CSE 380

Computer Operating Systems

Instructor: Insup Lee

University of Pennsylvania
Fall 2003
Lecture Note 1: Introduction

1
 What is an Operating System?

Operating systems provides an interface between hardware and user

programs, and makes hardware usable

2
Resource Abstraction and Sharing

q It is an extended machine providing abstraction of the

hardware
ß Hides the messy details which must be performed
ß Presents user with a virtual machine, easier to use

q It is a resource manager
ß Time on CPU is shared among multiple users/programs
ß Space in memory and on disks is shared among multiple
users/programs

3
Pentium Architecture

4
 Abstractions in OS

Hardware OS abstraction
q Disks q Files
q Memory q Programs
q Processors q Threads / Processes
q Network q Communication
q Monitor q Windows and GUI
q Keyboard q Input
q Mouse q Locator
5
 Sharing of Memory

Issues

q Allocation schemes
Program 1
q Protection from each other
Free space q Protecting OS code

q Translating logical addresses to physical

Program 3
q Swapping programs
Program 2
q What if physical memory is small: Virtual
memory
OS

6
 Timesharing

P1 OS P2 OS P1 OS P3 OS

q At any point, only one program can run on CPU

q Context switch: changing the program that has CPU

q When to switch (goal: to optimize the CPU usage)

ß When a program terminates

ß When a program has run “long enough”

ß When a program executes a system call or waits for I/O

ß When an external interrupt arrives (e.g. mouse click)

q OS must do all the book-keeping necessary for context switch, with minimum number
of instructions
7
 Challenges in OS
Why can’t Microsoft still get rid of all bugs in Windows ?

q Performance is critical

ß How to reduce the memory and time overhead due to OS

q Synchronization and deadlocks due to shared resources

q Scheduling of multiple programs

ß Fairness, response time, real-time applications

q Memory management

ß Virtual memory, paging, segmentation

q Security and Protection

ß Authorization, authentication, viruses

q Interrupt management and error handling

q Marketability and backward compatibility

8
 How does OS work?

q OS gets control of the CPU repeatedly

q Let’s look at two typical scenarios to get a glimpse of how

things work (we will get a more accurate and detailed
understanding as the course progresses)

q Basic knowledge about computer architecture is essential !

(Read Sec 1.4 to review CSE 240)

9
 Inside a CPU
q State of a running program
ß Registers
ß Program counter (PC)
ß Stack pointer
ß Program status word (PSW)
q Key distinction in PSW: user mode vs kernel (OS) mode
q Key instruction for OS calls: TRAP (switch to kernel mode)
q Many operations (such as accessing I/O devices) are possible only in
the kernel mode

10
 Different types of Memory

q Use of disks unavoidable (permanence and size)

q Access time is significantly slower for disks
11
 Sample Scenario 1

q Consider a statement to read from a file in a user program P

q User program stores parameters such as file-id, memory-address, number-

of-bytes, and system-call number of read, and executes TRAP instruction
to invoke OS

q Hardware saves the state of current program, sets the mode-bit in PSW
register in CPU to 1, and transfers control to a fixed location in OS code

q OS maintains an internal file table that stores relevant information about all
open files

12
 Sample Scenario 1 (continued)

q OS read routine examines the parameters, checks for errors (e.g. file must
be open), consults its file table, and determines the disk address from
where data is to be retrieved

q then it sets up registers to initiate transfer by the disk controller

q While disk controller is transferring data from disk to memory, OS can

suspend current program, and switch to a different program

q When OS routine finishes the job, it stores the status code, and returns
control to the user program P (hardware resets mode-bit)

q Note: Disk controller is accessed only by OS code (this is ensured by

hardware protection)
13
 Sample Scenario 2

q Consider an assignment x:=y in a program P

q Compiler assigns logical addresses, say Add1 and Add2, for program
variables in P’s data space

q When P is loaded in memory, OS assigns a physical base address to store

P and its data

q Compiled code looks like

Load (R, Add1); Store (R, Add2)

q While executing Load instruction the hardware translates the logical

address Add1 to a physical memory location (this is done by Memory
Management Unit MMU)
14
 Sample Scenario 2 (continued)

q However, OS may not keep all of P in memory all the time

q OS maintains an internal table, called page table, that keeps track of which blocks
of P are in memory

q If Add1 is not in memory, MMU generates a page fault, and transfers control to OS

q OS examines the cause, and initiates a disk transfer to load in the relevant block of
P

q OS needs to decide memory allocation for the block to be fetched (page

replacement algorithms)

q While this block is being fetched, P may be suspended using a context switch

15
 Brief History of Operating Systems

q 1940's -- First Computers

q 1950's -- Batch Processing
q 1960's -- Multiprogramming (timesharing)
q 1970's -- Minicomputers & Microprocessors
q 1980's -- Networking, Distributed Systems, Parallel
(multiprocessor) Systems
q 1990's and Beyond -- PCs, WWW, Mobile Systems,
embedded systems

16
 1940's -- First Computers

q Computer dedicated to one user/programmer at a time. Program

loaded manually by programmer, using console switches.
Debugging using console lights.
q Advantages:
ß Interactive (user gets immediate response)
q Disadvantages:
ß Expensive machine idle most of time, because people are slow.
ß Programming & debugging are tedious.
ß Each program must include code to operate peripherals -- error
prone, device dependencies.
q Libraries of subroutines to drive peripherals are example of
typical OS service.

17
 1950's -- Batch Processing

q User/programmer submits a deck of cards that describes a job to be

executed.
q Jobs submitted by various users are sequenced automatically by a
resident monitor.
q Tape drives available for batching of input and spooling of output.
q Advantages:
ß Computer system is kept busier.
q Disadvantages:
ß No longer interactive; longer turnaround time.
ß CPU is still idle for I/O-bound jobs.
q OS issues -- command processor (JCL), protection of resident monitor
from user programs, loading of user programs after monitor.

18
 Typical Batch System

Early batch system

ß bring cards to 1401
ß read cards to tape
ß put tape on 7094 which does computing
ß put tape on 1401 which prints output

19
 1960's -- Multiprogramming
(timesharing)
q The advent of the I/O processor made simultaneous I/O and CPU processing
possible.
q CPU is multiplexed (shared) among a number of jobs -- while one job waiting for
I/O, another can use CPU.
q Advantages:
ß Interactiveness is restored.
ß CPU is kept busy.
q Disadvantages:
ß Hardware and O.S. required become significantly more complex.
q Timesharing - switch CPU among jobs for pre-defined time interval
q Most O.S. issues arise from trying to support multiprogramming -- CPU scheduling,
deadlock, protection, memory management, virtual memory, etc.
q CTSS (Compatible Time Sharing System), Multics

20
 1970's - Minicomputers &
Microprocessors

q Trend towards many small to mid-range personal computers,

rather than a single mainframe.
q Early minicomputers and microprocessors were small, so there
was some regression to earlier OS ideas.
ß e.g. DOS on PC is still essentially a batch system similar to those
used in 1960, with some modern OS ideas thrown in (e.g.,
hierarchical file system).
q This trend changing rapidly because of powerful new
microprocessors.
q Also, the user interface (GUI) became more important.
q UNIX, DOS
21
 1980's - Networking

q Powerful workstations (e.g., PDP, VAX, Sunstations, etc.)

q Local area networks (e.g., Ethernet, Token ring) and long-distance
network (Arpanet)
q Networks organized with clients and servers
q Decentralization of computing requires more communication (e.g.,
resource sharing)
q O.S. issues -- network communication protocols, data encryption,
security, reliability, consistency of distributed data
q Real-Time Systems – timing constraints, deadlines, QoS (quality of
service)

22
 1990's and Beyond

q Parallel Computing (tera-flops)

q Powerful PCs, Multimedia computers

q High-speed, long-distance communication links to send large amounts of

data, including graphical, audio and video

q World Wide Web

q Electronic notebooks and PDAs using wireless communication

technologies

q Embedded computers: medical devices, cars, smartcards

q O.S. issues -- Large heterogeneous systems, mobile computing, utilization

of power, security, etc.
23
 Operating System Structure

q Monolithic Systems
q Layered Systems
q Virtual Machines
q Client-Server Model

24
 Operating System Structure (1)

Simple structuring model for a monolithic system

25
Operating System Structure (2)

Structure of the THE operating system

26
Operating System Structure (3)

Structure of VM/370 with CMS

27
Operating System Structure (4)

The client-server model

28
 Operating System Structure (5)

The client-server model in a distributed system

29