Lecture 33 - Introduction to Memory Management

National University of
Computer & Emerging Sciences

Lecture 33
Introduction to Memory Management

Dr. Ehtesham Zahoor

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Background
• Program must be brought (from disk) into memory and
placed within a process for it to be run

• A programmer will like to have memory

– Infinitely large
– Infinitely fast Registers
– Non volatile
• However, we have Memory Hierarchy: Cache

Main Memory
Disk
2
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory Manager

• The part of the OS that manages the Memory

Hierarchy
– Which part of memory is in use and which is not in use
– Allocate memory
– De-allocate memory
– Swapping between main memory and disks

3
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory Management Requirements

• Speed is not the only issue

– Relocation
– Protection
– Sharing
– Physical memory organisation
– ….

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Requirements: Relocation

• The programmer does not know where the

program will be placed in memory when it
is executed,
– it may be swapped to disk and return to main
memory at a different location (relocated)

• Memory references must be translated to

the actual physical memory address

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Relocation and Protection 800K

Partition 4
700K
• Different jobs will run at different addresses
Partition 3
• Suppose, the first instruction is
– call a procedure at absolute address 10 within the 400K
binary file
• If the process is loaded in 1st partition Partition 2

200K
Partition 1
Operating 100K
System
0

This call will jump inside the Operating system

6
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Relocation and Protection 800K

Partition 4
700K
• Solution:
• If the program is loaded in 1st partition, then Partition 3

– Call 100k + 10
• If the program is loaded in 2nd partition, then 400K

– Call 200k + 10 Partition 2

• So on… 200K
• This is called Relocation problem Partition 1
Operating 100K
System
0

7
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Relocation and Protection 800K

Partition 4
700K
• Solution:
Partition 3
– Add an offset to each address in the program
– The offset depends on the partition, e.g.,
400K
• if the Program is loaded in partition 1, add
100k to every address Partition 2
• if the Program is loaded in partition 2, add 200K
200k to every address Partition 1
Operating 100K
System
0
•A program can still generate an address that jumps
in the OS or other user’s code
•We need to Protect the code
8
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Requirements: Protection

• Processes should not be able to reference

memory locations in another process
without permission
• Impossible to check absolute addresses at
compile time
• Must be checked at run time

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Relocation and Protection

• Solution:
– Base and Limit Registers
• When a process is scheduled
– Base Register is loaded with the starting
address of the Partition
– Limit Register is loaded with the length of
the Partition

• Before referring to memory

– Add the base register contents to
generated memory address
– Also check against the Limit register for
protection

10
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management
Relocation and Protection
Addition and comparison has to be performed on every address
Use a hardware module

Absolute
Physical Address
Interrupt 800K
Partition 4
700K

Limit Register Comparator

Partition 3

400K
Partition 2
Base Register Adder
200K
Partition 1
Memory Management Unit Operating 100K
(MMU) System
Relative
Logical Address
Address 0

11
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory Management/Organization

• In good old days, memory management was

trivial as there was only one program
– No OS, eveything is responsibility of user program

• For the later uniprogramming systems, memory was

shared between
– A User program
– Operating system

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Uni-programming
• Run just one program at a time User
• Memory is shared between Program
– A User program
– Operating system
Operating
•The user types a command on the prompt
System
•The OS
Copies the program from the disk to memory
Executes the program
After execution, prompt is available again
•The new program is copied into RAM, overwriting the previous one 13
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Multi-programming systems

• Keep more than one process in memory

– Benefits?

• How to organize memory to achieve this?

• One basic approach is to divide memory

into n (usually unequal partitions)

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Multiprogramming with Fixed Partitions

•If the job size < Partition size, memory is wasted

•When a job arrives
its put into the input queue of

The smallest partition large enough to hold the job

15
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management
Multiprogramming with Fixed Partitions

 Disadvantage of Multiple Queues:

 When a large partition is empty
 And queues for small partition is
full
 Small jobs have to wait, even
though plenty of memory is free
 Alternative: Maintain a single
Queue

16
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Multiprogramming with Fixed Partitions

800K
 Whenever a partition becomes free, Partition 4
a job is selected 700K
 Closest to the front of the queue
Partition 3
 Smaller than the partition size
 Undesirable to waste a large
partition for smaller job 400K

 Search the queue, find the largest Partition 2

job that fits is
200K
 Unfair for smaller jobs Partition 1
 A job may not be skipped more than Operating 100K
System
k times 0

17
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management
Internal Fragmentation
Memory that is internal to a partition but not being used
800K
B 4
Partition
Internal Fragmentation 700K

A
Partition 3

Internal Fragmentation
400K

Partition 2

200K
Partition 1
Operating 100K
System 0
18

FAST, National University of Computer and Emerging Sciences, Islamabad

Lecture 33 - Introduction to Memory Management

Swapping
• If not enough space in memory for all the currently active
processes
• Excess processes must be kept on disk
– Fully  Swapping
– Partially  Virtual Memory
• Example of Swapping
– Round Robin Scheduling
• When the quantum expires
– Swap out the currently running process
– Swap in another process to freed memory space

19
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Swapping

• Another Example:
– Priority based scheduling
• If a higher priority process arrives
– Swap out a lower priority process
– Swap in the higher priority process
• When the higher priority process exits
– Swap in the lower priority process

20
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Swapping

The number, size

C and location of the
partition is decided
B D dynamically.
B
A
C A A
D
Operating
System

D is of Higher Priority, A is of Lower Priority!!!

B exits now, Swap in A 21
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management
Swapping
Its possible to combine all the holes into one big hole
This is called COMPACTION
EXTERNAL
E FRAGMENTATION
Memory External to all
partitions is Fragmented
C Enough total memory exists
to satisfy the request
But is not Contiguous
A
E Swapping has
D created Holes
Operating
System
D exits now
No Space for E !!!
Now E has enough space 22
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Compaction
• Compaction involves CPU overhead
• If addresses are not generated relative to the partition location:
– Then Compaction is not possible
• How much memory should be allocated for a process?
– If all the memory is allocated statically in a program
– Then, OS knows exactly the amount of memory to be allocated:
executable code + variables
• However, if memory is allocated dynamically (say using new)

23
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory allocation for Processes

• Problem may occur whenever a process tries to grow
• If a hole is adjacent to the growing process then it can be
allowed to grow
• If another process is adjacent to the growing process, then
– The growing process should be moved to a larger hole
– If a larger hole is not available, then, one or more processes should
be swapped out
– If a process cannot be swapped out (say, there is not enough
space on disk
• The growing process has to wait
• Or should be killed!!!

24
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory allocation for Processes

• If most of the processes grow as they run,
• It is better to allocate a little extra memory for a process

Room for growth

B Actual in use

Room for growth

A Actual in use
Operating
System
25
FAST, National University of Computer and Emerging Sciences, Islamabad
Lecture 33 - Introduction to Memory Management

Memory allocation for Processes

• If processes have two growing segments
– Data
• (as heap for dynamically allocated variables)
– Stack B-Stack
Room for growth
B-Data
•The memory can be used by either of the B-Program
two segments
•If it runs out the process either
Has to be moved to a larger
hole
A-Stack
Or swapped out Room for growth
Or Killed A-Data
A-Program
Operating System
26

FAST, National University of Computer and Emerging Sciences, Islamabad