Topic 6 Operating Systems

1
Topic 6Operating Systems
2
Applications software allow the user to
communicate with the system software. The
system software communicates with the hardware.
The operating system provides a set of system
calls that can be used by the programmer of an
application to work the computer system.
The operating system can also provide a library
of subroutines, often called an Application
Programming Interface (API)
3
The operating system is part of the system
software. It can be regarded as an interface in
three different ways.
First, it is an interface between the parts of a
computer system. These work at different speeds
and in different ways, and the operating system
enables them to communicate.
Second, the operating system is an interface
between an application package and hardware.
Software houses, that create application
packages, write them to run under a particular
operating system.
Third, the operating system is an interface
between the hardware and the user.
There are two kinds of user interface
command-line interfaces (CLIs) and graphical user
interfaces (GUIs).

• MS-DOS had a command line interface.
• Apple Macs from the start offered a GUI.

4

• Operating systems are expected to have similar
  features. These are
• memory management
• processing modes
• input/output
• filing system
• resource sharing
• user interface
• applications support
• security.

5

• Additional criteria required when looking at an
  operating system is-
• cost
• ease of installation
• ease of management
• reliability
• Performance
• range of platforms
• range of software
• customisation

6
What a process is The term process is basic to
the understanding of the workings of an operating
system. A programmer writes programs. The source
code of these programs will be in assembly
language or in some high level language like
Java. When a program is actually running, it
contains run-time data and a stack other
information necessary to keep the program running
include a program counter and a stack pointer.
All of these things together make up a process.
A process can be defined as a program in
execution.
Another name for process is task.
7
When you run a program, the operating system has
to convert it into a process. It is the process,
rather than the simple program, that the
processor executes. When the process terminates,
the program itself remains (on backing store),
fit to be converted into a process again if
needed.
6.3
8
Pupil task
Now log onto scholar and work through web
animation on layers.
9
An operating system needs room in memory for
itself and for the processes it creates, and
user programs need memory for themselves and the
data they use.
Part of the operating system, known as the memory
manager, allocates memory to these processes and
keeps track of the use of memory.
Some operating systems, such as early versions of
Windows (95, 98, NT) provides only a single-user
environment. Multi-user memory management is
provided by most modern operating systems.
Most operating systems support multi-tasking. In
multi-tasking, several applications or tasks are
(apparently) available at the same time, and the
user can switch easily between one application
and another.
10
Most operating systems also support
multi-threading. Threads are sometimes called
lightweight processes.
One process can have several threads.
The threads act independently of one another, but
share the same memory space. For example, in a
database application, one thread might run the
user interface and another might handle the
actual database.
Process loading and swapping If a program is to
be run (as a process), it needs to be copied into
memory from backing store. This action - by no
means instantaneous - is known as process
loading. Some systems, as we shall see, also
take processes out of memory and store them on
disc, to make room for another process. This is
known as process swapping.
11
Monoprogramming The simplest method of memory
management is to have just one process in memory
at a time, and to allow it to use all the memory
there is. One obvious problem with this approach
is that a program needs to be entirely
self-reliant and contain,
Multiprogramming Monoprogramming would not do on
a large computer that has lots of users. Having
several programs in memory at once is called
multiprogramming.
In multiprogramming, memory is divided into
parts, of one kind or another. Names for these
parts include partitions, segments, and pages.
Multitasking Animation Log onto scholar 6.3
using web animation..
12
Fixed partitions Using fixed partitions is a
relatively straightforward way of effecting
multiprogramming. The partitions are fixed in
size but dont have to be all the same size.
1000k
Partition 4 400k
Process D
700k
Partition 3 300k
Process C
400k
Partition 2 200k
Process B
200k
Partition 1 100k
Process A
100k
Operating system
13
Drawbacks to fixed partitions The main problem
with fixed partitions is that, processes dont
usually fit partitions exactly. As a rule, there
will be wasted space in each partition. A map of
memory, as in the diagram above, looks
fragmented. This waste of space is known as
internal fragmentation (internal because the
space wasted is inside the partitions). Another
drawback is that a process might come along that
is too big for any of the partitions.
14
Variable partitions One solution, is to allocate
to each process exactly the amount of memory it
needs. When a process turns up, it is loaded
into the next available area. This can go on
until there is no room left.
1000k
Available Space
750k
Process D
550k
Process C
325k
Process B
175k
Process A
100k
Operating system
15
The arrival of process E
Process C terminates
Process F can be executed once process B
terminates
16
Pupil task
looked at how the processor is switched between
tasks resident in memory (page 156 Scholar
multitasking animation)
17
Once everything is up and running, new processes
need to be fitted into the HOLES available.
(spaces in memory can be known as holes)
There are three policies to choose from- first
fit best fit worst fit
18
Method Description comment
First fit Incoming process is fitted into the first space big enough This method is quick and simple
Best fit Incoming process is fitted into the smallest space that can take it This method tends to leave small and therefore useless spaces
Worst fit Incoming process is fitted into the largest space going This method gets around the issue of best fit the space left will often be big enough for another process.
19
Information about holes is often kept in a linked
list. Coalescing holes is straightforward.
First fit will take the first hole large
enough the other two methods will need to scan
the entire list before a decision can be made.
Compaction Fragmentation can be kept down to one
hole (of whatever size) by shifting processes
around whenever one leaves memory. All remaining
processes are packed down (compacted) so as to
leave a single hole at the top.
In the previous series of memory maps, process F
arrived and process C terminated. Process F did
not have enough room. However, with compaction,
processes D and E could be moved down in memory,
leaving enough room at the top for F.
20
Process C Terminated
Memory after Compaction
1000k
Available Space
975k
Process E
675k
Process D
525k
Available Space
325k
Process B
175k
Process A
100k
Operating system
21
All such gains in sophistication have a cost in
terms of complexity of implementation. The more
intricate the algorithm, the more intricate the
code needed to make it happen. Furthermore, an
action such as compaction is not instantaneous
but takes time. The overheads that compaction
calls for tend, in fact, to be so great and make
things so slow that the scheme is rarely used in
practice.
Paging if we assume that processes are relatively
small and can fit into memory. From the very
early days, there has been the problem that some
programs are too large to fit into memory at all,
never mind into a fraction of it such as a
partition.
process frames
A 5
B 3
C 2
D 6
In a paged system, processes are chopped up into
a number of chunks all of the same size. These
chunks are called pages.
22
Complete questions 5 12 on pages 161/162.
23
Process management is to do with choosing which
process to run, and how long to run it for. This
is managed by part of the operating system called
the scheduler. A scheduler works according to a
particular scheduling algorithm. The algorithm is
subject to certain constraints

• it should be fair, so that each process gets its
  fair share of the processor
• it should be efficient, to maximise use of the
  processor
• it should achieve high throughput
• it should have a good response time.

Many processes are, obviously, started directly
by the user. Others are started by the operating
system, however processes can themselves start
other processes.
All these processes have a program counter,
registers, and other information. The processor
performs whats called a context switch when it
changes from one process to another
24
a process, once it had gained control of the
processor, would run to completion. Such an
approach is still tenable on PCs that expect to
have only one user at a time (DOS works in this
way), but wont do at all on a multi-user or
multi-tasking system.
One answer is cooperative scheduling where each
process would, in principle, run for a little and
get a bit of work done. The problem with this
is that processes tend to act selfishly and keep
the processor to themselves for as long as they
can. This is due to the operating system giving
up too much control. Cooperative scheduling
would work even worse in a multi-user system.
Somebody
The scheduler needs, then, to be able to suspend
one process, at any point, and give another
process a turn. The operating system keeps
control. Such scheduling is known as pre-emptive
scheduling.
25

• Process states
• A process enters the system and starts running.
  When its finished, it terminates.
• While its in the system, the process can be in
  one of three states
• running
• blocked
• ready.
• The process runs until obliged to wait for I/O or
  pre-empted by the scheduler (to give
• another process a turn).
• If it is pre-empted, it is still ready to use the
  processor.
• If it is waiting for I/O, it is described as
  blocked, and stays in that condition until the
  I/O is over
• look at Diagram 6.9 on page 163

There are 2 different scheduling
algorithms. Round robin Priority scheduling
26
Round robin Processes are put in a sort of
circular list. Each process in turn is assigned
a time-slice. If the process is still running
at the end of the time slice, it is pre-empted
and has to wait its turn again
Switching processes takes time and thus wastes
processor time. Too long a time slice might mean
irritatingly long waits.
Round robin is suitable to interactive systems (
not real time). The algorithm is also fair, in
that all jobs are treated as though they are
equally important
Priority scheduling In priority scheduling, each
process is given a priority, and the job with the
highest priority is processed. In many systems,
there is a queue of jobs for each level or
priority, and the jobs within a certain level are
treated to round-robin scheduling. Because
low-priority jobs might, on a busy system, never
get a turn, things are sometimes organised so
that jobs that have been waiting a long time have
their priority raised.
27
Three algorithms used for priority scheduling are
first come first served, shortest job first and
multi-level feedback queues.
First come, first served (FCFS) This algorithm is
straightforward, and works well enough for long
jobs.
Job Processingtime Turnaroundtime A 20 20 B
10 30 C 4 34 D 2 36 This gives an average
turnaround time of 30 minutes, which compares
badly with an average run-time of 9 minutes.
Shortest job first In batch systems, and other
systems where the running time of a process can
be predicted, we can use the principle that the
shortest jobs should be run first.
This gives an average turnaround time of 15
minutes the longest job does worse but the three
other jobs do a lot better. The
shortest-job-first approach gives optimal
turnaround times. On the other hand, long jobs on
busy systems might have to wait indefinitely
because shorter jobs keep getting in the way.
28
Multi-level Feedback Queues
Each of these methods has advantages and also
disadvantages. The multi-level feedback queue
algorithm tries to combine the best of each.
The top level queue has highest priority, but
each process in this queue is only allocated one
very short time slice.
for a longer time independently of the processor
(perhaps using DMA).
If it is a longer process, it is given a short
time slice, then moved down to the second level
queue, so that the top level queue is not held up.
The second level queue gets processor time if the
top level queue is empty. If a process is till
not completed within the longer time slice, it is
pushed down to the next queue.
The final queue is a round robin queue, which
continues to allocate time slices to processes
(when the processor is not being used by the
higher queues) until they are completed.
29
Pupil task
Now complete questions 13 -18 on page 166
30
Input and output (I/O)
Peripheral devices are, in general, different
from one another they work in different ways and
at different speeds.
In the I/O system, the software can be thought of
as organised in layers
the upper one presents an interface to the user
or application.
The device driver communicates with the device
controller, which is part of the hardware,
plugged into the computers I/O bus.
I/O Control System _________ Device Driver
I/O Bus
Device Controller ______________ Device
the lower one deals directly with the hardware
31
Device independence
The I/O system hides the physical details of a
device from the users and presents then with
straightforward means of access and control
Input/Output in Windows
A DLL is a piece of executable code (a device
driver, for example) that can be shared by
different processes. The routines in a DLL are
only linked into an application at run time. In
Windows, all the device drivers are dynamic link
libraries.
Plug and play
Adding a new device to a system has in the past
been rather a fiddly business. Often a new
control card has to be plugged into the computer.
Very often, switches on the device would have to
be set by hand.
Plug and play can be regarded as a sort of
convention or standard agreed between computer
manufacturers, chip manufacturers, device
manufacturers, and so on. The user has to supply
very little information during installation and
nothing at all from then on.
32
Direct memory access
A device controller will have its own memory on
board. This enables it to use Direct Memory
Access (DMA) to get data to and from the
computers memory. In DMA, the processor only
sets up the transfer. Thereafter, some clock
cycles are used by the I/O system for the data
transfer.
Virtual devices
A virtual device seems, to the user, like an
actual device but is, in fact, a simulation
maintained by the I/O system. A print spooler is
a virtual device, that takes data sent to the
printer, stores them temporarily on disc, and
prints them when the system is ready to do so.
Buffering
Buffering is used in sending blocks of data. Data
is transferred into the buffer until it is full.
Then the entire block is sent at once. This is
more efficient than having the data trickle
through the system.
Spooling
Spooling involves the input or output of data to
a tape or a disk. The print spooler stores the
data in files and sends it to the printer when it
is ready, using a print queue.
33
Pupil task
Question 19,20 21 on page 170
34
File Management

• The most familiar aspect of the Operating System
  to most users is whenever they carry out any of
  the following operations
• loading an application
• saving a document
• creating a new folder
• renaming a file
• deleting an unwanted application

A major function of the file management section
of the operating system is to map the logical
view (as seen by the user) on to the physical
view (as seen by the computer).
Logical View of files

• A file is a meaningful collection of data. The
  file will have
• a name,
• a file type,
• a location (within a folder on a disc),
• a size.
• other attributes such as its creation date,
• date last modified,
• read-only.

35
Folders and Directories.

• Applications install themselves into folders.
• Users arrange their work into folders.
• The arrangement of folders on the root containing
  folders that contain folders and so on, is often
  described as hierarchical.
• Paths
• In figure 6.11 (page 172), There is no confusion
  because the folders are on different paths their
  full names, so to speak, are different.

Physical View Most storage devices are
divided up into tracks and sectors, forming
blocks of a fixed size, perhaps 512 bytes. See
diagram page 173
The term cluster (or blocks) is also used to
describe the smallest section of a storage device
that the operating system will recognise.
36
Contiguous Allocation
the following sequence of user commands were
required using a disk with 30 clusters, each of
size 32K
Command 1 save file 1 (size 64K)
Command 2 save file 2 (size 97K)
Command 3 save file 3 (size 3K)
Command 4 save file 4 (size 220K)
37
Command 5 delete file 2
Command 6 save file 5 (220K)
Command 7 copy file 1 (64K)
Command 8 save file 6 (70K)
38
Command 9 save file 7 (190K)
This will need 6 clusters. These must be
contiguous, so clusters 6 and 7 or clusters 26 to
30 cannot be used. Although there are more than 6
free clusters, the file cannot be copied as the
clusters are not contiguous.
Command 10 add data to file 1
Perhaps file 1 is a word processed document, and
the user loads it up, adds some extra paragraphs,
then tries to save it. The system will be unable
to save the file as the next cluster (cluster 4)
is already in use.
39
Linked Allocation page 179
Indexed Allocation page 180
File Allocation Tables page 181(FAT)
The file management system in operation page
182 understand Diagram
40
Pupil task
REVISION QUESTIONS ON PAGE 185