Transcript Module 6: CPU Scheduling

Silberschatz, Galvin and Gagne ©2006
Operating System Principles
Chapter 5: Process Scheduling
Mi-Jung Choi
[email protected]
Dept. of Computer and Science
Ch05 - Process Scheduling
Chapter 5: Process Scheduling
 Basic Concepts
 Scheduling Criteria
 Scheduling Algorithms
 Multiple-Processor Scheduling
 Real-Time Scheduling
 Thread Scheduling
Operating System
-2-
Fall 2014
Ch05 - Process Scheduling
Basic Concepts
 Process Scheduling is

the basis of multi-programmed operating system
 Terminology


CPU scheduling, Process scheduling, Kernel Thread scheduling
used interchangeably, we use process scheduling
Operating System
-3-
Fall 2014
Ch05 - Process Scheduling
Basic Concepts
 In a single-processor system




Only one process can run at a time
Any others must wait until the CPU is free and can be rescheduled.
When the running process goes to the waiting state,
 the OS may select another process to assign CPU to improve
CPU utilization.
Every time one process has to wait, another process can take over
use of the CPU
 Process scheduling is


a fundamental function of operating-system.
to select a process from the ready queue and assign the CPU
 Maximum CPU utilization obtained with multiprogramming

by process scheduling
Operating System
-4-
Fall 2014
Ch05 - Process Scheduling
Diagram of Process State from ch.3
 It is important to realize that only one process can be running on any
processor at any instant.
 Many processes may be ready and waiting states.
Operating System
-5-
Fall 2014
Ch05 - Process Scheduling
Process Scheduling from ch.3
 Two types of queues: one ready queue, a set of device queues
 Two types of resources: CPU, I/O
 Arrow indicates the flow of processes in the system
Operating System
-6-
Fall 2014
Ch05 - Process Scheduling
CPU - I/O burst Cycle
 Process execution consists of

a cycle of CPU execution (CPU burst) and I/O wait (I/O burst)
 Process alternate between these two states


Process execution begins with a CPU burst, which is followed by
an I/O burst, and so on.
Eventually, the final CPU burst ends with an system call to
terminate execution.
 CPU burst distribution of a process

varies greatly from process to process and from computer to
computer
Operating System
-7-
Fall 2014
Ch05 - Process Scheduling
Alternating Sequence of CPU & I/O Bursts
CPU burst time
I/O burst time
CPU burst time
CPU burst time
Operating System
-8-
Fall 2014
Ch05 - Process Scheduling
Histogram of CPU-burst Times
 CPU burst distribution is generally characterized


as exponential or hyper-exponential
with large number of short CPU burst and small number of long CPU burst
 I/O bound process has many short CPU bursts
 CPU bound process might have a few long CPU bursts.
Operating System
-9-
Fall 2014
Ch05 - Process Scheduling
Process Scheduler
 is one of OS modules.
 selects one of the processes in memory that are ready to
execute, and allocates the CPU to the selected process.
 CPU scheduling decisions may take place when a
process:
1. switches from running to waiting state: I/O request, invocation of
wait() for the termination of other process
2. switches from running to ready state: when interrupt occurs
3. switches from waiting to ready: at completion of I/O
4. terminates
 Scheduling under 1 and 4 is non-preemptive
 Scheduling under 2 and 3 is preemptive
Operating System
- 10 -
Fall 2014
Ch05 - Process Scheduling
Non-preemptive vs. Preemptive
 Non-preemptive scheduling



Once the CPU has been allocated to a process, the process keeps the
CPU until it releases the CPU
either by terminating or by switching to the waiting state.
used by Windows 3.x
 Preemptive scheduling




Current running process can be switched with another at any time
 because interrupt can occur at any time
Most of modern OS provides this scheme. (Windows XP, Max OS, UNIX)
incurs a cost associated with access to shared data among processes
affects the design of the OS kernel
 Certain OS (UNIX) waits either for a system call to complete or for an
I/O block to take place before doing a context switch.
 protects critical kernel code by disabling and enabling the interrupt at
the entry and exit of the code.
Operating System
- 11 -
Fall 2014
Ch05 - Process Scheduling
Dispatcher
 Dispatcher module is a part of a Process Scheduler
 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
Operating System
- 12 -
Fall 2014
Ch05 - Process Scheduling
Context Switch from ch. 3
Operating System
- 13 -
Fall 2014
Ch05 - Process Scheduling
Scheduling Criteria
 Based on the scheduling criteria, the performance of various scheduling
algorithm could be evaluated.

Different scheduling algorithms have different properties.
 CPU utilization – ratio (%) of the amount of time while the CPU was
busy per time unit.
 Throughput – # of processes that complete their execution per time unit.
 Turnaround time – the interval from the time of submission of a process
to the time of completion. Sum of the periods spent waiting to get into
memory, waiting in the ready queue, executing on the CPU, and doing
I/O
 Waiting time – Amount of time a process has been waiting in the ready
queue, which is affected by scheduling algorithm
 Response time – In an interactive system, amount of time it takes from
when a request was submitted until the first response is produced, not
output (for time-sharing environment)
Operating System
- 14 -
Fall 2014
Ch05 - Process Scheduling
Optimization Criteria
 It is desirable to maximize:


The CPU utilization
The throughput
 It is desirable to minimize:



The turnaround time
The waiting time
The response time
 However in some circumstances, it is desirable to
optimize the minimum or maximum values rather than
the average.

Interactive systems, it is more important to minimize the variance
in the response time than minimize the average response time.
Operating System
- 15 -
Fall 2014
Ch05 - Process Scheduling
Process Scheduling Algorithms




First-Come, First-Served Scheduling (FCFS)
Shortest-Job-First Scheduling (SJF)
Priority Scheduling
Round-Robin Scheduling
 Our measure of comparison is the average waiting time.

CPU utilization, Throughput, Turn arround time,
Operating System
- 16 -
Fall 2014
Ch05 - Process Scheduling
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
0
24
P3
27
30
 Waiting time for P1 = 0; P2 = 24; P3 = 27
 Average waiting time: (0 + 24 + 27)/3 = 17
Operating System
- 17 -
Fall 2014
Ch05 - Process Scheduling
FCFS Scheduling
Suppose that the processes arrive in the order
P2 , P3 , P1
 The Gantt chart for the schedule is:
P2
0
P3
3
P1
6
30
 Waiting time for P1 = 6; P2 = 0; P3 = 3
 Average waiting time: (6 + 0 + 3)/3 = 3
 Much better than previous case
Operating System
- 18 -
Fall 2014
Ch05 - Process Scheduling
FCFS Scheduling
0
P0
P1
P2
10
20
I/O (20)
CPU (10)
CPU(6)
CPU(4)
10
CPU (4)
I/O(24)
14
P1
CPU(6)
50
CPU(9)
I/O(17)
P2
CPU(4)
40
CPU(11)
I/O(4) CPU (4)
P0 CPU(10)
Operating System
30
P2
CPU(4)
20
P0
I/O(6)
24
30
- 19 -
P1 CPU(9)
P0 CPU(11)
41
P2
CPU(4)
50
54
Fall 2014
Ch05 - Process Scheduling
FCFS Scheduling
 FCFS scheduling algorithm is non-preemptive


Once the CPU has been allocated to a process, that process keeps
the CPU until it releases the CPU, either by terminating or by
requesting I/O.
is particularly troublesome for time-sharing systems.
 Convoy effect occurs.




When one CPU-bound process with long CPU burst and many
I/O-bound process which short CPU burst.
All I/O bound process waits for the CPU-bound process to get off
the CPU while I/O is idle
All I/O- and CPU- bound processes executes I/O operation while
CPU is idle.
results in low CPU and device utilization
Operating System
- 20 -
Fall 2014
Ch05 - Process Scheduling
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
 Two schemes:


non-preemptive – once CPU given to the process it cannot be
preempted until completes its CPU burst
preemptive – if a new process arrives with CPU burst length less
than remaining time of current executing process, preempt. This
scheme is known as the Shortest-Remaining-Time-First (SRTF)
 SJF is optimal – gives minimum average waiting time for a
given set of processes
Operating System
- 21 -
Fall 2014
Ch05 - Process Scheduling
Example of Non-Preemptive SJF
Process Arrival Time
P1
0.0
P2
2.0
P3
4.0
P4
5.0
 SJF (non-preemptive)
P1
0
3
Burst Time
7
4
1
4
P3
7
P2
8
P4
12
16
 Average waiting time = (0 + 6 + 3 + 7)/4 = 4
Operating System
- 22 -
Fall 2014
Ch05 - Process Scheduling
Example of Preemptive SJF
Process Arrival Time
P1
0.0
P2
2.0
P3
4.0
P4
5.0
 SJF (preemptive)
P1
0
P2
2
P3
4
P2
5
Burst Time
7
4
1
4
P4
P1
11
7
16
 Average waiting time = (9 + 1 + 0 +2)/4 = 3
Operating System
- 23 -
Fall 2014
Ch05 - Process Scheduling
Preemptive SJF Scheduling
0
P0
P1
P2
10
20
I/O (20)
CPU (10)
CPU(6)
CPU(4)
Operating System
P1
CPU(6)
P2
CPU(4)
40
50
CPU(11)
CPU(9)
I/O(17)
I/O(4) CPU (4)
P0 P2
CPU(4)
CPU(1)
30
CPU (4)
I/O(24)
P0
CPU(9)
P1
I/O(4)
- 24 -
P1 CPU(9)
P2 P0
P2
I/O(2) CPU(4)
I/O(1)
P0
CPU(11)
Fall 2014
Ch05 - Process Scheduling
Non-preemptive SJF Scheduling
0
P0
P1
P2
10
CPU(6)
Operating System
P2
CPU(4)
50
CPU(9)
CPU (4)
I/O(24)
P1
CPU(6)
40
CPU(11)
I/O(17)
I/O(4) CPU (4)
P0 CPU(10)
30
I/O (20)
CPU (10)
CPU(4)
20
P2
CPU(4)
Idle (6)
- 25 -
P0 CPU(11)
P1 CPU(9)
P2
CPU(4)
Fall 2014
Ch05 - Process Scheduling
Determining Length of Next CPU Burst
 can only estimate the length
 can be done by using the length of previous CPU bursts,
using exponential averaging
1. tn  actual lenght of n th CPU burst
2.  n 1  predicted value for the next CPU burst
3.  is a real number, 0    1
4. Define :  n 1    tn  1    n .
 The value of tn contains our most recent information.
 n+1 stores the past history
 The parameter  controls the relative weight of recent and
past history in our prediction.
Operating System
- 26 -
Fall 2014
Ch05 - Process Scheduling
Prediction of the Length of the Next CPU Burst
 In this example, 0 = 10,  = ½
 1 =  x t0 + (1- ) x 0 = ½ x 6 + ½ x 10 = 8
 2 =  x t2 + (1- ) x 2 = ½ x 4 + ½ x 8 = 6
Operating System
- 27 -
Fall 2014
Ch05 - Process Scheduling
Examples of Exponential Averaging
=0


n+1 = n = n-1 = n-2 . … = 0
Recent history does not count
=1


n+1 =  tn
Only the actual last CPU burst counts
 If we expand the formula, we get:
n+1 =  tn + (1 - ) tn - 1 + …
+ (1 -  )j  tn -j + …
+ (1 -  )n +1 0
 Since both  and (1 - ) are less than or equal to 1, each
successive term has less weight than its predecessor
Operating System
- 28 -
Fall 2014
Ch05 - Process Scheduling
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
Non-preemptive
 SJF is a priority scheduling where priority is the predicted
next CPU burst time
 Problem  Starvation – low priority processes may never
execute
 Solution  Aging – as time progresses, increase the priority
of the process
Operating System
- 29 -
Fall 2014
Ch05 - Process Scheduling
Example of Non-Preemptive Priority
Process
P1
P2
P3
P4
P5
Burst Time
10
1
2
1
5
Priority
3
1
4
5
2
 Priority Scheduling (non-preemptive)
P2
0 1
P5
P1
P3
16
6
P4
18 19
 Average waiting time = (0 + 1 + 6 + 16 + 18)/5 = 8.2
Operating System
- 30 -
Fall 2014
Ch05 - Process Scheduling
Preemptive Priority Scheduling
0
P0(1)
P1(2)
10
20
I/O (20)
CPU (10)
CPU(6)
P2(3) CPU(4)
P1
CPU(6)
40
50
CPU(11)
CPU(9)
I/O(17)
I/O(4) CPU (4)
P0
CPU(10)
30
CPU (4)
I/O(24)
P0
P2
Idle:
P2
CPU(4) P2(I/O4) CPU(4)I/O(2)
P0
CPU(11)
P1 CPU(9)
P2
P2
I/O(2)CPU(4)
I/O is same to idle (원래 I/O에는 idle이 들어가야 합니다.)
Operating System
- 31 -
Fall 2014
Ch05 - Process Scheduling
Non-preemptive Priority Scheduling
0
10
P0 (1)
CPU (10)
P1 (2)
P2 (3)
CPU(6)
CPU(4)
I/O(4) CPU (4)
20
30
I/O (20)
40
50
CPU(11)
CPU(9)
I/O(17)
I/O(24)
CPU (4)
이 경우는 특수 case로 preemptive나 non-preemptive나 답이 같습니다. (즉 앞장과 답이 같습니다.)
Operating System
- 32 -
Fall 2014
Ch05 - Process Scheduling
Round Robin (RR) Scheduling
 Each process gets a small unit of CPU time (time quantum),
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
(= q time units) in chunks of at most n x q time units at once.
 No process waits more than (n-1) x q time units.
 Performance depends on the size of the time quantum.


q large  RR is same as FIFO
q small  provides high concurrency: each of n processes has its
own processor running at 1/n the speed of the real processor
Operating System
- 33 -
Fall 2014
Ch05 - Process Scheduling
Example of RR with Time Quantum = 20
Process
P1
P2
P3
P4
 The Gantt chart is:
P1
0
P2
20
37
P3
Burst Time
53
17
68
24
P4
57
P1
77
P3
97 117
P4
P1
P3
P3
121 134 154 162
 Typically, higher average turnaround than SJF, but better
response
Operating System
- 34 -
Fall 2014
Ch05 - Process Scheduling
RR Scheduling
0
P0(1)
P1(2)
10
CPU (10)
CPU(6)
P2(3) CPU(4)
20
30
I/O (20)
50
CPU(11)
CPU(9)
I/O(17)
I/O(4) CPU (4)
40
CPU (4)
I/O(24)
P0 P2 P0 P1 P2 P0 P1 P0 P1 P2 P0 P2
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
Idle
P1 I/O (11)
P1 P1 P1 P1 P0 P1 P0 P0 P2 P0 P2 P0
CPU(2)
CPU(2)
CPU1)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
CPU(2)
C
CPU
 Time Quantum = 2
 (9번째 P1과 10번째 P2는 바뀌어도 상관 없음)
 (20번째 P0와 21번째 P2도 바뀌어도 상관 없음)
Operating System
- 35 -
Fall 2014
Ch05 - Process Scheduling
RR Scheduling
0
P0(1)
P1(2)
10
CPU(6)
P0
CPU(5)
CPU (4)
I/O(24)
P0
CPU(5)
50
CPU(9)
I/O(17)
P1
CPU(5)
40
CPU(11)
I/O(4) CPU (4)
P2
CPU(4)
30
I/O (20)
CPU (10)
P2(3) CPU(4)
20
Idle
P1 I/O(15)
P2 P1
CPU(1)
CPU(4)
P0
CPU(5)
P1
CPU(5)
P0
P2
CPU(5) CPU(4
 Time Quantum = 5
Operating System
- 36 -
Fall 2014
Ch05 - Process Scheduling
Time Quantum and Context Switch Time
 The effect of context switching on the performance of RR
scheduling, for example one process of 10 time quantum.



quantum = 12 time units, finished in less than 1 time quantum
quantum = 6 time units, requires 2 quanta, 1 context switch
quantum = 1 time units, requires 10 quanta, 9 context switch
Operating System
- 37 -
Fall 2014
Ch05 - Process Scheduling
Round Robin (RR) Scheduling
 The time quantum q must be large with respect to context
switch, otherwise overhead is too high
 If the context switching time is 10% of the time quantum,
then about 10% of the CPU time will be spent in context
switching
 Most modern OS have time quanta ranging from 10 to 100
milliseconds,
 The time required for a context switch is typically less than
10 microseconds; thus the context-switch time is a small
fraction of the time quantum.
Operating System
- 38 -
Fall 2014
Ch05 - Process Scheduling
Turnaround Time varies with the Time Quantum
 The turnaround time
depends on the size of
the time quantum
 The average turnaround
time can be improved
if most processes finish
their next CPU burst
in a single time quantum.
 When SJF and RR used
 If quantum = 6 and 7, average turnaround time = 10.5
 If quantum = 1, average turnaround time = 11
Operating System
- 39 -
Fall 2014
Ch05 - Process Scheduling
Scheduling Algorithm with multi-Queues
 Multi-level Queue Scheduling
 Multi-level Feedback Queue Scheduling
Operating System
- 40 -
Fall 2014
Ch05 - Process Scheduling
Multilevel Queue
 Ready queue is partitioned into separate queues:
foreground (interactive)
background (batch)

The processes are permanently assigned to one queue, generally
based on some property, or process type.
 Each queue has its own scheduling algorithm


foreground – RR
background – FCFS
 Scheduling must be done between the queues


Fixed priority scheduling - serve all from foreground then from
background, Possibility of starvation.
Time slice scheduling – 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
Operating System
- 41 -
Fall 2014
Ch05 - Process Scheduling
Multilevel Queue Scheduling
 No process in the batch queue could run unless the queues with high
priority were all empty.
 If an interactive editing process entered the ready queue while a batch
process was running, the batch process would be preempted.
Operating System
- 42 -
Fall 2014
Ch05 - Process Scheduling
Multi-level Queue Scheduling
0
10
P0 (1)
CPU (10)
P1 (0)
P2 (0)
CPU(6)
CPU(4)
I/O(4) CPU (4)
20
30
I/O (20)
40
50
CPU(11)
CPU(9)
I/O(17)
I/O(24)
CPU (4)
 Two-level Queues: SJF with priority 0, FCFS with priority 1
 Fixed Priority Queue
 Preemptive
Operating System
- 43 -
Fall 2014
Ch05 - Process Scheduling
Multilevel Feedback Queue
 A process can move between the various queues; aging
can be implemented in 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
Operating System
- 44 -
Fall 2014
Ch05 - Process Scheduling
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 RR. 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 RR and receives 16 additional
milliseconds. If it still does not complete, it is preempted and
moved to queue Q2.
The job is serverd based on FCFS in queue Q2
Operating System
- 45 -
Fall 2014
Ch05 - Process Scheduling
Multilevel Feedback Queues
Operating System
- 46 -
Fall 2014
Ch05 - Process Scheduling
Multi-level Feedback Queue Scheduling
0
10
P0 (1)
CPU (10)
P1 (0)
CPU(6)
P2 (0)
CPU(4)
I/O(4) CPU (4)
20
30
I/O (20)
40
50
CPU(11)
CPU(9)
I/O(17)
I/O(24)
CPU (4)
 Three-level Queues:

RR with quantum 3, RR with quantum 6, FCFS with priority 1
Operating System
- 47 -
Fall 2014
Ch05 - Process Scheduling
Multiple-Processor Scheduling
 CPU scheduling more complex when multiple CPUs are available
 Homogeneous processors within a system or heterogeneous
processors within a system
 Asymmetric multiprocessing vs. Symmetric multiprocessing (SMP)
 Symmetric Multiprocessing (SMP) – each processor makes its own
scheduling decisions.
 Asymmetric multiprocessing – only one processor accesses the
system data structures, alleviating the need for data sharing.
 Load sharing on SMP system


keeps the workload evenly distributed across all processors in an SMP
system.
Push migration vs. Pull migration
Operating System
- 48 -
Fall 2014
Ch05 - Process Scheduling
Real-Time Scheduling
 Hard real-time systems – required to complete a critical
task within a guaranteed amount of time
 Soft real-time computing – requires that critical processes
receive priority over less fortunate ones
Operating System
- 49 -
Fall 2014
Ch05 - Process Scheduling
Thread Scheduling
 On operating systems that support kernel-level thread, it
is kernel-level threads, not processes, that are being
scheduled by the operating system.
 Local Scheduling – How the threads library decides
which thread to put onto an available LWP
 Global Scheduling – How the kernel decides which
kernel thread to run next
Operating System
- 50 -
Fall 2014
Ch05 - Process Scheduling
Summary
 CPU scheduling is the task of selecting a waiting process from the
ready queue and allocating the CPU to it.
 The CPU is allocated to the selected process by the dispatcher.
 FCFS scheduling is simple, cause short processes to wait for long time
 SJF scheduling is provably optimal, providing the shortest averaging
waiting time. But predicting the length of the next CPU bursts is difficult.
 Priority scheduling allocates the CPU to the heights priority process.
 Both priority and SJF may suffer from starvation. Aging is a technique to
prevent starvation.
 RR scheduling is more appropriate for a time-shared system.
 Major problem of RR scheduling is the selection of the time quantum.
 FCFS is non-preemptive, RR is preemptive, SJF and Priority may be
preemptive and non-preemptive.
 Multilevel queue allows different scheduling algorithm for each queue.
 Multilevel feedback queue allow process to move from one queue to
another.
Operating System
- 51 -
Fall 2014
Silberschatz, Galvin and Gagne ©2006
Operating System Principles
End of Chapter 5
Mi-Jung Choi
[email protected]
Dept. of Computer and Science