CSCI2402 Operating Systems and Computer Networks

1
CSCI2402 Operating Systems and Computer Networks

• Logical I/O

2
Logical I/O

• Managing the transfer of data to/from I/O devices
• Issues
• Range of different devices
• Complexity of operation
• Providing a simple and consistent interface for
  users
• Speed of transfer
• Speed disparity between CPU and mechanical
  devices
• Efficiency

3
Logical I/O

• Objectives
• Create a system which provides
• Device independence
• Simple, uniform file/device naming and access
• Efficiency max. CPU time running processes
• operate I/O device at max poss. speed

4
The Disk Surface

• The surface of a disk is logically divided into
  concentric circular tracks
• 1,000s are now common

• The surface is also divided into segments

• a track through a particular segment is a sector
• note often the term sector is confusingly used
  to refer to both segments and sectors!
• on typical PC disks, each sectorholds 512 bytes

5
Layers of a typical I/O system
user processes

• Device Independent
• Naming files
• Naming devices
• Device protection
• Buffering

• Device Dependent
• Maps logical requests into hardware operations.
• Error reporting
• Performs the op. and uses interrupts to signal
  successful read or write ops.

Device Driver
6
Logical I/O

• The User View (User Process)
• User sees a logical device and a standard
  interface
• Program I/O instructions translated into system
  calls to invoke OS functions
• Some I/O instructions provided by libraries
  linked to user programs
• Printing tasks often carried out via Spooling
• Spooling Directory Daemons

7
Logical I/O

• 2. The Device Independent Layer (I/O Control
  System)
• Provides a uniform logical interface to users
  which is device-independent
• Provides a logical file/device naming system to
  the user
• Implements any file/device protection strategy
• Provides info. about file sizes etc. but hides
  precise physical details
• Applies any buffering techniques to increase
  efficiency

8
Buffering

• Buffering
• Without Buffering

OS
User Process
Is this efficient ?
9
Buffering

• b) With Single Buffering

OS
User Process
Buffer
10
Buffering

• c) With Double Buffering

OS
User Process
Buffer B
Buffer A
11
Device Driver

• 3. The Device Driver (device-dependent)
• When it receives a logical request for e.g. a
  file on disk, it must
• Figure out where the file is on the disk
• Check to see if the drive motor is running
• Determine if the arm is positioned over the
  correct sector
• etc. etc. etc.
• i.e. Decide which device hardware operations are
  necessary and in which sequence.

12
Device Driver

• then give appropriate commands to hardware.
• Then if OK (and relevant!)
• Pass data back to I/O Control System
• Or if not OK
• Try and correct (e.g. retry, ignore etc.) or pass
  back an error report)

13
Disk Scheduling Algorithms
14
Disk Scheduling Concepts

• The operating system is generally responsible for
  using the hardware efficiently
• Whenever a process needs I/O to or from a disk,
  it issues a system call to the operating system
• the system call request specifies
• whether the operation is input or output
• what the disk address for the operation is
• what the memory address of the transfer is
• the number of bytes to be transferred
• The OS maps block numbers into head and sector
  physical addresses
• If the drive (and its controller) are free, the
  request can be serviced immediately
• if not, the request will have to wait in a
  pending queue

15
Performance Parameters

• To the operating system a disk drive can be
  characterised in terms of performance parameters
• the transfer rate
• the rate at which data flow between the drive and
  computer
• measured (now) in Mb per second
• e.g. 33 Mb/s or 66 Mb/s typical for modern PC
  drives
• the seek time
• time to move the disk arm to the required sector
• e.g. 9 ms average seek time

16
First-Come, First-Served

• As in process scheduling, the simplest algorithm
  is to keep a queue of all pending requests and
  service them in the order they joined the queue
• this is (again) called first-come, first-served
  (FCFS)

17
FCFS Example

• Consider a queue, for a disk with 200 sectors,
  with requests for I/O to blocks on sectors
• 98, 183, 37, 122, 14, 124, 65, 67
• the disk heads are initially positioned at
  cylinder 53

• a total head movement of 640 sectors has occurred

18
Shortest-Seek(-Time)-First

• The shortest-seek-first (or shortest-seek-time-fir
  st) algorithm improves the FCFS situation by
  trying to reduce the head movement distance by
  sorting the pending I/O requests according to
  closeness
• on a hard disk the seek time is proportional to
  distance
• For each block, the pending queue is examined
• the nearest block in the queue is serviced next
• Note that this algorithm is similar to SJF
• it too may result in starvation
• if near blocks join the queue then blocks that
  are far away will never get serviced
  (particularly if the disk is busy)

19
SS(-T-)F Example

• The same queue as for FCFS
• 98, 183, 37, 122, 14, 124, 65, 67
• the disk heads are initially positioned at sector
  53

• the total head movement has been reduced to 236

20
Elevator Algorithm

• The disk arm starts moving in one direction
• it continues in the current direction, servicing
  all pending requests as it passes over them
• when it gets to the last pending request in the
  direction it is going, it reverses direction and
  services all the pending requests
• this is the same as the algorithm used in most
  elevators (lifts)
• This improves FCFS without suffering starvation
• middle blocks are favoured if the disk is heavily
  used
• if the direction is up and the arm is at sector 1
  when a sectior 0 request arrives, the block must
  wait until all blocks are serviced upwards, and
  again downwards, before its turn

21
Elevator Example

• The same I/O request queue again
• 98, 183, 37, 122, 14, 124, 65, 67 heads at 53
• the initial direction is DOWN

• the total head movement has been reduced to 208

22
Selecting an Algorithm

• Which is the best algorithm to use?
• SSF is common in practice as it usually performs
  slightly better than the elevator algorithm (on
  lightly loaded systems)
• the elevator algorithm is a good candidate for
  general purpose as it performs well without
  starvation
• The choice of disk arm scheduling algorithm may
  well be influenced by the file allocation method
• contiguous allocation ? many short head movements

23
Logical I/O

• Summary

Make an I.O call
Naming, Protection, Blocking, Buffering
Map logical requests into hardware ops
Wake up driver when I/O completed
Perform I/O operation