known as the context of the thread
WINDOWS THREADS (CONT.)

 The primary data structures of a thread

include:
 ETHREAD (executive thread block) – includes
pointer to process to which thread belongs and
to KTHREAD, in kernel space
 KTHREAD (kernel thread block) – scheduling
and synchronization info, kernel-mode stack,
pointer to TEB, in kernel space
 TEB (thread environment block) – thread id,
user-mode stack, thread-local storage, in user
space
 LINUX THREADS
 Linux refers to them as tasks rather than threads
 Thread creation is done through clone() system call

 clone() allows a child task to share the address

space of the parent task (process)
 Flags control behavior

 struct task_struct points to process data

structures (shared or unique)
Lecture 3 – Part 3

CPU Scheduling
CONTENT

 Basic Concepts
 Scheduling Criteria

 Scheduling Algorithms

 Thread Scheduling

 Multiple-Processor Scheduling
BASIC CONCEPTS
 Maximum CPU
utilization obtained with
multiprogramming
 CPU–I/O Burst Cycle –
Process execution
consists of a cycle of
CPU execution and I/O
wait
 CPU burst followed by
I/O burst
 CPU burst distribution
is of main concern
CPU SCHEDULER
Short-term scheduler selects from among the
processes in ready queue, and allocates the CPU to one
of them
Queue may be ordered in various ways
CPU scheduling decisions may take place when a
process:
1. Switches from running to waiting state
2. Switches from running to ready state
3. Switches from waiting to ready
4. Terminates
Scheduling under 1 and 4 is nonpreemptive
All other scheduling is preemptive
Consider access to shared data
Consider preemption while in kernel mode
Consider interrupts occurring during crucial OS activities
DISPATCHER
 Dispatcher module gives control of the CPU to the
process selected by the short-term scheduler; this
involves:
 switching context
 switching to user mode
 jumping to the proper location in the user program to
restart that program
 Dispatch latency – time it takes for the
dispatcher to stop one process and start another
running
SCHEDULING CRITERIA
 CPU utilization – keep the CPU as busy as possible
 Throughput – # of processes that complete their
execution per time unit
 Turnaround time – amount of time to execute a
particular process
 Waiting time – amount of time a process has been
waiting in the ready queue
 Response time – amount of time it takes from when
a request was submitted until the first response is
produced, not output (for time-sharing environment)
SCHEDULING ALGORITHM OPTIMIZATION CRITERIA

 Max CPU utilization

 Max throughput

 Min turnaround time

 Min waiting time

 Min response time

FIRST- COME, FIRST-SERVED (FCFS) SCHEDULING
Process Burst Time
P1 24
P2 3
P3 3
 Suppose that the processes arrive in the order: P1 , P2 , P3
The Gantt Chart for the schedule is:

P1 P2 P3
0 24 27 30

 Waiting time for P1 = 0; P2 = 24; P3 = 27

 Average waiting time: (0 + 24 + 27)/3 = 17
FCFS SCHEDULING (CONT.)

Suppose that the processes arrive in the order:

P2 , P3 , P1
 The Gantt chart for the schedule is:

P2 P3 P1
0 3 6 30

 Waiting time for P1 = 6; P2 = 0; P3 = 3

 Average waiting time: (6 + 0 + 3)/3 = 3

 Much better than previous case

SHORTEST-JOB-FIRST (SJF) SCHEDULING

 Associate with each process the length of its next

CPU burst
 Use these lengths to schedule the process with the
shortest time
 SJF is optimal – gives minimum average waiting
time for a given set of processes
 The difficulty is knowing the length of the next CPU
request
 Could ask the user
 EXAMPLE OF SJF
ProcessArril Time Burst Time
P1 0.0 6
P2 2.0 8
P3 4.0 7
P4 5.0 3

 SJF scheduling chart

P4 P1 P3 P2
0 3 9 16 24

 Average waiting time = (3 + 16 + 9 + 0) / 4 = 7

EXAMPLE OF SHORTEST-REMAINING-TIME-FIRST
 Now we add the concepts of varying arrival times and preemption to the
analysis
ProcessAarri Arrival TimeTBurst Time
P1 0 8
P2 1 4
P3 2 9
P4 3 5
 Preemptive SJF Gantt Chart

P1 P2 P4 P1 P3
0 1 5 10 17 26

 Average waiting time = [(10-1)+(1-1)+(17-2)+5-3)]/4 = 26/4 = 6.5 msec

ROUND ROBIN (RR)
 Each process gets a small unit of CPU time (time
quantum q), usually 10-100 milliseconds. After this
time has elapsed, the process is preempted and added
to the end of the ready queue.
 If there are n processes in the ready queue and the
time quantum is q, then each process gets 1/n of the
CPU time in chunks of at most q time units at once.
No process waits more than (n-1)q time units.
 Timer interrupts every quantum to schedule next
process
 Performance
 q large  FIFO
 q small  q must be large with respect to context switch,
otherwise overhead is too high
EXAMPLE OF RR WITH TIME QUANTUM = 4

Process Burst Time

P1 24
P2 3
P3 3
 The Gantt chart is:

P1 P2 P3 P1 P1 P1 P1 P1
0 4 7 10 14 18 22 26 30

 Typically, higher average turnaround than SJF, but better

response
 q should be large compared to context switch time
 q usually 10ms to 100ms, context switch < 10 usec
 PRIORITY SCHEDULING
 A priority number (integer) is associated with each process

 The CPU is allocated to the process with the highest priority

(smallest integer  highest priority)
 Preemptive
 Nonpreemptive

 SJF is priority scheduling where priority is the inverse of

predicted next CPU burst time

 Problem  Starvation – low priority processes may never

execute

 Solution  Aging – as time progresses increase the priority of

the process
 EXAMPLE OF PRIORITY SCHEDULING
ProcessAarri Burst TimeTPriority
P1 10 3
P2 1 1
P3 2 4
P4 1 5
P5 5 2

 Priority scheduling Gantt Chart

 Average waiting time = 8.2 msec

MULTILEVEL QUEUE
 Ready queue is partitioned into separate queues, eg:
 foreground (interactive)
 background (batch)
 Process permanently in a given queue
 Each queue has its own scheduling algorithm:
 foreground – RR
 background – FCFS
 Scheduling must be done between the queues:
 Fixed priority scheduling; (i.e., serve all from foreground then
from background). Possibility of starvation.
 Time slice – each queue gets a certain amount of CPU time
which it can schedule amongst its processes; i.e., 80% to
foreground in RR
 20% to background in FCFS
MULTILEVEL QUEUE SCHEDULING
MULTILEVEL FEEDBACK QUEUE
 A process can move between the various queues; aging
can be implemented this way
 Multilevel-feedback-queue scheduler defined by the
following parameters:
 number of queues
 scheduling algorithms for each queue
 method used to determine when to upgrade a process
 method used to determine when to demote a process
 method used to determine which queue a process will enter
when that process needs service
EXAMPLE OF MULTILEVEL FEEDBACK QUEUE
 Three queues:
 Q0 – RR with time quantum 8 milliseconds
 Q1 – RR time quantum 16 milliseconds
 Q2 – FCFS

 Scheduling
 A new job enters queue Q0 which is served FCFS
 When it gains CPU, job receives 8 milliseconds

 If it does not finish in 8 milliseconds, job is

moved to queue Q1
 At Q1 job is again served FCFS and receives 16
additional milliseconds
 If it still does not complete, it is preempted and

moved to queue Q2
MULTIPLE-PROCESSOR SCHEDULING
 CPU scheduling more complex when multiple CPUs are
available
 Homogeneous processors within a multiprocessor
 Asymmetric multiprocessing – only one processor accesses
the system data structures, alleviating the need for data
sharing
 Symmetric multiprocessing (SMP) – each processor is
self-scheduling, all processes in common ready queue, or each
has its own private queue of ready processes
 Currently, most common
 Processor affinity – process has affinity for processor on
which it is currently running
 soft affinity
 hard affinity
 Variations including processor sets
MULTIPLE-PROCESSOR SCHEDULING – LOAD BALANCING

 If SMP, need to keep all CPUs loaded for efficiency

 Load balancing attempts to keep workload evenly
distributed
 Push migration – periodic task checks load on
each processor, and if found pushes task from
overloaded CPU to other CPUs
 Pull migration – idle processors pulls waiting
task from busy processor