INPUT/OUTPUT

CHAPTER #
10
Operating Systems
Chapter 10 - Outline
 Input/Output organization
 Objectives of I/O system
 Unix I/O system
 MS-DOS I/O system
 Device drivers in windows
 Disk structure
 Seek time
 Latency
 Disk scheduling
 FCFS
 SSTF
 Scan scheduling
Input/Output Organization
 Input/output systems is one of the important part of system like
processor and memory
 It is least satisfactory system because of its slowness and lack of
consistency
 Mode of operation of I/O devices is different, it is difficult for
operating system to handle with any generality
 The computer communicates with I/O devices by mean of an I/O bus
system
 Each I/O device has an associated hardware controller attached to this
bus system
 I/O device performs its activity independently of and with the
processor activity
 Most computers use a better technique of direct memory access
(DMA) which offers faster data rates
Characteristics of I/O devices
 Data rate
 Disk: 2Mbytes/sec
 Keyboard: 10-15bytes/sec
 Unit of transfer
 Disk: blocks (e.g. block of 1024bytes)
 Screen: single character
 Operations
 Disk: read, write, seek etc
 Printer: write, move paper
 Error conditions
 Disk: read errors
 Printer: paper out
Structure of I/O System

The I/O system consist of a number of layers separating the user

at one end and the physical devices to other end
 Application Program
 An I/O activity is expressed within the application
 Such high level language instructions are translated into corresponding
system calls which invoke operating system functions
 Input/output Control System (IOCS)
 Part of the operating system which deals with I/O related system calls
 It performs initial processing and validation on the request and routes
it to appropriate handler (device driver) at the next stage
 It is also where the general management of I/O interrupts takes place
Structure of I/O System
 Device Drivers
 It is a software module which manages the communication with, and
control of a specific I/O device
 It converts user’s logical request into specific commands directed to
the device
 Device Controllers
 A device controller is a hardware unit attached to the I/O bus of the
computer
 It provides a hardware interface between the computer and the I/O
device
 Device
 I/O devices are designed to be used in a wide range of different
computer systems
 I/O devices can be categorized into block and character devices
 A virtual device is a simulation of an actual device by the operating
system, e.g. print spooler
Unix I/O System
 Unix kernel treat all I/O devices and files identically
 That is, devices are also represented as special files which appear as
entries within the disk directory system
 All data transfer between a process and a file or device are
handled as a stream of bytes
 The /dev directory holds a list of special files which
correspond to the devices attached to the system
 For example:
 /dev/console system console
 /dev/tty01 user terminal 1
 /dev/rmt0 magnetic tape
 /dev/dsk/os0 hard disk drive 0
 /dev/dsk/f03h 3.1/2 floppy drive
 /dev/lp line printer
MS DOS I/O System
 The early MS-DOS computers had built-in device drivers for
screen, keyboard, magnetic disk, serial port and parallel port
 MS-DOS I/O program services have two components
 DOS services
 Refer to higher level facilities often utilizes the BIOS services
 ROM BIOS
 Refer to ROM-based set of routines which provide low level services for
managing I/O operations on hardware
 IO.SYS file manages the I/O operations of the systems and
corresponds to IOCS
 Whereas, MSDOS.SYS file is the main operating system file
 DOS standard I/O devices names are
 A: first diskette drive
 Lpt1: parallel port 1
 C: hard disk drive
I/O Device Drivers in Windows

 In windows device drivers are implemented as

dynamic link libraries (DLL’s)
 DLL’s contain shareable code for standard drivers
such as display and keyboard drivers
 A new device driver can be loaded without effecting
the earlier one
 A range of optional drivers can be made available
 The plug & play concept is used to make installation
and configuration easy
Disk Structure
 Each disk has a flat circular shape
 Its surfaces are covered with magnetic material
 Data are stored on the surfaces by the read/write
head
 The disk surface is divided into concentric tracks
 Within a track the information is written in blocks
called sectors
 The block is fundamental unit of transfer
 Information is stored in form of tiny magnetic spots
 A head is moved in and out to access different tracks
Moving-head Disk Mechanism
Disk Structure
 Information on the disk is referenced as by a multi part
address, which includes the drive number, the surface and
the track
 The tracks on different surfaces are called a cylinder
 The time to read write a disk block is determined by three
factors;
 i.e. seek time, latency and actual transfer time
 Seek time
 It is the time required to move the head to proper track for read/write
operation
 Rotational Latency
 It is the time to wait for the required block/sector to rotate below the
head
Disk Scheduling

 When a process needs disk I/O, it issues a system call

to the OS;
 If disk is free the request will be served immediately
 But, if disk is busy and serving a request , then the
additional request will be queued
 There are different approaches to serve disk I/O
requests from the queue
 First come first serve (FCFS) scheduling
 Shortest seek time first (SSTF) scheduling
 Scan scheduling
FCFS Scheduling
 The request at the head of the queue will be served first
 FCFS is the simplest form of the disk scheduling
 Its algorithm is easy to implement
 Consider a queue of disk requests;
 98, 183, 37, 122, 14, 124, 65 and 67

 If the reader head is at track 53, it will first move from 53 to 98, then
to 183 and so on
 The total head movement is 640 tracks
 If the requests for the tracks 37 and 14 could be served together, the
total head movement could be decreased substantially
 In this way, the average time to service each request would decrease,
improving disk throughput
FCFS Scheduling (5b)
 Disk track current location= track no 98
 183, 37,53, 122, 14, 124, 65 and 67
 Total head tracks = find answer
14 37 53 65 67 98 122 124 183
 FCFS Scheduling (5b)
 Disk track current location= track no 98
 183, 37,53, 122, 14, 124, 65 and 67
 Find the solution of Total head tracks for the above given
scenario?

Disk track current location= track no 0 and track 400 and track 50
18, 370,153, 12, 14, 112, 300,123,90,87,66,345,234
Find the solution of Total head tracks for the above given scenario?
SSTF Scheduling
 The shortest seek time first disk scheduling algorithm selects the
request with minimum seek time from the current head position
 SSTF results in a total head movements of only 236 tracks
 Comparing with FCFS, SSTF results in substantial improvements
 Assume that we have just two requests in queue, 14 and 186
 If a request near 14 are arrives while serving that request, it will be
served next
 Request for track 186 may wait indefinitely, if more requests near
14 are arriving continuously
SSTF Scheduling (5c)

 98, 183, 37, 122, 14, 124, 65 and 67

14 37 53 65 67 98 122 124 183
Scan Scheduling
 In this algorithm, the head continuously scans the disk from end to
end
 The read-write head starts at one end of the disk, and moves towards
the other end
 Assume that head is moving from track 53 towards 0, the head
movement would service 37, 14 and then moves to 0
 At track 0, the head would reverse and move to other end of the disk
serving requests at 65, 67, 98, 122, 124 and 183
 Any new request in front of head will be served immediately
 A request just behind the head will have to wit until the head moves
to the end of the disk and reverse its direction
 Scan algorithm is also called Elevator algorithm
 This algorithm is little bit better than SSTF
Scan Scheduling (5c,5d,5a,5B)

 98, 183, 37, 122, 14, 124, 65 and 67

14 37 53 65 67 98 122 124 183