Memory Management

Rab Nawaz Khan Jadoon

Department of Computer Science Assistant Professor
COMSATS Lahore
DCS
Pakistan
COMSATS Institute of
Information Technology

Operating System Concepts

 Objectives

 Primary memory (real memory, physical

memory) of computer system has been a major
influence on OS design.
 Main memory is still expansive compared to
secondary storage.

Department of Computer Science

 Memory Management
Operating System Release date Minimum Recomm.
memory req Memory
Windows 1.0 Nov 1985 256 KB

Windows 2.03 Nov 1987 320 KB

Windows 3.0 Mar 1990 896 KB I MB

Windows 3.1 Apr 1992 2.6 MB 4 MB

Windows 95 Aug 1995 8 MB 16 MB

Windows NT 4.0 Aug 1996 32 MB 96 MB

Windows 98 Jun 1998 24 MB 64 MB

Windows ME Sep 2000 32 MB 128 MB

Windows 2000 Feb 2000 64 MB 128 MB

Professional
Windows XP Oct 2001 64 MB 128 MB
Home
Windows XP prof Oct 2001 128 MB 256 MB

Windows Vista Jan 2007 512 MB 1 GB

Windows 7 Oct 2009 1 GB 2 GB

Department of Computer Science 3

 Memory Hierarchy

Department of Computer Science 4

 Memory Manager

 It is the component of the OS concerned with

the system’s memory organization scheme and
memory management strategies.
 It determines, how available memory space is
allocated to processes and how to respond to
changes in a memory process’s memory usage.
 It also interact with special purpose memory
management Hardware to improve performance.

Department of Computer Science 5

 Memory management strategies
Memory
Management
Strategies

Fetch Placement Replacement

strategies strategies strategies

Demand
First fit
fetch

Anticipatory
Best fit
fetch

Worst fit

Department of Computer Science 6

 Memory Management Strategies
cont…

 Fetch Strategy
 It determines when to move the next piece of a
program or data to main memory from the secondary
memory.
 Demand Fetch
 In the conventional methods, the system places the next
piece of the code or data in main memory when a running
program references it.
 That is, programs are loaded into the RAM area when they
are needed.

Department of Computer Science 7

 Memory Management Strategies
cont…

 Anticipatory strategy
 The program or data which is loaded on the basis of
prediction that it will be reference in the near future.
 This will increase the performance but if the loaded
pages are not used, then there is an inefficient
memory utilization.

Department of Computer Science 8

 Memory Management Strategies
cont…
 Placement strategy
 It determines where in main memory the system
should place incoming program or data pieces.
 First Fit Strategy
 Place the incoming page into the first available hole in the
memory
 It is faster but leads to memory waste.
 It allows the system to make a placement decision quickly.
 Best Fit Strategy
 Place the incoming page into the hole in which it fits tight.
 Best use of memory space.
 It is slower in making allocation.

Department of Computer Science 9

 Memory Management Strategies
cont…

 Worst Fit Strategy

 Worst-fit memory allocation is opposite to best-fit.
 It allocates free available block to the new job and it
is not the best choice for an actual system.
 Place the incoming job in the largest available whole
 Problem  Memory fragmentation  Memory
waste.

Department of Computer Science 10

 Memory Management Strategies
cont…

 Replacement strategy
 When memory is too full to accommodate a new
program, the system must remove some of program
or data currently resides in memory.
 The system’s replacement strategy determines which
piece to remove for making room for incoming jobs.

Department of Computer Science 11

 Contiguous vs non contiguous
Memory Allocations

 To execute a program in the early systems, the

OS had to find enough contiguous main
memory to accommodate the entire program.
 If the program size is larger than the available
memory, the program could not execute it.
 This is done in contiguous memory allocation, in
which all the program pages are loaded into the
adjacent memory allocations.

Department of Computer Science 12

 Contiguous vs non contiguous
Memory Allocations

 Non Contiguous Memory Allocation

 In noncontiguous memory allocation, a program is
divided into blocks or segments, that the system may
place in non adjacent slots in main memory
 This allows making use of holes (unused gaps) in
memory that would be too small to hold whole
programs.
 The OS thereby incurs more overhead as for as using
these unused holes for incoming pages.

Department of Computer Science 13

 Quiz
 When is noncontiguous preferable to contiguous
memory allocation???

ANSWER!
When the available memory contain no area large enough to hold the incoming programs in
one Contiguous space, but sufficient smaller pieces of memory are available, that in total are
large enough.

Department of Computer Science 14

 Overlays
 Contiguous memory allocation limited the size
of programs that could execute on the system
 One way in which software designer could overcome
the memory limitation was to create overlays, which
allowed the system to execute programs larger than
main memory.
 Programmer divides the program into logical
sections.
 When the program does not need the memory for
one section, the system can replace some or all of its
with the memory for a needed section.
 Overlays enables a programmer to extend main
memory.
Department of Computer Science 15
 Overlay structure
0
Operating System User portion with memory requirements
larger than available portion of main memory
a
Portion of user code and Initialization Processing Output
data than must remain phase phase phase
in main memory for b
b b
duration of execution
b

Overlay Area A

C
c

A Load initialization phase at b and run

B Load Processing phase at b and run

C Load Output phase at b and run

Department of Computer Science 16

 Overlay

 Managed overlay requires careful and time

consuming planning and the programmer often
must have detailed knowledge of the system
memory organization.
 Later on virtual memory system obviate the need for
programmer controlled overlays.

Department of Computer Science 17

 Protection in single user system
 Protection in single user contiguous memory
allocation systems can be implemented with a
single boundary register built into the
processor.
 The boundary register contains the memory address
at which the user’s program begins.
 Each time a process reference a memory address the
system determines if the request is for an address
greater than or equal to that stored in the boundary
register.
 If yes, system service the memory request.
 Else program is trying to access OS, request is terminated
with an error message.
Department of Computer Science 18
 Protection in single user system
0
Operating System Processor
a

Boundary
User Area
register
a

b Note:
Unused Area Operating System prevents the user
c from accessing the addresses less
than a.

Memory protection with single user contiguous memory allocation

Department of Computer Science 19

 Fixed Partition Multiprogramming
 Even with batch processing OSs, a single user
systems still wasted a considerable amount of
the computing resource.
 A processor wait until IO finished. Because IO
speeds were extremely slow compared with
processor speed, the CPU was severely underutilized.
 Designer saw that they could further increase CPU
utilization by implementing multiprogramming
systems, in which several users simultaneously
compete for the system resources.
 The process currently waiting for IO yields the
processor if another process is ready to do
calculation.
Department of Computer Science 20
 Fixed Partition Multiprogramming

 Thus IO and processing operation done

simultaneously.
 Greatly increases CPU utilization and system
throughput.
 To take advantage of Multiprogramming several
processes must reside in the computer’s main
memory at the same time.
 When one process requests IO, the CPU may switch
to another process and continue without any delay
associated with loading programs from secondary
storage.

Department of Computer Science 21

 Fixed Partition Multiprogramming

 The earlier multiprogramming systems used

fixed partition multiprogramming.
 In which, system divides main memory into a
number of fixed size partitions.
 Each partition holds a single job, and the system switches
the processor rapidly between jobs to create the illusion of
simultaneously.

Department of Computer Science 22

 Fixed Partition Multiprogramming

 In the early multiprogramming systems, the

programmer translated a job using an absolute
assembler or compiler (comp. , linking, and
loading),
 A job had its precise location in memory determined
before it was launched and could run only in a
specific partition.
 This restriction led to wasted memory because the
job had to wait for its designated partition (if
occupied), even if other partitions were available.

Department of Computer Science 23

 Fixed Partition Multiprogramming
0
Operating system
Job Queue for partition 1
a
These Jobs run
only in partition 1 Partition 1
Job Queue for partition 2 b
These Jobs run Partition 2
only in partition 2

Job Queue for partition 3 c

These Jobs run
only in partition 3 Partition 3

Fixed partition multiprogramming with absolute translation and loading

Department of Computer Science 24

 Fixed Partition Multiprogramming
0
Operating system
Job Queue for partition 1
a
No jobs waiting
for partition 1 Partition 1
Job Queue for partition 2 b
No jobs waiting Partition 2
for partition 2

Job Queue for partition 3 c

Partition 3
(in use)

Memory waste under Fixed partition multiprogramming with

absolute translation and loading

Department of Computer Science 25

 Fixed Partition Multiprogramming
0
Operating system
a
Partition 1
Job Queue
b

jobs ….. Partition 2

Partition 3

Fixed partition multiprogramming with relocateable translation and loading

Department of Computer Science 26

 Memory protection in
Multiprogramming

 The system can delimit each partition with two

boundary registers,
 Low and
 High
 Also called the base and limit register
 If a process issues a request for memory, the system
checks whether the requested is greater than or equal to
the process low boundary register value and less then the
high register value.
 If so, the system honors the request. Else terminate the
request with error.

Department of Computer Science 27

 Protection in Multiprogramming
0
Operating system
Processor
a
2 Currently active
Partition 1 partition
b Low Boundary
b
Partition 2 c High Boundary

Partition 3

Memory protection in contiguous allocation multiprogramming systems

Department of Computer Science 28

 Fragmentation

 One problem prevalent in all memory

organizations is that of fragmentation (loss of
memory).
 This states that certain area of the available RAM
that system cannot use.
 Fixed partition multiprogramming suffers from
internal fragmentation, which occurs when the size
of the process’s memory and data is smaller than
that of the partition in which the process execute.

Department of Computer Science 29

 Fragmentation
0
Operating system
a
Partition 1 Used Memory
b
Partition 2 Unused Memory

Partition 3

Internal Fragmentation in a Fixed Partition Multiprogramming System

Department of Computer Science 30

 Quiz 1
 Explain the need for relocating compilers,
assemblers and loaders???

Answer
Before such tools, programmer manually specified the partition into which
Their program had to be loaded, which potentially wasted memory and
Processor utilization and reduce application portability.

Department of Computer Science 31

 Quiz 2
 Explain the benefits and drawbacks of large and
small partition sizes???

Answer
Larger partition allows large program to run, but result in internal
fragmentation For small programs. Small partition reduce the amount of
internal fragmentation And increase the level of multiprogramming by allowing
more programs to reside In memory at once, but limit program size.

Department of Computer Science 32

 Variable Partition Multiprogramming

 This scheme allows a process to occupy only as

much space as needed (up to the amount of
available main memory).
 This scheme is called variable partition
multiprogramming.
 The figure on next slide shows the contiguous
allocation schemes where a process must occupy
adjacent memory allocations.
 The queue at the top contains available jobs and
information about their memory requirements.

Department of Computer Science 33

 Variable partitioning

Department of Computer Science 34

 Holes in variable partitioning

 Variable partitioning does not suffer from

internal fragmentation, because the process’s
partition is exactly the size of the process.
 Holes can be created upon process execution /
completion.
 These holes cannot accommodate the incoming
processes which are larger than these available
holes.
 These holes are called external fragmentation.

Department of Computer Science 35

 Memory holes in variable partition
multiprogramming

Department of Computer Science 36

 Defragmentation Techniques

 There are two techniques to overcome the

problems of external fragmentation created in
variable partitioning system.
 The techniques are,
 Coalescing (Merging adjacent locations)
 Compaction

Department of Computer Science 37

 Coalescing Memory Holes

Department of Computer Science 38

 Memory compaction Technique

Department of Computer Science 39

Department of Computer Science 40