Operating Systems: Introduction

1
Operating Systems Introduction

• 1. Historical Development
• 2. The OS as a Resource Manager
• 3. Definitions
• 4. The Process

2
The OS as a Resource Manager

• Resources
• 4 functions
• Keep track of resource
• Enforce policy -
• Allocate
• Reclaim

3
The OS as a Resource Manager

• Memory management
• Keep track
• If multiprogramming
• Allocate
• Reclaim (pi relinquishes it or terminates)

4
The OS as a Resource Manager

• Processor management
• Keep track of
• Scheduling
• Allocate
• Reclaim pi relinquishes P, terminates, exceeds
  TS

5
The OS as a Resource Manager

• Device management
• Keep track of devices, channels, control units
• Efficient allocation policy
• Allocate
• Reclaim Normally automatic termination

6
The OS as a Resource Manager

• Information management
• Keep track Location, use, status ( )
• Who gets use of resources, enforce protection,
  provide accessing routines
• Allocate (Open file)
• Deallocate (Close file)

7
Processor management lower module (P,V), Process
Scheduler
Level 5
Kernel
Level 4
Level 3
Memory management
Level 2
Level 1
Processor management upper module (messages,
create-destroy process)
Bare machine
Device management (I/O traffic control)
Info management (File system)
8
Supervisor process (Job scheduler)
Jobs
Hierarchical OS Structure
Process 1
Process 2
SPOOLing
Process 3
User-created process
I/O process
I/O process working with User process 3
Layer 1
Layer 2
Kernel (Layer 0)
9
Definitions

• Operating System (OS)
• Software governing control of resources
• (
  )
• Interface
• Processor Hardware which interprets and
  executes instructions
• Process (task)
• Job Set of modules reqd to perform some task

10
Definitions

• Multiprogramming
• Privileged instruction
• Instruction available only to
• Executed in supervisor (executive) state as
  opposed to problem (user) state.
• Interrupt
• Mechanism forcing processor to take notice of
  event

11
I.4 The Process

• Process (pi) A program in a state of execution
• Life (Higher level view)
• 1. Run pi
• 2. Wait For event
• 3. Ready

RUN
Wait for I/O completion
pi P
READY
WAIT
I/O complete
12
Life of a Process

• Lower level view

COMPLETE
RUN
WAIT
READY
HOLD
SUBMIT
13
Processor Management and Scheduling

• 1. Introduction
• 2. Basic Concepts
• 3. Processor Scheduling Algorithms

14
Introduction

• Given 2 jobs, A and B.
• Each executes for 1 sec, waits for 1 sec
• Repeat this for 60 cycles 2 min
• 1. Lets run A, then B

A B
2 min
2 min
Elapsed Time 4 min Compute Time 2 min CPU
Utilization 2/4 50
15


Introduction

• 2. Now multi-program A and B

A B
Elapsed Time 2 min
Compute Time 2 min
CPU Utilization 2/2 100

• Compare to 1 A finishes at same time,
• B in half the time.

16

Introduction

• So now we have multiprogramming
• N jobs in memory
• Job i P
• When i waits for I/O, j P
• Overlap CPU and I/O to keep P busy
• Benefits Increased CPU utilization and
  higher throughput.
• Throughput Work done in given period.
• E.g. 10 jobs per hour.

17
Basic Concepts

• Process a program in a state of execution.
• E.g. Batch job, transaction (in T-S system)
• Also called job, program, task, activity.
• Process behavior pi alternate between
  execution(CPU burst) I/O(I/O burst).
• Generally begins and ends with CPU bursts.

18
Basic Concepts

• pi change state as they execute
• READY, RUN, WAIT, HOLD.
• Process Control Block (PCB)
• Process is represented internally by its PCB
• Active representation of passive entity, the
  PGM
• The only tangible part of a process

19

• PCB
• Process id
• Current State
• Priority
• Other CPU Scheduling info
• State info PC (address of NI to be executed)
  register, cc contents
• Memory management info base-bounds
  registers, page tables
• Accounting info amount of CPU time, account
• I/O status info I/O requests pending, I/O
  devices allocated, list of open files
• Pointer to list of all pi in same state
• etc.

20
Basic Concepts

• Scheduling queues
• Ready queue List of processes (PCBs) in ready
  state (awaiting assignment of P)
• Device queue List of pi waiting on this device
• I/O queue pi waiting for I/O (once served, pi
  moves to Ready queue).

. . .
Head Tail
PCB i
PCB j
PCB n
Registers
Registers
Registers
. . .
. . .
. . .
Queue
header
21
Basic Concepts

• Schedulers
• Job (long-term) scheduler
• Selects job from spool queue to enter system,
  loads it into memory
• Processor (CPU, short-term) scheduler
• Selects ready pi and dispatches (assigns P to) it

22

• Difference How often they execute
• Job scheduler
• Infrequently once steady state is reached
• Controls degree of multiprogramming (number of pi
  in memory)
• Processor scheduler
• Must select new pi very frequently (every 10 ms)
• FAST or much of the processor time is spent in
  scheduling !

23
Basic Concepts

• Dispatcher
• Assigns P to pi selected by Processor scheduler
• Loads pis registers
• Changes to user mode
• Jumps to proper address to (re)start it

24
Processor Scheduling Algorithms

• Problem Which pi in Ready queue gets P?
• Performance Criteria (Comparing algorithms)
• Throughput Work done in given period.
• CPU utilization P busy time / Total elapsed
  time
• Want it as busy as possible (40-90).
• Turnaround time Interval between submission to
  completion of job (Batch OS).
• Wait time Time spent by pi in Ready queue.
• Response time Interval from submission until
  response produced (Interactive system).

25
Processor Scheduling Algorithms

• Lets optimize as follows
• Maximize CPU utilization, throughput
• Minimize turnaround time (TT), wait, response
  time.
• Operationally
• Optimize average or max or min
• e.g. Minimize the max response time Minimize
  variance in response time.

26
Common Scheduling Algorithms

• (a) First Come First Served (FCFS)
• (b) Shortest Job First (SJF)
• (c) Priority
• Preemptive vs Non-preemptive
• (d) Round Robin
• (e) Multilevel Queues

27
Scheduling Algorithms

• First Come First Served (FCFS)
• Implementation FIFO queue
• pj enters ready q
• pi dispatched
• Consider Job CPU burst A 24
  B 3 C 3

pi
pk
pj
...
RQ
pk
pj
...
RQ
28

• FCFS
• Suppose jobs arrive in order A, B, C and are
  served FCFS.
• TT For A 24, For B 27, For C 30
• Avg TT (24 27 30) / 3 27
• Now, suppose they arrive in order B, C, A
• Avg TT (3 6 30) / 3 13

A
B
C
24
27
30
0
A
B
C
30
6
3
0
29
Scheduling Algorithms

• Shortest Job First (SJF)
• Associates with job the length of its next CPU
  burst
• Example of priority scheduling (job with shortest
  next CPU burst gets highest priority)
• Consider Job CPU burst A
  6 B 3 C 8 D
  7

30

• SJF
• Avg TT (3 9 16 24) / 4 13
• This algorithm can be proved optimal
• Gives min avg wait time for given set of jobs
• Problem Length of next CPU burst
• Batch jobs Time limit supplied by user
• Processes Cant know but can predict

B3
D7
C8
A6
0
3
9
16
24
31
Scheduling Algorithms

• Priority Scheduling
• Can be assigned who is paying, type of work
• Can be computed
• Time limits, memory req., ratio avg I/O burst to
  CPU burst, number of open files
• Major problem Starvation
• Low priority Ready pi can wait indefinitely for
  CPU
• Eventually, load lightens and job gets run, or
  system crashes and job is lost
• Remedy
• Aging Increase priority of low priority job
  over time

32
Scheduling algorithms

• Round Robin (RR)
• Designed for TS systems
• Given defined TS (10-100 ms).
• CPU scheduler traverses Ready queue, selects pi
  to be dispatched for interval ? TS
• Implementation
• Ready queue FIFO queue
• CPU scheduler picks 1st job in Ready queue
• Sets timer to interrupt after 1 TS
• Dispatches process

33

Scheduling algorithms

• Multilevel Queue
• Good when jobs easily classified into groups
• e.g. foreground (FG - interactive)
  and background (BG - batch)
• In a multilevel scheduling algorithm
• Ready queue partitioned into separate queues
• Each pi assigned permanently to one queue
• Due to memory size, job type,
• Each queue has its own scheduling algorithm
• FG RR, BG FCFS
• Must have scheduling between the queues
• e.g. Fixed priority preemptive (such as FG BG)