5. Process Scheduling
1
CPU-I/O Burst Cycle
Process execution consists of a cycle of
CPU execution and I/O wait.
 Process alternates between those two
states.

2
Alternating sequence of CPU and
I/O bursts
3
Usually large number of short CPU bursts
and a small number of long CPU bursts.
 I/O-bound process: typically many short
CPU bursts.
 CPU-bound process: a few long CPU
bursts.

4
Histogram of CPU-burst durations
of a sample process
5
CPU (short-term) scheduler
Choose a process from the ready queue
to execute on a CPU.
 The ready queue is not necessarily first-in
first-out (FIFO).
 Records in the queues are generally the
PCBs of the processes.
 Can be either nonpreemptive or
preemptive.

6
Nonpreemptive scheduling

Invoke CPU scheduler only when it is
absolutely needed.
1. When a running process terminates,
2. When a running process changes to waiting
state (waiting for I/O or an event).



Old operating system like windows 3.11
use nonpreemptive scheduling.
More like cooperative scheduling
Processes must be willing to give up
CPU by itself.
7
More on nonpreemptive scheduling
Once the CPU has been allocated to a
process, the process keeps the CPU until
it releases the CPU.
 Used by Windows 3.x, previous versions
of Mac OS.
 For certain hardware platforms that lacks
some special hardware (e.g. a timer), only
nonpreemptive scheduling can be used.

8
Some issues from nonpreemtive
If a process crashes
 If a process runs indefinitely (infinite loop)
 Responsiveness

9
Preemptive scheduling

In addition to the two cases above, also
invoke CPU scheduler when it is not
absolutely needed, e.g.
◦ When a new process enters the ready queue.
◦ When the currently running process runs for
certain amount of time.
◦ Process maybe forcefully/unwillingly swap out
of CPU right.
10
More on preemptive scheduling
Used by Windows 95 and newer, Mac OS
X.
 OS needs to make sure that a process is
not preempted while it is updating shared
data.

◦ => Overhead in handling this.
11

Preemption also affects the design of the
kernel.
◦ Example: a process is preempted while the
system is modifying the I/O queue for the
process.

So needs mechanism to deal with such
problem.
12
Dispatcher
The dispatcher is the module that gives
control of the CPU to the process
selected by the short-term scheduler.
 The dispatcher must be able to do the
following:

◦ Switching context
◦ Switching to user mode
◦ Jumping to the proper location in the user
program to resume that program.
13
The dispatcher should be as fast as
possible because it is invoked often (i.e.
every process switch).
 The time it takes for the dispatcher to
stop one process and start another
running is known as the dispatch
latency.

14
Scheduling criteria
There are different criteria in choosing
the “best” process by the CPU scheduler.
 There are also many different CPU
scheduling algorithms.

15
Some of the criteria for good scheduling
algorithms are:
 CPU utilization.
◦ We want to keep the CPU as busy as possible.
◦ Can range from 0 to 100 percent
(conceptually).
16

Throughput.
◦ Throughput is the number of processes that
are completed per time unit.
17

Turnaround time.
◦ How long does it take to execute a process?
 From the process’s point-of-view, it should be short
18
The turnaround time is the interval from
the time of submission of a process to the
time of completion.
 This includes the time

◦
◦
◦
◦
Spent waiting to get into memory,
Waiting in the ready queue,
Executing on a CPU, and
Doing I/O.
19

Waiting time.
◦ Waiting time is the sum of the periods (time)
spent waiting in the ready queue.
20

Response time.
◦ In an interactive system, turnaround time may
not be the best criterion.
◦ E.g. users may prefer to get one output after 5
minutes then the rest after another 5 minutes
than to get all output after 9 minutes.
21
Response time is the time from the
submission of a request until the first
response is produced.
 This is the time it takes to start responding,
not the time it takes to output all of the
response.
 The turnaround time is generally limited by
the speed of the output device.

22

Usually, it is desirable to maximize
◦ CPU utilization and
◦ Throughput

And to minimize
◦ Turnaround time,
◦ Waiting time, and
◦ Response time.
23

In most cases, we optimize the average
measure (over all processes).
◦ E.g. Maximize the average CPU utilization
over all processes in the system. Minimize the
average turnaround time.
24

Under some circumstances, however,
better to optimize the minimum or
maximum values rather than the average.
◦ For example, to guarantee that all users get
good service. -> minimize the maximum
response time.
25
For interactive system, perhaps more
important to minimize the variance in the
response time than to minimize the
average response time.
 Reason: a system with reasonable and
predictable response time may be
considered more desirable than one that
is faster but highly variable.
 However, minimizing variance is difficult.

26
CPU/IO Queues
Scheduling algorithms can be used in
different parts of OS/other software
 Printing queue
 Balancing (while maximize the utilization)

◦ CPU
◦ I/O
27