Operating System18

1
Operating System(18)

• Minyi Guo
• Dept. of Computer Science and Engineering,
• Shanghai Jiaotong University

2
Input/Output

• Principles of I/O hardware
• Principles of I/O Software
• I/O software layers
• Disks/Clocks/Char-Oriented Terminals
• GUI/Network Terminals
• Power management

3
Disks

• Come in a variety of types
• Magnetic disks
• hard disk and floppy disks
• Optical disks
• CD-ROM, CD-RW, DVD

4
Magnetic Disks

• Integral part of a computer system
• Some contains a microcontroller
• IDE Disks
• Some have little electronics and just deliver a
  simple serial bit of stream
• Let the device controller do all the work

5
Magnetic Disks

• Some disks offer parallel processing
• Doing seek on two ore more drives
• Overlapped seeks
• Read/write while seeking on another drive
• Floppy disk cannot offer parallelism

6
Disk Drives

• Disks are usually organized into drives
• A drive may have up to 16 disks
• With each disk having its own read/write head
• Read data from one disk at a time

7
Tracks and Cylinders

• The surface of a disk is divided in tracks
• A track is one cycle
• All tracks of a disk drive forms a cylinder
• A disk drive can have up to 16 disks
• Each track is further divided into sectors
• A sector is the basic unit of I/O

8
Disk Drives
9
Sectors

• Numbered from 0 to some maximum
• Sectors can either be continuous or interleave
• Continuous logically adjacent physical
  adjacent
• Interleave logically adjacent ! physical
  adjacent

10
Sectors

• Gaps are maintained between sectors
• This allows continuous read

11
RAID

• Redundant Array of Inexpensive Disks
• Redundant Array of Independent Disks
• Goal improve performance and reliability
• Idea use multiple disks (drives) as 1 unit
• Key technique stripe

12
RAID

• Depending on how stripe is implemented
• RAID can be classified into 6 levels
• RAID 0 stripe in blocks
• Performance improved, better for large request
• Bad for small request, worse reliability, no
  redundancy

13
RAID

• RAID 1 full redundancy, may or may not stripe
• Write twice, performance of read is up to twice
  as good
• RAID 2 stripe in bits, many dedicated ECCs

14
RAID

• RAID 3 stripe in words, less ECC
• RAID 4 stripe in blocks with dedicated ECC
• RAID 5 stripe in blocks with round-robin ECC
• RADI 2 and 3 require disk synchronization

15
Raid levels 0 through 2
16
Raid levels 3 through 5
17
Disk Read/Write Process

• First move head to the correct cylinder
• seek
• Wait for the sector to pass by the head
• rotation
• Actually read/write data
• transfer

18
Seek Time

• Seek time
• Time for I/O head to reach desired track.
• Track-to-track time
• Time to move from track to an adjacent track
• Average Seek time
• Average time to reach a track for random access

19
Delay time

• Rotational Delay or Latency
• Time for I/O head to reach the wanted data as the
  disk rotate
• One half of a rotation on average
• At 7200 rpm, this is 8.3/2 4.2ms

20
Transfer time

• Time for data to move under the I/O head
• At 7200 rpm
• Number of sectors read / Number of sectors per
  track 8.3ms

21
Disk Spec Example

• 16.8 GB disk with 10 platters
• 1.68GB/platter
• 13,085 tracks/platter
• 256 sectors/track
• 512 bytes/sector

22
Disk Spec Example

• Track-to-track seek time 2.2 ms
• Average seek time 9.5ms
• Interleaving factor of 3
• 5400RPM

23
Disk Access Cost Example

• Read a 1MB file divided into 2048 records of 512
  bytes (1 sector) each
• Assume all records are on 8 contiguous tracks
• First track 9.5 11.1/2 3 x 11.1 48.4 ms
• Remaining 7 tracks 2.2 11.1/2 3 x 11.1
  41.1ms
• Total 48.4 7 41.1 335.7ms

24
How Much to Read

• Read time for one track
• 9.5 11.1/2 3 x 11.1 48.4ms
• Read time for one sector
• 9.5 11.1/2 (1/256)11.1 15.1ms

25
How Much to Read

• Read time for one byte
• 9.5 11.1/2 15.05 ms
• Nearly all disk drives read/write one sector at
  every I/O access
• Also referred to as a page

26
Physical Geometry

• Different tracks have different length
• Outer tracks are longer than inner tracks
• thus tracks may have differing of sectors
• But this cause difficult to OS
• Solution is to provide a virtual geometry

27
Physical Geometry of a Disk
Physical geometry of a disk with two zones
A possible virtual geometry for this disk
28
IBM 75GXP Specification

• Size 461,
• Capacity 76GB
• RPM 7200
• Platters 5
• Sectors per track 370-792
• Data rate 174-374Mbps

29
IBM 75GXP Virtual Specification

• Number of tracks 34327
• Sectors per track 512
• Average seek time 9.2ms
• Track density 28350/in

30
Disk Formatting

• Disk must be formatted to be useful
• Format involves writing basic information into
  the disk
• Sector information
• Boot information, Partition tables
• Reserved spare sectors
• Formatted capacity is 20 lower than the
  unformatted capacity

31
Disk Formatting

• A disk sector
• The preamble starts with a certain bit pattern
  that allows the hardware to recognize the start
  of a sector, also contains the cylinder and
  sector
• ECC contains error correction code for data
  recovery

32
Cylinder Skew

• To allow continuous read of data across
  cylinders, the numbering of sectors in
  neighboring cylinders are gapped
• To allow continuous read of data in the same
  track, sector may also be skewed

33
Cylinder Skew
34
Sector Interleaving
No interleaving Single
interleaving Double interleaving
35
Disk Arm Scheduling

• Time required to read or write a disk block
  determined by 3 factors
• Seek time
• Rotational delay
• Actual transfer time
• Seek time dominates

36
Disk Arm Scheduling

• Scheduling goal is to minimize total seek time
• Scheduling Algorithm
• Shortest Seek First (SSF)
• Elevator Scheduling

37
Disk Arm Scheduling
Initial position
Pending requests

• Shortest Seek First scheduling algorithm

38
Disk Arm Scheduling
Elevator algorithm for scheduling requests
39
Error Handling

• Disk defect is inevitable
• Defect in uniform substrate
• Defect in fine oxide coating
• Motion defect increase as time goes
• by Read/write head moves into the wrong place
• Environment pollution
• Dust and speck cause temporary RW errors

40
Error Handling

• Disk defect cause bad blocks
• If the defect is small, let the ECC correct
• Otherwise, let entire sector go bad and
• deal with bad sector in controller or OS

41
Error Handling

• 2 approaches to handle bad sectors
• Deal with them in controller
• Disk tested before shipping
• bad sector marked and substituted with spare
• Deal with them in operating system
• OS check for bad sectors and record them

42
Error Handling
(a) Found a bad sector in a disk track (b) Find a
replace sector (c) Shift all sectors to bypass
the bad sector