Chapter 2: Processes & Threads

Chapter 2
 Processes and threads
 Processes
 Threads
 Scheduling
 Interprocess communication
 Classical IPC problems

CS 1550, cs.pitt.edu (origin 2

aly modified by Ethan L. Mil Chapter 2
 What is a process?
 Code, data, and stack
 Usually (but not always) has its own address space
 Program state
 CPU registers
 Program counter (current location in the code)
 Stack pointer
 Only one process can be running in the CPU at any
given time!

CS 1550, cs.pitt.edu (origin 3

aly modified by Ethan L. Mil Chapter 2
 The process model
Single PC Multiple PCs  Multiprogramming of four
(CPU’s point of view) (process point of view) programs
 Conceptual model
A  4 independent processes
B A D  Processes run sequentially
C
B  Only one program active at any
C instant!
B  That instant can be very short…
D

D
C
B
A
Time
CS 1550, cs.pitt.edu (origin 4
aly modified by Ethan L. Mil Chapter 2
 When is a process created?
 Processes can be created in two ways
 System initialization: one or more processes created when
the OS starts up
 Execution of a process creation system call: something
explicitly asks for a new process
 System calls can come from
 User request to create a new process (system call executed
from user shell)
 Already running processes
 User programs
 System daemons

CS 1550, cs.pitt.edu (origin 5

aly modified by Ethan L. Mil Chapter 2
 When do processes end?
 Conditions that terminate processes can be
 Voluntary
 Involuntary
 Voluntary
 Normal exit
 Error exit
 Involuntary
 Fatal error (only sort of involuntary)
 Killed by another process

CS 1550, cs.pitt.edu (origin 6

aly modified by Ethan L. Mil Chapter 2
 Process hierarchies
 Parent creates a child process
 Child processes can create their own children
 Forms a hierarchy
 UNIX calls this a “process group”
 If a process exits, its children are “inherited” by the
exiting process’s parent
 Windows has no concept of process hierarchy
 All processes are created equal

CS 1550, cs.pitt.edu (origin 7

aly modified by Ethan L. Mil Chapter 2
 Process states
 Process in one of 5 states
Created  Created
 Ready
1  Running
 Blocked
Ready  Exit
2  Transitions between states
5 1 - Process enters ready queue
3 2 - Scheduler picks this process
Blocked 3 - Scheduler picks a different
Running
(waiting) process
4 4 - Process waits for event (such as
I/O)
5 - Event occurs
7 6 - Process exits
6 7 - Process ended by another
7
process
Exit

CS 1550, cs.pitt.edu (origin 8

aly modified by Ethan L. Mil Chapter 2
 Processes in the OS
 Two “layers” for processes
 Lowest layer of process-structured OS handles interrupts,
scheduling
 Above that layer are sequential processes
 Processes tracked in the process table
 Each process has a process table entry
Processes

0 1 … N-2 N-1

Scheduler
CS 1550, cs.pitt.edu (origin 9
aly modified by Ethan L. Mil Chapter 2
 What’s in a process table entry?
Process management File management
May be Registers Root directory
stored Program counter Working (current) directory
on stack CPU status word File descriptors
Stack pointer User ID
Process state Group ID
Priority / scheduling parameters
Process ID
Parent process ID
Memory management
Signals
Pointers to text, data, stack
Process start time
or
Total CPU usage
Pointer to page table

CS 1550, cs.pitt.edu (origin 10

aly modified by Ethan L. Mil Chapter 2
 What happens on a trap/interrupt?
1. Hardware saves program counter (on stack or in a
special register)
2. Hardware loads new PC, identifies interrupt
3. Assembly language routine saves registers
4. Assembly language routine sets up stack
5. Assembly language calls C to run service routine
6. Service routine calls scheduler
7. Scheduler selects a process to run next (might be
the one interrupted…)
8. Assembly language routine loads PC & registers
for the selected process

CS 1550, cs.pitt.edu (origin 11

aly modified by Ethan L. Mil Chapter 2
 Threads: “processes” sharing memory
 Process == address space
 Thread == program counter / stream of instructions
 Two examples
 Three processes, each with one thread
 One process with three threads

Process 1 Process 2 Process 3 Process 1

User
space

Threads Threads
System Kernel Kernel
space
CS 1550, cs.pitt.edu (origin 12
aly modified by Ethan L. Mil Chapter 2
 Process & thread information

Per process items

Address space
Open files
Child processes
Signals & handlers
Accounting info
Global variables

Per thread items Per thread items Per thread items

Program counter Program counter Program counter
Registers Registers Registers
Stack & stack pointer Stack & stack pointer Stack & stack pointer
State State State

CS 1550, cs.pitt.edu (origin 13

aly modified by Ethan L. Mil Chapter 2
 Threads & Stacks
Thread 1 Thread 2 Thread 3

User space

Process
Thread 1’s
stack

Thread 3’s
stack

Thread 2’s
stack
Kernel

=> Each thread has its own stack!

CS 1550, cs.pitt.edu (origin 14
aly modified by Ethan L. Mil Chapter 2
 Why use threads?
 Allow a single application When in the Course of human events, it
becomes necessary for one people to
dissolve the political bands which have
We hold these truths to be self-evident,
that all men are created equal, that they
are endowed by their Creator with
destructive of these ends, it is the Right
of the People to alter or to abolish it,
and to institute new Government,

to do many things at once connected them with another, and to

assume among the powers of the earth,
the separate and equal station to which
the Laws of Nature and of Nature's God
certain unalienable Rights, that among
these are Life, Liberty and the pursuit
of Happiness.--That to secure these
rights, Governments are instituted
laying its foundation on such principles
and organizing its powers in such form,
as to them shall seem most likely to
effect their Safety and Happiness.

Simpler programming model

entitle them, a decent respect to the among Men, deriving their just powers Prudence, indeed, will dictate that
 opinions of mankind requires that they
should declare the causes which impel
from the consent of the governed, --
That whenever any Form of
Governments long established should
not be changed for light and transient
them to the separation. Government becomes causes; and accordingly all

 Less waiting
 Threads are faster to create
or destroy
 No separate address space
 Overlap computation and
I/O
 Could be done without
threads, but it’s harder
Kernel
 Example: word processor
 Thread to read from keyboard
 Thread to format document
 Thread to write to disk
CS 1550, cs.pitt.edu (origin 15
aly modified by Ethan L. Mil Chapter 2
 Multithreaded Web server
Dispatcher while(TRUE) {
thread getNextRequest(&buf);
Worker handoffWork(&buf);
thread }

while(TRUE) {
waitForWork(&buf);
lookForPageInCache(&buf,&page);
if(pageNotInCache(&page)) {
Kernel
readPageFromDisk(&buf,&page);
Web page }
cache returnPage(&page);
Network
connection }

CS 1550, cs.pitt.edu (origin 16

aly modified by Ethan L. Mil Chapter 2
 Three ways to build a server
 Thread model
 Parallelism
 Blocking system calls
 Single-threaded process: slow, but easier to do
 No parallelism
 Blocking system calls
 Finite-state machine
 Each activity has its own state
 States change when system calls complete or interrupts
occur
 Parallelism
 Nonblocking system calls
 Interrupts
CS 1550, cs.pitt.edu (origin 17
aly modified by Ethan L. Mil Chapter 2
 Implementing threads
Process

Thread

Kernel Kernel

Run-time Thread Process Process Thread

system table table table table

User-level threads Kernel-level threads

+ No need for kernel support + More flexible scheduling
- May be slower than kernel threads + Non-blocking I/O
- Harder to do non-blocking I/O - Not portable

CS 1550, cs.pitt.edu (origin 18

aly modified by Ethan L. Mil Chapter 2
 Scheduling
 What is scheduling?
 Goals
 Mechanisms
 Scheduling on batch systems
 Scheduling on interactive systems
 Other kinds of scheduling
 Real-time scheduling

CS 1550, cs.pitt.edu (origin 19

aly modified by Ethan L. Mil Chapter 2
 Why schedule processes?
 Bursts of CPU usage alternate with periods of I/O wait
 Some processes are CPU-bound: they don’t make many I/O
requests
 Other processes are I/O-bound and make many kernel
requests

Total CPU usage

CPU bound

CPU bursts I/O waits

I/O bound
Total CPU usage
Time
CS 1550, cs.pitt.edu (origin 20
aly modified by Ethan L. Mil Chapter 2
 When are processes scheduled?
 At the time they enter the system
 Common in batch systems
 Two types of batch scheduling
 Submission of a new job causes the scheduler to run
 Scheduling only done when a job voluntarily gives up the CPU
(i.e., while waiting for an I/O request)
 At relatively fixed intervals (clock interrupts)
 Necessary for interactive systems
 May also be used for batch systems
 Scheduling algorithms at each interrupt, and picks the next
process from the pool of “ready” processes

CS 1550, cs.pitt.edu (origin 21

aly modified by Ethan L. Mil Chapter 2
 Scheduling goals
 All systems
 Fairness: give each process a fair share of the CPU
 Enforcement: ensure that the stated policy is carried out
 Balance: keep all parts of the system busy
 Batch systems
 Throughput: maximize jobs per unit time (hour)
 Turnaround time: minimize time users wait for jobs
 CPU utilization: keep the CPU as busy as possible
 Interactive systems
 Response time: respond quickly to users’ requests
 Proportionality: meet users’ expectations
 Real-time systems
 Meet deadlines: missing deadlines is a system failure!
 Predictability: same type of behavior for each time slice

CS 1550, cs.pitt.edu (origin 22

aly modified by Ethan L. Mil Chapter 2
 Measuring scheduling performance
 Throughput
 Amount of work completed per second (minute, hour)
 Higher throughput usually means better utilized system
 Response time
 Response time is time from when a command is submitted until
results are returned
 Can measure average, variance, minimum, maximum, …
 May be more useful to measure time spent waiting
 Turnaround time
 Like response time, but for batch jobs (response is the completion of
the process)
 Usually not possible to optimize for all metrics with the same
scheduling algorithm

CS 1550, cs.pitt.edu (origin 23

aly modified by Ethan L. Mil Chapter 2
 First Come, First Served (FCFS)
 Goal: do jobs in the order
Current job queue
they arrive
4 3 6 3  Fair in the same way a bank
teller line is fair
A B C D  Simple algorithm!
 Problem: long jobs delay
FCFS scheduler
every job after them
 Many processes may wait for
Execution order a single long job
4 3 6 3

A B C D

CS 1550, cs.pitt.edu (origin 24

aly modified by Ethan L. Mil Chapter 2
 Shortest Job First (SJF)
 Goal: do the shortest job
Current job queue first
4 3 6 3
 Short jobs complete first
 Long jobs delay every job
A B C D after them
 Jobs sorted in increasing
SJF scheduler
order of execution time
 Ordering of ties doesn’t
matter
Execution order  Shortest Remaining Time
3 3 4 6 First (SRTF): preemptive
form of SJF
B D A C  Problem: how does the
scheduler know how long a
job will take?
CS 1550, cs.pitt.edu (origin 25
aly modified by Ethan L. Mil Chapter 2
 Three-level scheduling
CPU
CPU scheduler
Arriving
jobs
Main
memory
Input Admission Memory
queue scheduler scheduler

 Jobs held in input queue until moved into memory

 Pick “complementary jobs”: small & large, CPU- & I/O-intensive
 Jobs move into memory when admitted
 CPU scheduler picks next job to run
 Memory scheduler picks some jobs from main memory and
moves them to disk if insufficient memory space
CS 1550, cs.pitt.edu (origin 26
aly modified by Ethan L. Mil Chapter 2
 Round Robin (RR) scheduling
 Round Robin scheduling
 Give each process a fixed A B C D E
time slot (quantum)
 Rotate through “ready”
processes
 Each process makes some E
progress D
 What’s a good quantum? C
 Too short: many process B
switches hurt efficiency A
 Too long: poor response to
interactive requests Time
 Typical length: 10–50 ms

CS 1550, cs.pitt.edu (origin 27

aly modified by Ethan L. Mil Chapter 2
 Priority scheduling
 Assign a priority to each process High “Ready” processes
 “Ready” process with highest
priority allowed to run Priority 4
 Running process may be
interrupted after its quantum Priority 3
expires
Priority 2
 Priorities may be assigned
dynamically Priority 1
 Reduced when a process uses
CPU time Low
 Increased when a process waits
for I/O
 Often, processes grouped into
multiple queues based on
priority, and run round-robin per
queue
CS 1550, cs.pitt.edu (origin 28
aly modified by Ethan L. Mil Chapter 2
 Shortest process next
 Run the process that will finish the soonest
 In interactive systems, job completion time is unknown!
 Guess at completion time based on previous runs
 Update estimate each time the job is run
 Estimate is a combination of previous estimate and most
recent run time
 Not often used because round robin with priority
works so well!

CS 1550, cs.pitt.edu (origin 29

aly modified by Ethan L. Mil Chapter 2
 Lottery scheduling
 Give processes “tickets” for CPU time
 More tickets => higher share of CPU
 Each quantum, pick a ticket at random
 If there are n tickets, pick a number from 1 to n
 Process holding the ticket gets to run for a quantum
 Over the long run, each process gets the CPU m/n of
the time if the process has m of the n existing tickets
 Tickets can be transferred
 Cooperating processes can exchange tickets
 Clients can transfer tickets to server so it can have a
higher priority

CS 1550, cs.pitt.edu (origin 30

aly modified by Ethan L. Mil Chapter 2
 Policy versus mechanism
 Separate what may be done from how it is done
 Mechanism allows
 Priorities to be assigned to processes
 CPU to select processes with high priorities
 Policy set by what priorities are assigned to processes
 Scheduling algorithm parameterized
 Mechanism in the kernel
 Priorities assigned in the kernel or by users
 Parameters may be set by user processes
 Don’t allow a user process to take over the system!
 Allow a user process to voluntarily lower its own priority
 Allow a user process to assign priority to its threads
CS 1550, cs.pitt.edu (origin 31
aly modified by Ethan L. Mil Chapter 2
 Scheduling user-level threads
Process A Process B
 Kernel picks a process to
run next
 Run-time system (at user
level) schedules threads
 Run each thread for less than
process quantum
 Example: processes get 40ms
each, threads get 10ms each
Kernel
 Example schedule:
Run-time Thread Process A1,A2,A3,A1,B1,B3,B2,B3
system table table  Not possible:
A1,A2,B1,B2,A3,B3,A2,B1

CS 1550, cs.pitt.edu (origin 32

aly modified by Ethan L. Mil Chapter 2
 Scheduling kernel-level threads
Process A Process B
 Kernel schedules each
thread
 No restrictions on ordering
 May be more difficult for
each process to specify
priorities
 Example schedule:
A1,A2,A3,A1,B1,B3,B2,B3
Kernel  Also possible:
Process A1,A2,B1,B2,A3,B3,A2,B1
Thread
table table

CS 1550, cs.pitt.edu (origin 33

aly modified by Ethan L. Mil Chapter 2