UNIT 1

AN INTRODUCTION TO OPERATING SYSTEMS

Application software performs specific task for the user.

System software operates and controls the computer system and provides a platform to
run application software.
An operating system is a piece of software that manages all the resources of a computer
system, both hardware and software, and provides an environment in which the user can
execute his/her programs in a convenient and efficient manner by hiding underlying
complexity of the hardware and acting as a resource manager.
Why OS?
1. What if there is no OS?
a. Bulky and complex app. (Hardware interaction code must be in app’s code
base)
b. Resource exploitation by 1 App.
c. No memory protection.
2. What is an OS made up of?
a. Collection of system software.

An operating system function -

- Access to the computer hardware.
- interface between the user and the computer hardware
- Resource management (Aka, Arbitration) (memory, device, file, security, process etc)
- Hides the underlying complexity of the hardware. (Aka, Abstraction)
- facilitates execution of application programs by providing isolation and protection.

User

Application programs
Operating system
Computer hardware

The operating system provides the means for proper use of the resources in the

operation of the computer system.

 LEC-2: Types of OS
OS goals –

• Maximum CPU utilization

• Less process starvation

• Higher priority job execution

Types- of operating systems –

- Single process operating system [MS DOS, 1981]
- Batch-processing operating system [ATLAS, Manchester Univ., ate 1950s – ear y 1960s]
- Multiprogramming operating system [THE, Dijkstra, ar y 1960s]
- Multitasking operating system [CTSS, MIT, arly 1960s]
- Multi-processing operating system [Windows NT]
- Distributed system [LOCUS]

Real time OS [ATCS]

 Single process OS, only 1 process executes at a time from the ready queue. [Oldest]

Batch-processing OS,
2. Firstly, user prepares his job using punch cards.

1.
3. Then, he submits the job to the computer operator.
Operator collects the jobs from different users and sort the jobs into batches with
4. similar needs.
5. Then, operator submits the batches to the processor one by one.
CodeHelp

-All the jobs of one batch are executed together.

- Priorities cannot be set, if a job comes with some higher priority.
- May lead to starvation. (A batch may take more time to com ete)

CPU may become idle in case of I/O operations.

Multiprogramming
memory
increases CPU utilization by keeping multiple jobs (code and data)
in the so that the CPU always has one to x cute in case some job gets busy with

I/O.-
-
Single CPU
- Context switching for proc ss s.
- Switch happens wh n curr nt proc ss go s to wait state.
CPU i le time r uc .

Multitasking is a logical xt nsion of

multipr- gramming.
- Single CPU
Able to run m re than ne task
- simultane usly.
- ntext switching and time sharing used.
- Increases resp nsiveness.
PU idle time is further reduced.

Multi-processing OS, more than 1 CPU in a single computer.

- Increases reliability, 1 CPU fails, other

- can work
- Better throughput.
Lesser process starvation, (if 1 CPU is
working on some process, other can be
executed on other CPU.
Distributed OS,
- OS manages many bunches of resources, >=1 CPUs, >=1 memory, >=1 GPUs, etc
- Loosely connected autonomous, interconnected computer nodes.
- collection of independent, networked, communicating, and physically separate
computational nodes.
RTOS
- Real time error free, computations within tight-time boundaries.
- Air Traffic control system, ROBOTS etc.
 LEC-3: Multi-Tasking vs Multi-Threading

Program: A Program is an executable file which contains a certain set of instructions written
to complete the specific job or operation on your computer.
• It’s a compiled code. Ready to be executed.
• Stored in Disk

Process: Program under execution. Resides in Computer’s primary memory (RAM).

Thread:
• Single sequence stream within a process.
CodeHelp

• An independent path of execution in a process.

• Light-weight process.
• Used to achieve parallelism by dividing a process’s tasks which are inde endent ath
of execution.
• E.g., Multiple tabs in a browser, text editor (Wh n you are ty ing in an editor, s ell-
checking, formatting of text and saving the t xt are done concurr nt y by multi le
threads.)

Multi-Tasking Multi-Thr ading

The execution of more than one task A proc ss is divid d into s veral different
simultaneously is called as multitasking. sub-tasks called as threads, which has its
own path of execution. This concept is
call as multithreading.
Concept of more than 1 proc ss s b ing Conc pt of more than 1 thread. Threads are
context switched. cont xt switched.
No. of CPU 1. No. of CPU >= 1. (Better to have more than
1)
Isolation and mem ry pr tecti n exists. No isolation and memory protection,
OS must all cate separate mem ry and resources are shared among threads of that
resources to each pr gram that CPU is process.
executing. OS allocates memory to a process; multiple
threads of that process share the same
memory and resources allocated to the
process.

Scheduling:
Threads are scheduled for execution based on their priority. Even though threads are
executing within the runtime, all threads are assigned processor time slices by the operating
system.

Difference between Thread Context Switching and Process Context Switching:

Thread Context switching Process context switching
OS saves current state of thread & switches OS saves current state of process &
to another thread of same process. switches to another process by restoring its
state.
Doesn’t includes switching of memory Includes switching of memory
address space. address space.
(But Program counter, registers & stack are
included.)
Fast switching. Slow switching.
CPU’s cache state is preserved. CPU’s cache state is flushed.
 LEC-4: Components of OS

1. Kernel: A kernel is that part of the operating system which interacts directly with the

hardware and performs the most crucial tasks.

a. Heart of OS/Core component

b. Very first part of OS to load on start-up.
2. User space: Where application software runs, apps don’t have privileged access to the
underlying hardware. It interacts with kernel.
a. GUI
b. CLI

A shell, also known as a command interpreter, is that part of the operating system that

receives commands from the users and gets them executed.

Functions of Kernel:

Types CodeHelp
1. Process management:
a.
Scheduling processes and threads on the CPUs.
b.

c. Creating & deleting both user and system process.

d. Suspending and resuming processes
Providing mechanisms for process synchronization or process
2. communication.
Memory management:
a.
Allocating and deallocating memory space as per need.
b.

Keeping track of which part of memory are currently being used and by
3. which process.
File management:
a.
Creating and deleting fil s.
b.

c. Creating and deleting dir ctori s t organize files.

d. Mapping files into s condary storage.
4. Backup support onto a stabl storage media.
I/O management: to manage and control I/O perations and I/O devices
a. Buffering ( ata copy between two devices), caching and spooling.
i. Spooling

1. Within differing speed two jobs.

2. Eg. Print spooling and mail spooling.

ii. Buffering
1. Within one job.
iii. 2. Eg. Youtube video buffering
aching

1. Memory caching, Web caching etc.

1. Monolithic kernel
a. All functions are in kernel itself.
b. Bulky in size.
c. Memory required to run is high.
d. Less reliable, one module crashes -> whole kernel is down.
e. High performance as communication is fast. (Less user mode, kernel mode
overheads)
f. Eg. Linux, Unix, MS-DOS.
 2. Micro Kernel
a. Only major functions are in kernel.
i. Memory mgmt.
ii. Process mgmt.
b. File mgmt. and IO mgmt. are in User-space.
c. smaller in size.
d. More Reliable
e. More stable
f. Performance is slow.
g. Overhead switching b/w user mode and kernel mode.

h. Eg. L4 Linux, Symbian OS, MINIX etc.

3. Hybrid Kernel: Advantages of both worlds. (File mgmt. in User space and rest in Kernel
a.
b. space. )
c. Combined approach.
CodeHelp

d. Speed and design of mono.

e. Modularity and stability of micro.
f. Eg. MacOS, Windows NT/7/10
4. IPC also happens but lesser overheads

Nano/Exo kernels…
Q.

How will communication happen between user m and k rn l mod ?

IPC

Ans. Inter process communication ( ).

1. Two processes executing independently, having independent memory space (Memory

2. protection), But some may need to communicate. to work.

Done by shared memory and message passing

 LEC-5: System Calls

How do apps interact with Kernel? -> using system calls.

Eg. Mkdir laks

- Mkdir indirectly calls kernel and asked the file mgmt. module to create a new directory.
- Mkdir is just a wrapper of actual system calls.
- Mkdir interacts with kernel using system calls.

Eg. Creating a process.

- User executes a process. (User space)
- Gets system call. (US)
- Exec system call to create a process. (KS)
- Return to US.

Transitions from US to KS done by software interrupts.

System calls are implemented in C.

A system call is a mechanism using which a user program can request a service from the kernel
for which it does not have the permission to perform.
User programs typically do not have permission to perform operations like accessing I/O devices
and communicating other programs.

System Calls are the only way through which a process can go into kernel mode from user mode.
Types of System Calls:
1) Process Control
a. end, abort
b. load, execute
c. create process, terminate process
d. get process attributes, set process attributes
e. wait for time
f. wait event, signal event
g. allocate and free memory

2) File Management
a. create file, delete file
b. open, close
c. read, write, reposition
d. get file attributes, set file attributes

3) Device Management
a. request device, release device
b. read, write, reposition
c. get device attributes, set device attributes
d. logically attach or detach devices

4) Information maintenance
a. get time or date, set time or date
b. get system data, set system data
c. get process, file, or device attributes
d. set process, file, or device attributes

5) Communication Management
a. create, delete communication connection
b. send, receive messages
c. transfer status information
d. attach or detach remote devices

Examples of Windows & Unix System calls:

Category Windows Unix

Process Control CreateProcess() fork()

ExitProcess() exit()
WaitForSingleObject() wait()

File Management CreateFile() open ()

ReadFile() read ()
WriteFile() write ()
CloseHandle() close ()
SetFileSecurity() chmod()
InitlializeSecurityDescriptor() umask(
SetSecurityDescriptorGroup() chown()

Device Management SetConsoleMode() ioctI()

ReadConsole() read()
WriteConsole() write()

Information Management GetCurrentProcessID() getpid ()

SetTimer() alarm ()
Sleep() sleep ()

Communication CreatePipe() pipe ()

CreateFileMapping() shmget ()
MapViewOfFile() mmap()
 LEC-6: What happens when you turn on your computer?

i. PC On

ii. CPU initializes itself and looks for a firmware program (BIOS) stored in
BIOS Chip (Basic input-output system chip is a ROM chip found on
mother board that allows to access & setup computer system at most
basic level.)
1. In modern PCs, CPU loads UEFI (Unified extensible firmware
interface)
iii. CPU runs the BIOS which tests and initializes system hardware. Bios
loads configuration settings. If something is not appropriate (like
missing RAM) error is thrown and boot process is stopped.
This is called POST (Power on self-test) process.
CodeHelp

(UEFI can do a lot more than just initialize hardware; it’s really a tiny
operating system. For example, Intel CPUs have the Intel Management
Engine. This provides a variety of features, including powering Inte ’s
Active Management Technology, which allows for remote management
iv. of business PCs.)

will handoff responsibility for booting your PC to your OS’s

BIOS

bootloader.

1. BIOS looked at the MBR (master boot r cord), a sp cial boot

sector at the beginning of a disk. The MBR contains code that
loads the rest of the operating syst m, known as a “bootload r.”
The BIOS executes the bootloader, which takes it from there and
begins booting the actual operating system—Windows or Linux,
for example.
In other words,
the BIOS or UEFI xamin s a storage device on your system to
look for a small program, ith r in th MBR or on an EFI system
v. partition, and runs it.
The bootloa er is a small program that has the large task of booting the
(Boots Kern l then, User Space)
rest f the operating syst m . Windows
uses a b tl a er named Win ows Boot Manager (Bootmgr.exe), most

Linux systems use GRUB, and Macs use something called boot.efi
 Lec-7: 32-Bit vs 64-Bit OS

1. A 32-bit OS has 32-bit registers, and it can access 2^32 unique memory
addresses. i.e., 4GB of physical memory.
2. A 64-bit OS has 64-bit registers, and it can access 2^64 unique memory addresses. i.e.,
17,179,869,184 GB of physical memory.
3. 32-bit CPU architecture can process 32 bits of data & information.
4. 64-bit CPU architecture can process 64 bits of data & information.
5. Advantages of 64-bit over the 32-bit operating system:
a. Addressable Memory: 32-bit CPU -> 2^32 memory addresses, 64-bit CPU -> 2^64
CodeHelp

memory addresses.
b. Resource usage: Installing more RAM on a system with a 32-bit OS doesn't im act
performance. However, upgrade that system with excess RAM to the 64-bit version of
Windows, and you'll notice a difference.
c. Performance: All calculations take place in the registers. When you’re erforming math in
your code, operands are loaded from memory into r gist rs. So, having arger registers
allow you to perform larger calculations at the same time.
32-bit processor can execute 4 bytes of data in 1 instruction cyc whi 64-bit means that
processor can execute 8 bytes of data in 1 instruction cycle.
(In 1 sec, there could be thousands to billons of instruction cycl s d p nding upon a
processor design)
d. Compatibility: 64-bit CPU can run both 32-bit and 64-bit OS. While 32-bit CPU can only
run 32-bit OS.
. Better Graphics p rformance: 8-byt s graphics calculations make graphics-intensive apps
run faster.
 Lec-8: Storage Devices Basics

What are the different memory present in the computer system?

CodeHelp
1. Register: Smallest unit of storage. It is a part of CPU itself.
A register may hold an instruction, a storage addr ss, or any data (such as bit sequence or individual
characters).
Registers are a type of comput r m mory us to quickly accept, store, and transfer data and
instructions that are being us imm iat ly by the CPU.
2. Cache: Additional memory syst m that t mporarily stores frequently used instructions and data for
quicker processing by the CPU.
3. Main Mem ry: RAM.
4. Sec ndary Mem ry: St rage me ia, on which computer can store data & programs.

Comparison
1. ost:
a. Primary st rages are c stly.
b. Registers are most expensive due to expensive semiconductors & labour.
c. Secondary storages are cheaper than primary.
2. Access Speed:
a. Primary has higher access speed than secondary memory.
b. Registers has highest access speed, then comes cache, then main memory.
3. Storage size:
a. Secondary has more space.
4. Volatility:
a. Primary memory is volatile.
b. Secondary is non-volatile.
 Lec-9: Introduction to Process

1. What is a program? Compiled code, that is ready to execute.

2. What is a process? Program under execution.
3. How OS creates a process? Converting program into a process.
STEPS:
a. Load the program & static data into memory.
b. Allocate runtime stack.
c. Heap memory allocation.
d. IO tasks.
e. OS handoffs control to main ().
CodeHelp

4. Architecture of process:

5. Attributes of process:
a. Feature that allows i ntifying a proc ss uniquely.
b. Process table
i. All proc ss s are b ing track by OS using a table like data structure.
ii. Each entry in that table is process control block (PCB).
c. PCB: St res inf /attributes of a process.
i. Data structure used for each process, that stores information of a process such as
pr cess id, pr gram counter, process state, priority etc.
6. P B structure:

Registers in the PCB, it is a data structure. When a processes is running and it's time slice expires,
the current value of process specific registers would be stored in the PCB and the process would be
swapped out. When the process is scheduled to be run, the register values is read from the PCB and
written to the CPU registers. This is the main purpose of the registers in the PCB.
 Lec-10: Process States | Process Queues

1. Process States: As process executes, it changes state. Each process may

be in one of the following states.
a. New: OS is about to pick the program & convert it into process. OR the
process is being created.
b. Run: Instructions are being executed; CPU is allocated.
c. Waiting: Waiting for IO.
d. Ready: The process is in memory, waiting to be assigned to a processor.
e. Terminated: The process has finished execution. PCB entry removed from process table.

2. Process Queues: a. Job Queue:

i. Process s in n w state.
ii. Present in s condary m mory.
iii. Job Sch ular (Long t rm sch dular (LTS)) picks process
from the pool and loa s th m into m mory for x cution.
b. Ready Queue:
i. Pr cesses in Rea y state.
ii. Present in main memory.
iii. CPU Sche ular (Short-term schedular) picks process from
ready queue and dispatch it to CPU.
c. Waiting Queue:
i. Pr cesses in Wait state.
3. Degree of multi-programming: The number of processes in the memory.
a. LTS controls degree of multi-programming.
4. Dispatcher: The module of OS that gives control of CPU to a process selected by STS.
 LEC-11: Swapping | Context-Switching | Orphan process | Zombie process

1. Swapping
a. Time-sharing system may have medium term schedular (MTS).
b. Remove processes from memory to reduce degree of multi-programming.
c. These removed processes can be reintroduced into memory, and its execution can be continued
where it left off. This is called Swapping.
d. Swap-out and swap-in is done by MTS.
e. Swapping is necessary to improve process mix or because a change in memory requirements has
overcommitted available memory, requiring memory to be freed up.
f. Swapping is a mechanism in which a process can be swapped temporarily out of main memory (or
move) to secondary storage (disk) and make that memory available to other processes. At some
CodeHelp

later time, the system swaps back the process from the secondary storage to main memory.

2. Context-Switching
a. Switching the CPU to anoth r proc ss r quires p rforming a state save of the current process and a
state restore of a diff r nt proc ss.
b. When this occurs, the k rn l saves the cont xt of the old process in its PCB and loads the saved
context of the new proc ss sch ul to run.
c. It is pure overhea , b cause the syst m do s no useful work while switching.
d. Speed varies from machine to machin , d p nding on the memory speed, the number of registers
that must be c pied etc.
3. Orphan pr cess
a. The pr cess wh se parent process has been terminated and it is still running.
b. Orphan pr cesses are ad pted by init process.
c. Init is the first pr cess f OS.
4. Zombie pr cess / Defunct pr cess
a. A zombie process is a process whose execution is completed but it still has an entry in the process
table.
b. Zombie processes usually occur for child processes, as the parent process still needs to read its
child’s exit status. Once this is done using the wait system call, the zombie process is eliminated
from the process table. This is known as reaping the zombie process.
c. It is because parent process may call wait () on child process for a longer time duration and child
process got terminated much earlier.
. As entry in the process table can only be removed, after the parent process reads the exit status of
child process. Hence, the child process remains a zombie till it is removed from the process table.
 LEC-12: Intro to Process Scheduling | FCFS | Convoy Effect

1. Process Scheduling
a. Basis of Multi-programming OS.
b. By switching the CPU among processes, the OS can make the computer more productive.
c. Many processes are kept in memory at a time, when a process must wait or time quantum expires,
the OS takes the CPU away from that process & gives the CPU to another process & this pattern
continues.
2. CPU Scheduler
a. Whenever the CPU become ideal, OS must select one process from the ready queue to be executed.
b. Done by STS.
3. Non-Preemptive scheduling
CodeHelp

a. Once CPU has been allocated to a process, the process keeps the CPU until it releases CPU either by
terminating or by switching to wait-state.
b. Starvation, as a process with long burst time may starve ess burst time rocess.
c. Low CPU utilization.
4. Preemptive scheduling
a. CPU is taken away from a process after time quantum xpir s a ong with t rminating or switching
to wait-state.
b. Less Starvation
c. High CPU utilization.
5. Goals of CPU scheduling
a. Maximum CPU utilization
b. Minimum Turnaround time (TAT).
c. Min. Wait-time
d. Min. response time.
e. Max. throughput of syst m.
6. Throughput: No. of process s compl t p r unit time.
7. Arrival time (AT): Time wh n proc ss is arriv at th r ady queue.
8. Burst time (BT): The time require by the proc ss for its xecution.
9. Turnar und time (TAT): Time taken from first time process enters ready state till it terminates. (CT - AT)
10. Wait time (WT): Time pr cess spen s waiting for CPU. (WT = TAT – BT)
11. Response time: Time urati n between process getting into ready queue and process getting CPU for the
first time.
12. ompleti n Time ( T): Time taken till process gets terminated.
13. F FS (First come-first serve):
a. Whichever process comes first in the ready queue will be given CPU first.
b. In this, if one rocess has longer BT. It will have major effect on average WT of diff processes, called
onvoy effect.
c. onvoy Effect is a situation where many processes, who need to use a resource for a short time, are
blocked by one rocess holding that resource for a long time.
i. This cause poor resource management.
 LEC-13: CPU Scheduling | SJF | Priority | RR
1. Shortest Job First (SJF) [Non-preemptive]
a. Process with least BT will be dispatched to CPU first.
b. Must do estimation for BT for each process in ready queue beforehand,
Correct estimation of BT is an impossible task (ideally.)
c. Run lowest time process for all time then, choose job having lowest BT at that instance.
d. This will suffer from convoy effect as if the very first process which came is
Ready state is having a large BT.
e. Process starvation might happen.
f. Criteria for SJF algos, AT + BT.
2. SJF
CodeHelp
[Preemptive]
a. Less starvation.
b. No convoy effect.
c. Gives average WT less for a given set of processes as schedu ing short job before a long one
decreases the WT of short job more than it increases the WT of the ong rocess.
3. Priority Scheduling [Non-preemptive]
a. Priority is assigned to a process when it is creat d.
b. SJF is a special case of general priority scheduling with priority inv rs y ro ortional to BT.
4. Priority Scheduling [Preemptive]
a. Current RUN state job will be preempted if n xt job has high r priority.
b. May cause indefinite waiting (Starvation) for low r priority jobs. (Possibility is they won’t get
executed ever). (True for both preemptive and non-pr mptive v rsion)
i. Solution: Ageing is the solution.
ii. Gradually increase priority of process that wait so long. E.g., increase priority by 1 every 15
minutes.
5. Round robin scheduling (RR)
a. Most popular
b. Like FCFS but pre mptive.
c. Designed for time sharing syst ms.
. Criteria: AT + time quantum (TQ), Do sn’t d p nd on BT.
e. No pr cess is g ing to wait forever, hence very low starvation. [No convoy effect]
f. Easy to implement.
g. If TQ is small, m re will be the context switch (more overhead).
 LEC-14: MLQ | MLFQ

1. Multi-level queue scheduling (MLQ)

a. Ready queue is divided into multiple queues depending upon priority.
b. A process is permanently assigned to one of the queues (inflexible) based on
some property of process, memory, size, process priority or process type.
c. Each queue has its own scheduling algorithm. E.g., SP -> RR, IP -> RR & BP -> FCFS.

CodeHelpd.Systemprocess:CreatedbyOS(Highestpriority)

Interactive process (Foreground process): Needs user input (I/O).

Batch process (Background process): Runs silently, no us r input r quir d.
e. Scheduling among different sub-queues is implem nt d as fix d priority preemptive
scheduling. E.g., foreground queue has absolute priority ov r background queue.
f. If an interactive process comes & batch process is currently executing. Then, batch process will
be preempted.
g. Problem: Only after completion of all the processes from the top-level ready queue, the
further level ready queues will be scheduled.
This came starvation for low r priority proc ss.
h. Convoy effect is pres nt.

2. Multi-level feedback queue sch uling (MLFQ)

a. Multiple sub-queues are present.
b. Allows the process to move between queues. The idea is to separate processes according to
the characteristics f their BT. If a process uses too much CPU time, it will be moved to
lower
pri rity queue. This scheme leaves I/O bound and interactive processes in the higher-priority
queue.
In additi n, a pr cess that waits too much in a lower-priority queue may be
moved to a higher priority queue. This form of ageing prevents starvation.
c. Less starvation then MLQ.
. It is flexible.
e. an be configured to match a specific system design requirement.
Sample MLFQ design:
 µ

CodeHelp
3. Comparison:
FCFS SJF PSJF Priority P- RR MLQ MLFQ
Priority
Design Simple Complex Complex Complex Complex Sim Com ex Com lex
Preemption No No Yes No Yes Yes Yes Yes
Convoy Yes Yes No Yes Ys No Yes Yes
effect
Overhead No No Yes No Ys Ys Yes Yes
 LEC-15: Introduction to Concurrency
1. Concurrency is the execution of the multiple instruction sequences at the same time. It
happens in the operating system when there are several process threads running in parallel.
2. Thread:
• Single sequence stream within a process.
• An independent path of execution in a process.
• Light-weight process.
• Used to achieve parallelism by dividing a process’s tasks which are
independent path of execution.
• E.g., Multiple tabs in a browser, text editor (When you are typing in an editor, spell
checking, formatting of text and saving the text are done concurrently by multi le threads.)
3. Thread Scheduling: Threads are scheduled for execution based on their
riority. Even though threads are executing within the runtime, all threads
are assigned rocessor time s ices by the operating system.
4. Threads context switching
• OS saves current state of thread & switch s to anoth r thr ad of same process.
• Doesn’t includes switching of memory addr ss space. (But Program
counter, registers & stack are included.)
• Fast switching as compared to process switching
• CPU’s cache state is preserved.
5. How each thread get access to the CPU?
• Each thread has its own program count r.
• Depending upon the thr ad sch duling algorithm, OS schedule these threads.
• OS will fetch instructions corr sponding to PC of that thread and execute instruction.
6. I/O or TQ, based context switching is one h re as well
• We have TCB (Thr ad control block) like PCB for state storage
management while perf rming c ntext switching.
7. Will single CPU system w uld gain by multi-threading technique?
• Never.
• As two threads have to c ntext switch for that single CPU.
• This w n’t give any gain.
8. Benefits f Multi-threading.
• Responsiveness
• Resource sharing: Efficient resource sharing.
• Economy: It is more economical to create and context switch threads.
1. Also, allocating memory and resources for process creation is
costly, so better to divide tasks into threads of same process.
• Threads allow utilization of multiprocessor architectures to a greater scale and efficiency.
 LEC-16: Critical Section Problem and How to address it

1. Process synchronization techniques play a key role in maintaining

the consistency of shared data
2. Critical Section (C.S)
a. The critical section refers to the segment of code where processes/threads
access shared resources, such as common variables and files, and perform
write operations on them. Since processes/threads execute concurrently, any
process can be interrupted mid-execution.
3. Major Thread scheduling issue
a. Race Condition
i. A race condition occurs when two or more threads can access shared
data and they try to change it at the same time. Because the thread
scheduling algorithm can swap between threads at any time, you don't
know the order in which the threads wi attem t to access the shared
data. Therefore, the r sult of the change in data is de endent on the
thread scheduling algorithm, i. ., both thr ads are "racing" to
access/change the data.
4. Solution to Race Condition
a. Atomic operations: Make Critical code s ction an atomic op ration, i.e.,
Executed in one CPU cycle.
b. Mutual Exclusion using locks.
c. Semaphores
5. Can we use a simple flag variable to solve the problem of race condition?
a. No.
6. Peterson’s solution can be us to avoid race condition but holds good for only 2
process/threa s.
7. Mutex/Locks
a. L cks can be used to implement mutual exclusion and avoid race condition
by all wing nly ne threa /process to access critical section.
b. Disa vantages:
i. ntenti n: ne thread has acquired the lock, other threads will be
busy waiting, what if thread that had acquired the lock dies, then all
ther threads will be in infinite waiting.
ii. Deadlocks
iii. Debugging
CodeHelpiv.Starvationofhighprioritythreads.
LEC-17: Conditional Variable and Semaphores for Threads synchronization

1. Conditional variable
a. The condition variable is a synchronization primitive that lets the thread wait
until a certain condition occurs.
b. Works with a lock
c. Thread can enter a wait state only when it has acquired a lock. When a
thread enters the wait state, it will release the lock and wait until another
thread notifies that the event has occurred. Once the waiting thread enters
the running state, it again acquires the lock immediately and starts executing.
d. Why to use conditional variable?
CodeHelp

i. To avoid busy waiting.

e. Contention is not here.
2. Semaphores
a. Synchronization method.
b. An integer that is equal to number of r sourc s
c. Multiple threads can go and execute C.S concurr nt y.
d. Allows multiple program threads to acc ss the finite instance of resources
whereas mutex allows multiple thr ads to acc ss a single shared resource
one at a time.
e. Binary semaphore: value can be 0 or 1.
i. Aka, mutex locks
f. Counting semaphore
i. Can range ov r an unr strict d domain.
ii. Can be us d to control acc ss to a given resource consisting of a finite
numb r of instanc s.
g. To overcome the n for busy waiting, we can modify the definition of
the wait () and signal () s maphore op rations. When a process executes the
wait () perati n and fin s that the semaphore value is not positive, it must
wait. H wever, rather than engaging in busy waiting, the process car block
itself. The bl ck- peration places a process into a waiting queue associated
with the semaph re, and the state of the process is switched to the Waiting
state. Then c ntr is transferred to the CPU scheduler, which selects another
pr cess to execute.
h. A process that is blocked, waiting on a semaphore S, should be restarted
when some other process executes a signal () operation. The process is
restarted by a wakeup () operation, which changes the process from the
waiting state to the ready state. The process is then placed in the ready
queue.
 Lec-20: The Dining Philosophers problem

CodeHelp
1. We have 5 philosophers.
2. They spend their life just b ing in two stat s:
a. Thinking
b. Eating
3. They sit on a circular table surroun by 5 chairs (1 each), in the center of table is a bowl of
noodles, and the table is laid with 5 single forks.
4. Thinking state: When a ph. Thinks, he doesn’t interact with others.
5. Eating state: When a ph. Gets hungry, he tries to pick up the 2 forks adjacent to him (Left and
Right). He can pick ne f rk at a time.
6. One can’t pick up a f rk if it is already taken.
7. When ph. Has b th f rks at the same time, he eats without releasing forks.
8. S luti n can be given using semaphores.
a. Each fork is a binary semaphore.
b. A ph. alls wait() operation to acquire a fork.
c. Release fork by calling signal().
. Semaphore fork[5]{1};
9. Although the semaphore solution makes sure that no two neighbors are
eating simultaneously but it could still create Deadlock.
10. Suppose that all 5 ph. Become hungry at the same time and each picks up
their left fork, then All fork semaphores would be 0.
11. When each ph. Tries to grab his right fork, he will be waiting for ever (Deadlock)
12. We must use some methods to avoid Deadlock and make the solution work
a. Allow at most 4 ph. To be sitting simultaneously.
b. Allow a ph. To pick up his fork only if both forks are available and to
do this, he must pick them up in a critical section (atomically).
c. Odd-even rule.
an odd ph. Picks up first his left fork and then his right fork,
whereas an even ph. Picks up his right fork then his left fork.
13. Hence, only semaphores are not enough to solve this problem.
We must add some enhancement rules to make deadlock free solution.
 LEC-21: Deadlock Part-1

1. In Multi-programming environment, we have several processes competing for finite number of

resources
2. Process requests a resource (R), if R is not available (taken by other process), process enters in a
waiting state. Sometimes that waiting process is never able to change its state because the resource,
it has requested is busy (forever), called DEADLOCK (DL)
3. Two or more processes are waiting on some resource’s availability, which will never be available as
it is also busy with some other process. The Processes are said to be in Deadlock.
4. DL is a bug present in the process/thread synchronization method.
5. In DL, processes never finish executing, and the system resources are tied up, preventing other jobs
CodeHelp

from starting.
6. Example of resources: Memory space, CPU cycles, files, locks, sockets, IO devices etc.
7. Single resource can have multiple instances of that. E.g., CPU is a resource, and a system can have 2
CPUs.
8. How a process/thread utilize a resource?
a. Request: Request the R, if R is free Lock it, lse wait ti it is avai ab .
b. Use
c. Release: Release resource instance and make it available for oth r processes

9. Deadl ck Necessary C n iti n: 4 Con ition should hold simultaneously.

a. Mutual Exclusi n
i. Only 1 pr cess at a time can use the resource, if another process requests that
res urce, the requesting process must wait until the resource has been released.
b. H ld & Wait
i. A process must be holding at least one resource & waiting to acquire additional
resources that are currently being held by other processes.
c. No-preemption
i. Resource must be voluntarily released by the process after completion of
execution. (No resource preemption)
. Circular wait
i. A set {P0, P1, … ,Pn} of waiting processes must exist such that P0 is waiting for a
resource held by P1, P1 is waiting for a resource held by P2, and so on.
10. Methods for handling Deadlocks:
a. Use a protocol to prevent or avoid deadlocks, ensuring that the
system will never enter a deadlocked state.
b. Allow the system to enter a deadlocked state, detect it, and recover.
 c. Ignore the problem altogether and pretend that deadlocks never occur
in system. (Ostrich algorithm) aka, Deadlock ignorance.
11.To ensure that deadlocks never occur, the system can use either a
deadlock prevention or deadlock avoidance scheme.
12. Deadlock Prevention: by ensuring at least one of the necessary
conditions cannot hold. a. Mutual exclusion
i. Use locks (mutual exclusion) only for non-sharable resource.
ii. Sharable resources like Read-Only files can be
accessed by multiple processes/threads.
iii. However, we can’t prevent DLs by denying the mutual-exclusion condition,
CodeHelp
because some resources are intrinsically non-sharable.
b. Hold & Wait
i. To ensure H&W condition never occurs in the system, we must guarantee that,
whenever a process requests a resource, it doesn’t ho d any other resource.
ii. Protocol (A) can be, each process has to request and be a ocated a l its resources
before its execution.
iii. Protocol (B) can be, allow a proc ss to r qu st r sourc s on y when it has none. It
can request any additional resourc s aft r it must have r as d a the resources
that it is currently allocated.
c. No preemption
i. If a process is holding some resources and request another resource that cannot
be immediately allocated to it, then all the resources the process is currently
holding are pr mpt d. The proc ss will restart only when it can regain its old
resourc s, as w ll as the n w one that it is requesting. (Live Lock may occur).
ii. If a proc ss r qu sts some r sourc s, we first check whether they are available. If
yes, we allocate th m. If not, we ch ck whether they are allocated to some other
process that is waiting for a ditional resources. If so, preempt the desired resource
fr m waiting pr cess and allocate them to the requesting process.
d. Circular wait
i. To ensure that this con ition never holds is to impose a proper ordering of
res urce all cati n.
ii. P1 and P2 b th require R1 and R1, locking on these resources should be like, both
try to ck R1 then R2. By this way which ever process first locks R1 will get R2.
 LEC-22: Deadlock Part-2

1. Deadlock Avoidance: Idea is, the kernel be given in advance info concerning which resources will
use in its lifetime.
By this, system can decide for each request whether the process should wait.
To decide whether the current request can be satisfied or delayed, the system must consider the
resources currently available, resources currently allocated to each process in the system and the
future requests and releases of each process.
a. Schedule process and its resources allocation in such a way that the DL never occur.
b. Safe state: A state is safe if the system can allocate resources to each process (up to its
CodeHelp

max.) in some order and still avoid DL.

A system is in safe state only if there exists a safe sequence.
c. In an Unsafe state, the operating system cannot prevent rocesses from requesting
resources in such a way that any deadlock occurs. It is not necessary that a unsafe states
are deadlocks; an unsafe state may lead to a dead ock.
d. The main key of the deadlock avoidance m thod is wh n v r the r quest is made for
resources then the request must only be approv d on y in the case if the resulting state is a
safe state.
e. In a case, if the system is unable to fulfill the r qu st of all proc ss s, then the state of the
system is called unsafe.
f. Scheduling algorithm using which DL can be avoided by finding safe state. (Banker
Algorithm)
2. Banker Algorithm
a. When a process r qu sts a s t of r sourc s, the system must determine whether allocating
these resources will ave the syst m in a safe state. If yes, then the resources may be
allocated to the proc ss. If not, th n the proc ss must wait till other processes release
enough resourc s.
3. Deadl ck Detecti n: Systems haven’t implemented deadlock-prevention or a deadlock avoidance
technique, then they may empl y DL etection then, recovery technique.
a. Single Instance f Each resource type (wait-for graph method)
i. A deadl ck exists in the system if and only if there is a cycle in the wait-for graph.
In rder to detect the deadlock, the system needs to maintain the wait-for graph
and peri dically system invokes an algorithm that searches for the cycle in the
wait-for graph.
b. Multiple instances for each resource type
i. Banker Algorithm
4. Recovery from Deadlock
a. Process termination
i. Abort all DL processes
ii. Abort one process at a time until DL cycle is eliminated.
b. Resource preemption
i. To eliminate DL, we successively preempt some resources from processes and

give these resources to other processes until DL cycle is broken.

 LEC-24: Memory Management Techniques | Contiguous Memory Allocation

1. In Multi-programming environment, we have multiple processes in the main memory (Ready

Queue) to keep the CPU utilization high and to make computer responsive to the users.
2. To realize this increase in performance, however, we must keep several processes in the memory; that is, we
must share the main memory. As a result, we must manage main memory for all the different processes.

3. Logical versus Physical Address Space

a. Logical Address
i. An address generated by the CPU.
ii. The logical address is basically the address of an instruction or data used by a process.
CodeHelp
iii. User can access logical address of the process.
iv. User has indirect access to the physical address through logical address.
v. Logical address does not exist physically. Hence, aka, Virtual address.
vi. The set of all logical addresses that are generated by any rogram is referred to as Logical
Address Space.
vii. Range: 0 to max.
b. Physical Address
i. An address loaded into the memory-addr ss r gist r of the hysical memory.
ii. User can never access the physical addr ss of the Program.
iii. The physical address is in the memory unit. It’s a location in the main memory physically.
iv. A physical address can be accessed by a us r indir ctly but not dir ctly.
v. The set of all physical addresses corr sponding to the Logical addr sses is commonly
known as Physical Address Space.
vi. It is comput d by the M mory Manag ment Unit (MMU).
vii. Range: (R + 0) to (R + max), for a base value R.
c. The runtime mapping from virtual to physical address is done by a hardware device called the
memory-manag m nt unit (MMU).
d. The user's program mainly g n rat s the logical address, and the user thinks that the program is
running in this logical a r ss, but the program mainly needs physical memory in order to
complete its execution.

4. How OS manages the isolation and protect? (Memory Mapping and Protection)
a. OS provides this Virtual Address Space (VAS) concept.
b. To separate memory space, we need the ability to determine the range of legal addresses that the
process may access and to ensure that the process can access only these legal addresses.
c. The relocation register contains value of smallest physical address (Base address [R]); the limit
register contains the range of logical addresses (e.g., relocation = 100040 & limit = 74600).
d. Each logical address must be less than the limit register.
e. MMU maps the logical address dynamically by adding the value in the relocation register.
f. When CPU scheduler selects a process for execution, the dispatcher loads the relocation and
limit registers with the correct values as part of the context switch. Since every address generated by the CPU
(Logical address) is checked against these registers, we can protect both OS and other users’ programs and
data from being modified by running process.
g. Any attempt by a program executing in user mode to access the OS memory or other uses’
memory results in a trap in the OS, which treat the attempt as a fatal error.
h. Address Translation

5. CodeHelpAllocationMethodonPhysicalMemory

a. Contiguous Allocation
b. Non-contiguous Allocation
6. Contiguous Memory Allocation
a. In this scheme, each proc ss is contain d in a single contiguous block of memory.
Fixed Partitioningi.Thmain m mory is divid d into partitions of equal or different sizes.
b. ii.iii.Limitations:1.InternalFragmentation:ifthesizeoftheprocessislesserthen the total size ofthepartitionthensomesizeofthepartitiongetswastedandremainunused.Thisiswastageofthememoryandcalledinternalfragmentation.2.ExternalFragmentation:Thetotalunusedspaceofvariouspartitionscannotbe

used to load the processes even though there is space

available but not in the contiguous form.
3. Limitation on process size: If the process size is larger than the size
of maximum sized partition then that process cannot be loaded into
the memory. Therefore, a limitation can be imposed on the process
size that is it cannot be larger than the size of the largest partition.
4. Low degree of multi-programming: In fixed partitioning, the degree of
multiprogramming is fixed and very less because the size of the partition cannot be varied according
to the size of processes.
c. Dynamic Partitioning
i. In this technique, the partition size is not declared initially. It is
declared at the time of process loading.

CodeHelpii.

iii. Advantages over fixed partitioning

1. No internal fragmentation
2. No limit n size of process
3. Better degree of multi-programming
iv. Limitation
1. External fragmentation
 LEC-25: Free Space Management

1. Defragmentation/Compaction
a. Dynamic partitioning suffers from external fragmentation.
b. Compaction to minimize the probability of external fragmentation.
c. All the free partitions are made contiguous, and all the loaded partitions are brought together.
d. By applying this technique, we can store the bigger processes in the memory. The free partitions
are merged which can now be allocated according to the needs of new processes. This technique is
also called defragmentation.
e. The efficiency of the system is decreased in the case of compaction since all the free spaces will be
transferred from several places to a single place.
2. How free space is stored/represented in OS?
CodeHelp

a. Free holes in the memory are represented by a free list (Linked-List data structure).
3. How to satisfy a request of a of n size from a list of free holes?
a. Various algorithms which are implemented by the Operating System in order to find out the holes
in the linked list and allocate them to the processes.
b. First Fit
i. Allocate the first hole that is big enough.
ii. Simple and easy to implement.
iii. Fast/Less time complexity
c. Next Fit
i. Enhancement on First fit but starts s arch always from last allocat d hole.
ii. Same advantages of First Fit.
d. Best Fit
i. Allocate small st hole that is big nough.
ii. Lesser int rnal fragm ntation.
iii. May create many small hol s and cause major external fragmentation.
iv. Slow, as r quir to it rate whole fr holes list.
. Worst Fit
i. Allocate the larg st hole that is big nough.
ii. Sl w, as required to iterate whole free holes list.
iii. Leaves larger h les that may accommodate other processes.
 LEC-26: Paging | Non-Contiguous Memory Allocation

1. The main disadvantage of Dynamic partitioning is External Fragmentation.

a. Can be removed by Compaction, but with overhead.
b. We need more dynamic/flexible/optimal mechanism, to load processes in the partitions.
2. Idea behind Paging
a. If we have only two small non-contiguous free holes in the memory, say 1KB each.
b. If OS wants to allocate RAM to a process of 2KB, in contiguous allocation, it is not possible, as we
must have contiguous memory space available of 2KB. (External Fragmentation)
c. What if we divide the process into 1KB-1KB blocks?
3. Paging
a. Paging is a memory-management scheme that permits the physical address space of a
CodeHelp

process to be non-contiguous.
b. It avoids external fragmentation and the need of compaction.
c. Idea is to divide the physical memory into fixed-sized b ocks ca ed Frames, a ong with divide
logical memory into blocks of same size called Pages. (# Page size = Frame size)
d. Page size is usually determined by the processor archit cture. Traditiona y, ages in a system had
uniform size, such as 4,096 bytes. However, proc ssor d signs oft n a ow two or more, sometimes
simultaneous, page sizes due to its benefits.
e. Page Table
i. A Data structure stores which page is mapp d to which frame.
ii. The page table contains the base addr ss of ach page in the physical memory.
f. Every address generated by CPU (logical address) is divided into two parts: a page number (p) and
a page offset (d). The p is us d as an ind x into a page table to get base address the corresponding
frame in physical m mory.

g. Page table is stored in main memory at the time of process creation and its base address is stored
in process control block (PCB).
h. A page table base register (PTBR) is present in the system that points to the current page table.
hanging page tables requires only this one register, at the time f context-switching.
4. How Paging avoids external fragmentation?
a. Non-contiguous allocation of the pages of the process is allowed in the random free frames of the
physical memory.
5. Why paging is slow and how do we make it fast?
a. There are too many memory references to access the desired location in physical memory.
6. Translation Look-aside buffer (TLB)
a. A Hardware support to speed-up paging process.
b. It’s a hardware cache, high speed memory.
c. TBL has key and value.
d. Page table is stores in main memory & because of this when the memory
references is made the translation is slow.
e. When we are retrieving physical address using page table, after getting frame
address corresponding to the page number, we put an entry of the into the TLB. So that next
time, we can get the values from TLB directly without referencing actual page table. Hence,
make paging process faster.

CodeHelpf.TLBhit,TLBcontainsthemappingfortherqustedlogicaladdress.

g. Address space i ntifi r (ASIDs) is stor d in ach entry of

TLB. ASID uniquely identifies each
process and is us to provi a r ss space protection and allow to TLB to
contain entries for
several ifferent proc ss s. Wh n TLB att mpts to resolve virtual page numbers, it ensures
that
thASID f r the currently executing process matches the ASID associated
with virtual page. If it esn’t match, the attempt is treated as TLB miss.
 LEC-27: Segmentation | Non-Contiguous Memory Allocation

1. An important aspect of memory management that become unavoidable with paging

is separation of user’s view of memory from the actual physical memory.
2. Segmentation is memory management technique that supports the user view of memory.
3. A logical address space is a collection of segments, these segments are
based on user view of logical memory.
4. Each segment has segment number and offset, defining a segment.
<segment-number, offset> {s,d}
5. Process is divided into variable segments based on user view.
6. Paging is closer to the Operating system rather than the User. It divides all the processes into the form of
pages although a process can have some relative parts of functions which need
to be loaded in the same CodeHelppage.
7. Operating system doesn't care about the User's view of the process. It may divide the
same function into different pages and those pages may or may not be loaded at
the same time into the memory. It decreases the efficiency of the system.
8. It is better to have segmentation which divides the proc ss into the s gm nts. Each s
gment contains the same type of functions such as the main function can be includ
d in one s gm nt and the ibrary functions can be included in the other segment.

9.
10. Advantages:
a. No internal fragmentation.
b. One segment has a contiguous allocation, hence efficient working within segment.
c. The size of segment table is generally less than the size of page table.
. It results in a more efficient system because the compiler keeps the same
type of functions in one segment.
11. Disadvantages:
a. External fragmentation.
b. The different size of segment is not good that the time of swapping.
12. Modern System architecture provides both segmentation and paging implemented in
some hybrid approach.
 LEC-28: What is Virtual Memory? || Demand Paging || Page Faults

1. Virtual memory is a technique that allows the execution of processes that are not completely in the
memory. It provides user an illusion of having a very big main memory. This is done by treating a part of
secondary memory as the main memory. (Swap-space)
2. Advantage of this is, programs can be larger than physical memory.
3. It is required that instructions must be in physical memory to be executed. But it limits the size of a
program to the size of physical memory. In fact, in many cases, the entire program is not needed at the
same time. So, we want an ability to execute a program that is only partially in memory would give
many benefits:
CodeHelp
a. A program would no longer be constrained by the amount of physical memory that is
available.
b. Because each user program could take less physical memory, more rograms could be run at
the same time, with a corresponding increase in CPU uti ization and through ut.
c. Running a program that is not entirely in memory wou d benefit both the system and the
user.
4. Programmer is provided very large virtual memory wh n on y a sma r hysical memory is available.
5. Demand Paging is a popular method of virtual m mory manag m nt.
6. In demand paging, the pages of a process which are l ast us d, g t stor d in the s condary memory.
7. A page is copied to the main memory when its demand is mad , or page fault occurs. There are various
page replacement algorithms which are used to d t rmine the pag s which will be replaced.
8. Rather than swapping the entire process into memory, we use Lazy Swapp r. A lazy swapper never
swaps a page into memory unless that page will be needed.
9. We are viewing a proc ss as a s qu nce of pag s, rather than one large contiguous address space, using
the term Swapper is t chnically incorr ct. A swapp r manipulates entire processes, whereas a Pager is
concerned with individual pag s of a proc ss.
10. How Demand Paging works?
a. When a proc ss is to be swapp -in, the pag r guesses which pages will be used.
b. Instead of swapping in a whole proc ss, the pager brings only those pages into memory. This,
it avoi s rea ing into m mory pag s that will not be used anyway.
c. Ab ve way, OS ecreases the swap time and the amount of physical memory needed.
. The valid-invalid bit scheme in the page table is used to distinguish between pages that are
in mem ry and that are on the isk.
i. Valid-invalid bit 1 means, the associated page is both legal and in memory.
ii. Valid-invalid bit 0 means, the page either is not valid (not in the LAS of the process)
r is valid but is currently on the disk.
CodeHelpthepageasthroughithadalwaysbeeninmemory.

f. . If a process never attempts to access some invalid bit pag , the roc ss wi be
executed successfully without even the need pag s pr s nt in the swap space.

i. Check an internal table (in PCB of the process) to determine whether the reference
was valid or an invalid m mory access.
ii. If r f. was invalid proc ss throws xception.
If r f. is valid, pag r will swap-in the page.
iii. We find a fr frame (from fr -frame list)
iv. Sch ule a isk op ration to r ad the desired page into the newly allocated frame.
v. Wh n isk r ad is compl t , we modify the page table that, the page is now in
mem ry.
vi. Restart the instruction that was interrupted by the trap. The process can now access
 i.
j. Pure Demand Paging
i. In extreme case, we can start executing a process with no pages in
memory. When OS sets the instruction pointer to the first instruction
of the process, which is not in the m mory. The proc ss imm diately
faults for the page and page is brought in the memory.
ii. Nev r bring a page into m mory until it is required.
k. We use locality of r f r nce to bring out r asonable performance from demand paging.
11. Advantages of Virtual m mory
a. The egree of multi-programming will be increased.
b. User can run large apps with less real physical memory.
12. Disadvantages f Virtual Mem ry
a. The system can bec me slower as swapping takes time.
b. Thrashing may ccur.
 LEC-29: Page Replacement Algorithms

1. Whenever Page Fault occurs, that is, a process tries to access a page which is not currently present in a
frame and OS must bring the page from swap-space to a frame.
2. OS must do page replacement to accommodate new page into a free frame, but there might be a possibility
the system is working in high utilization and all the frames are busy, in that case OS must replace one of the
pages allocated into some frame with the new page.
3. The page replacement algorithm decides which memory page is to be replaced. Some allocated page is
swapped out from the frame and new page is swapped into the freed frame.
4. Types of Page Replacement Algorithm: (AIM is to have minimum page faults)
a. FIFO
i. Allocate frame to the page as it comes into the memory by replacing the oldest page.
CodeHelp

ii. Easy to implement.

iii. Performance is not always good
1. The page replaced may be an initialization modu e that was used long time ago
(Good replacement candidate)
2. The page may contain a h avily us d variab that was initia ized early and is in
content use. (Will again cause page fau t)
iv. Belady’s anomaly is present.
1. In the case of LRU and optimal page r plac m nt algorithms, it is seen that
the number of page faults will be r duc d if we incr ase the number of
frames. owever, Balady found that, In FIFO page r plac ment algorithm, the
number of page faults will get increased with the increment in number of
frames.
2. This is the strange b havior shown by FIFO algorithm in some of the cases.
b. Optimal page replac m nt
i. Find if a page that is n v r r f r nc d in future. If such a page exists, replace this page
with new page.
If no such page xists, find a page that is referenced farthest in future. Replace this page
with new page.
ii. L west page fault rate among any algorithm.
iii. Difficult to implement as OS requires future knowledge of reference string which is
kind f imp ssible. (Similar to SJF scheduling)
c. Least-recently used (LRU)
i. We can use recent past as an approximation of the near future then we can replace the
page that has n t been used for the longest period.
ii. an be implemented by two ways
1. Counters
a. Associate time field with each page table entry.
b. Replace the page with smallest time value.
2. Stack
a. Keep a stack of page number.
b. Whenever pag is referenced, it is removed from the stack & put on
the top.
c. By this, most recently used is always on the top, & least
recently used is always on the bottom.
d. As entries might be removed from the middle of the
stack, so Doubly linked list can be used.
d. Counting-based page replacement – Keep a counter of the number of
references that have been made to each page. (Reference counting)
i. Least frequently used (LFU)
1. Actively used pages should have a large reference count.
2. Replace page with the smallest count.
ii. Most frequently used (MFU)
1. Based on the argument that the page with the smallest count
was probably just brought in and has yet to be used.
iii. Neither MFU nor LFU replacement is common.
 LEC-30: Thrashing

1. Thrashing
a. If the process doesn’t have the number of frames it needs to support pages in active use, it will
quickly page-fault. At this point, it must replace some page. However, since all its pages are in active use, it must
replace a page that will be needed again right away. Consequently, it quickly faults again, and again, and again,
replacing pages that it must bring back in immediately.
b. This high paging activity is called Thrashing.
c. A system is Thrashing when it spends more time servicing the page faults
than executing processes.

CodeHelp
d. Technique to Handle Thrashing
i. Working s t mo l
1. This mo l is bas on the concept of the Locality Model.
2. The basic principle stat s that if we allocate enough frames to a process to
acc mm ate its current locality, it will only fault whenever it moves to some
new l cality. But if the allocated frames are lesser than the size of the current
l cality, the process is bound to thrash.
ii. Page Fault frequency
1. Thrashing has a high page-fault rate.
2. We want to control the page-fault rate.
3. When it is too high, the process needs more frames. Conversely, if the page-fault
rate is too low, then the process may have to many frames.
4. We establish upper and lower bounds on the desired page fault rate.
5. If pf-rate exceeds the upper limit, allocate the process another frame, if pf-rate
fails falls below the lower limit, remove a frame from the process.
6. By controlling pf-rate, thrashing can be prevented.