1

Chapter 5 : Input & Chapter 5 : Input & OutputOutput

I/O hardware (classification, I/O hardware (classification, device drivers)device drivers)

I/O techniques (programmed, I/O techniques (programmed, interrupt driven, DMA)interrupt driven, DMA)

Structuring I/O softwareStructuring I/O software Disks (performance, arm Disks (performance, arm

scheduling, common disk errors)scheduling, common disk errors) RAID configurationsRAID configurations

2

I/O HardwareI/O Hardware

Classification of I/O devicesClassification of I/O devices Device controllersDevice controllers

3

Classification of I/O Classification of I/O DevicesDevices

Block devicesBlock devices Information is stored in fixed size blocksInformation is stored in fixed size blocks Block sizes range from Block sizes range from 512512--3276832768 bytes bytes I/O is done by reading/writing blocksI/O is done by reading/writing blocks Hard disks, floppies, CD ROMS, tapes are Hard disks, floppies, CD ROMS, tapes are

in this categoryin this category Character devicesCharacter devices

I/O is done as characters (ie., no blocking)I/O is done as characters (ie., no blocking) Terminals, printers, mouse, joysticks are in Terminals, printers, mouse, joysticks are in

this categorythis category

4

Some typical device, network, and data base ratesSome typical device, network, and data base rates

5

Device ControllersDevice Controllers

A controller is an electronic card (PC’s) or a unit A controller is an electronic card (PC’s) or a unit (mainframes) which performs blocking(mainframes) which performs blocking (from serial (from serial bit stream)bit stream), analog signal generation (to move the , analog signal generation (to move the disk arm, to drive CRT tubes in screens), execution disk arm, to drive CRT tubes in screens), execution of I/O commandsof I/O commands

System bus

CPU Memory Controller Controller Controller

Disk

Motherboard

6

I/O TechniquesI/O Techniques

Programmed I/OProgrammed I/O Interrupt-driven I/OInterrupt-driven I/O Direct memory access (DMA)Direct memory access (DMA)

7

Programmed I/OProgrammed I/O

The processor issues an I/O The processor issues an I/O command on behalf of a process to command on behalf of a process to an I/O modulean I/O module

The process The process busy-waitsbusy-waits for the for the operation to be completed before operation to be completed before proceedingproceeding

8

Interrupt-driven I/OInterrupt-driven I/O Processor issues an I/O command on Processor issues an I/O command on

behalf of a processbehalf of a process Process is suspended and the I/O startsProcess is suspended and the I/O starts Processor may execute another processProcessor may execute another process Controller reads a block from the drive Controller reads a block from the drive

serially, bit by bit into controller’s serially, bit by bit into controller’s internal bufferinternal buffer. A checksum is computed . A checksum is computed to verify no reading errorsto verify no reading errors

When I/O is finished, the processor is When I/O is finished, the processor is interrupted to notify that the I/O is overinterrupted to notify that the I/O is over

OS reads controller’s buffer a byte (or OS reads controller’s buffer a byte (or word) at a time into a memory buffer.word) at a time into a memory buffer.

9

Direct Memory Access Direct Memory Access (DMA)(DMA)

A DMA module controls the exchange of data A DMA module controls the exchange of data between main memory and an I/O devicebetween main memory and an I/O device

The processor sends a request for the transfer The processor sends a request for the transfer of a block of data to the DMA module (block of a block of data to the DMA module (block address, memory address and number of address, memory address and number of bytes to transfer) and continues with other bytes to transfer) and continues with other workwork

DMA module interrupts the processor when DMA module interrupts the processor when the entire block has been transferredthe entire block has been transferred

When OS takes over, it does not have to copy When OS takes over, it does not have to copy the disk block to memory; it is already there.the disk block to memory; it is already there.

10

DMA (Cont.)DMA (Cont.) DMA unit is capable of transferring data DMA unit is capable of transferring data

straight from memory to the I/O devicestraight from memory to the I/O device Cycle StealingCycle Stealing: DMA unit makes the CPU : DMA unit makes the CPU

unable to use the bus until the DMA unit unable to use the bus until the DMA unit has finishedhas finished

Instruction execution cycle is suspended, Instruction execution cycle is suspended, NOTNOT interrupted interrupted

DMA has less number of interrupts (usually DMA has less number of interrupts (usually one when I/O is finished to acknowledge). one when I/O is finished to acknowledge). Interrupt driven I/O may need one Interrupt driven I/O may need one interrupt for each character if device has interrupt for each character if device has no internal buffers.no internal buffers.

11

Direct Memory Access Direct Memory Access (DMA)(DMA)

Operation of a DMA transferOperation of a DMA transfer

12

Hardware

Interrupt handlers

Device drivers

Device-independent software

User space software

I/O request I/O reply

I/O system calls (library)

Naming, protection, blocking,

buffering, allocation

Setup device registers; check status

Wakeup driver when I/O completed

Perform I/O operation

Structuring I/O SoftwareStructuring I/O Software

13

User-Space I/O SoftwareUser-Space I/O Software

Library of I/O procedures (ie., Library of I/O procedures (ie., system calls) such as system calls) such as

bytes-read = read (file_descriptor, bytes-read = read (file_descriptor, buffer, buffer, bytes to be read) bytes to be read)

Spooling provides virtual I/O devicesSpooling provides virtual I/O devices

14

Device-Independent I/O Device-Independent I/O SoftwareSoftware

Uniform interface for device Uniform interface for device drivers (ie., different devices)drivers (ie., different devices)

Device namingDevice naming Mapping of symbolic device names Mapping of symbolic device names

to proper device driversto proper device drivers Device protectionDevice protection

In a multi-user system you can not In a multi-user system you can not let all users access all I/O deviceslet all users access all I/O devices

15

Device-Independent I/O Device-Independent I/O Software (Cont.)Software (Cont.)

Provide device independent block Provide device independent block sizesize Physical block sizes for different devices Physical block sizes for different devices

may differ, so we have to provide the may differ, so we have to provide the same logical block sizessame logical block sizes

BufferingBuffering Storage allocation on block devices Storage allocation on block devices

such as diskssuch as disks Allocating and releasing dedicated Allocating and releasing dedicated

devices such as tapesdevices such as tapes

16

Device-Independent I/O Device-Independent I/O Software (Cont.)Software (Cont.)

Error reportingError reporting When a bad block is encountered, the When a bad block is encountered, the

driver repeats the I/O request several driver repeats the I/O request several times and issues an error message if times and issues an error message if data can not be recovereddata can not be recovered

17

Device DriversDevice Drivers

One driver per device or device classOne driver per device or device class Device driverDevice driver

Issues I/O commandsIssues I/O commands Checks the status of I/O device (eg. Checks the status of I/O device (eg.

Floopy drive motor)Floopy drive motor) Queues I/O requestsQueues I/O requests

18

Device DriversDevice Drivers (Cont.) (Cont.)

(a) Without a standard driver interface(a) Without a standard driver interface

(b) With a standard driver interface(b) With a standard driver interface

19

Interrupt HandlersInterrupt Handlers

When an I/O is issued, the process is When an I/O is issued, the process is suspended until I/O is finishedsuspended until I/O is finished

When the I/O is finished, the hardware When the I/O is finished, the hardware causes an interrupt and the execution causes an interrupt and the execution is directed to a special routine is directed to a special routine (interrupt handler)(interrupt handler)

Interrupt handler notifies the device Interrupt handler notifies the device driver which in turn passes this to the driver which in turn passes this to the upper layersupper layers

20

DisksDisks

Heads

Cylinder

Sector

Track

21

DisksDisks (Cont.) (Cont.)

Disk parameters for the original IBM PC floppy Disk parameters for the original IBM PC floppy disk and a Western Digital WD 18300 hard diskdisk and a Western Digital WD 18300 hard disk

22

Disk Performance Disk Performance ParametersParameters

Seek time: Time to move disk arm to Seek time: Time to move disk arm to the required trackthe required track

Rotational delay (rotational latency): Rotational delay (rotational latency): Wait for the correct block to be Wait for the correct block to be under the headunder the head

Seek Rotational Delay

Data Transfer

Access time

23

Approximate Formulas Approximate Formulas for Disk Performance for Disk Performance

Parameters Parameters Seek time (TSeek time (Tss) = ) = m * n + sm * n + swhere where mm = a constant depending on the = a constant depending on the disk drivedisk drive

n n = number of tracks traversed = number of tracks traversed s s = startup time = startup time

Rotational delay (TRotational delay (Trr) = 1 / (2*) = 1 / (2*rr))wherewhere r r is the rotation speed in revolutions is the rotation speed in revolutions per secondper second

Transfer time (TTransfer time (Ttt) = ) = bb / ( / (r*Nr*N))where where bb = number of bytes to be = number of bytes to be transferredtransferred NN = number of bytes on a track = number of bytes on a track

Average access time (TAverage access time (Ta a ) = T) = Ts s + T + Tr r + + TTtt

24

Example (From Example (From “Stallings” Book) “Stallings” Book)

Read a file of 256 sectors (or 8 Read a file of 256 sectors (or 8 tracks) from a disk drive with the tracks) from a disk drive with the following characteristicsfollowing characteristics Average seek time = 20 msecAverage seek time = 20 msec Transfer rate = 1 Mb/sTransfer rate = 1 Mb/s Bytes per sector = 512Bytes per sector = 512 32 sectors per track32 sectors per track Disk rotates at 3600 rpm Disk rotates at 3600 rpm

Consider two cases:Consider two cases: Contiguously StoredContiguously Stored Randomly StoredRandomly Stored

25

File is Stored File is Stored ContiguouslyContiguously

Time to read one track isTime to read one track is

seek + latency + data transfer (1 track - seek + latency + data transfer (1 track - one revolution) =one revolution) =

20 msec + 8.3 msec + 16.7 msec = 45 20 msec + 8.3 msec + 16.7 msec = 45 msecmsec

No seek time for the other 7 tracksNo seek time for the other 7 tracks All tracks = first track + other 7 tracksAll tracks = first track + other 7 tracks

45 msec + 7 * 25 (8.3+16.7) = 220 45 msec + 7 * 25 (8.3+16.7) = 220 msecmsec

26

Randomly StoredRandomly Stored

Time to read one sector (randomly) isTime to read one sector (randomly) is

seek + latency + data transfer (1 seek + latency + data transfer (1 sector) =sector) =

20 msec + 8.3 msec + 0.5 msec = 20 msec + 8.3 msec + 0.5 msec = 28.8 msec28.8 msec

Time to read 256 sectors = 256 * 28.8 Time to read 256 sectors = 256 * 28.8 = 7.37 seconds= 7.37 seconds

27

Disk Scheduling PoliciesDisk Scheduling Policies The order in which sectors are The order in which sectors are

read from the disk has a read from the disk has a tremendous effect on I/O tremendous effect on I/O performance)performance)

Scheduling AlgorithmsScheduling Algorithms FIFOFIFO SSF (Shortest seek first)SSF (Shortest seek first) SCAN (Elevator algorithm) SCAN (Elevator algorithm) C-SCAN (One-way elevator)C-SCAN (One-way elevator) FSCANFSCAN

28

First in, First out (FIFO)First in, First out (FIFO) Disk driver accepts request one at a Disk driver accepts request one at a

time and carries them in that ordertime and carries them in that order No starvationNo starvation Example: Requests for 1, 36, 16, 34, 9, Example: Requests for 1, 36, 16, 34, 9,

12 when positioned on cylinder 11 12 when positioned on cylinder 11 (mean movement = 18.5 cylinders)(mean movement = 18.5 cylinders)

111 12 16 3634

0 10 20 30 40

9

29

Shortest Seek First (SSF)Shortest Seek First (SSF)

Request which requires shortest seek is chosenRequest which requires shortest seek is chosen Possibility of starvation (if requests are clustered)Possibility of starvation (if requests are clustered) Same example : mean movement = 10.2 cylindersSame example : mean movement = 10.2 cylinders

111 12 16 3634

0 10 20 30 40

9

30

SCAN (Elevator SCAN (Elevator Algorithm)Algorithm)

Disk arm moves in one direction, performing Disk arm moves in one direction, performing all requests until no more are needed in that all requests until no more are needed in that direction, then turns around and comes backdirection, then turns around and comes back

Same example : mean movement = 10.0 Same example : mean movement = 10.0 cylinderscylinders

111 12 16 3634

0 10 20 30 40

9

31

SCAN (Cont.)SCAN (Cont.)

Favours Favours Tracks nearest to both Tracks nearest to both innermost and outermost innermost and outermost cylinderscylinders

Latest-arriving requestsLatest-arriving requests

32

C-SCAN (One-way C-SCAN (One-way Elevator)Elevator)

Modification of SCAN where Modification of SCAN where scanning direction is one way scanning direction is one way onlyonly

Once arm reaches the end it Once arm reaches the end it moves back to the startmoves back to the start

33

FSCANFSCAN SSTF, SCAN and C-SCAN SSTF, SCAN and C-SCAN

may suffer from "arm may suffer from "arm stickiness” (starvation for stickiness” (starvation for some requests)some requests)

If multiple new requests If multiple new requests keep arriving for the same keep arriving for the same track the arm gets "stuck"track the arm gets "stuck"

The solution is to maintain The solution is to maintain multiple queuesmultiple queues

34

FSCAN (Cont.)FSCAN (Cont.) Two queues, one being used for Two queues, one being used for

the scan and the other for new the scan and the other for new requests during the scanrequests during the scan

When a scan begins, all new When a scan begins, all new requests are in one of the requests are in one of the queues, with the other being queues, with the other being emptyempty

During the scan, all new During the scan, all new requests are put into the queue requests are put into the queue that was initially emptythat was initially empty

Thus, service of new requests is Thus, service of new requests is deferred until all the old deferred until all the old requests have been processedrequests have been processed

35

Common Disk ErrorsCommon Disk Errors Programming error (e.g., request Programming error (e.g., request

for nonexistant sector)for nonexistant sector) This type of error should not This type of error should not

occur if programming (software occur if programming (software development) is done carefullydevelopment) is done carefully

If such an error is encountered, If such an error is encountered, probably the only thing to do is probably the only thing to do is to terminate the request and to terminate the request and notify the user or programmernotify the user or programmer

36

Common Disk Errors Common Disk Errors (Cont.)(Cont.)

Transient checksum error (e.g., Transient checksum error (e.g., usually caused by dust on the head. usually caused by dust on the head. Mostly for floppies)Mostly for floppies) The read or write operation is The read or write operation is

repeated for a couple of timesrepeated for a couple of times If the operation is not successful If the operation is not successful

the block is marked as bad (Bad the block is marked as bad (Bad CRC)CRC)

Re-formating may cure the errorRe-formating may cure the error

37

Common Disk Errors Common Disk Errors (Cont.)(Cont.)

Permanent checksum error (e.g., Permanent checksum error (e.g., disk block physically damaged)disk block physically damaged) Bad blocks are marked so that Bad blocks are marked so that

device drivers do not access themdevice drivers do not access them Controller error (e.g., controller Controller error (e.g., controller

refuses to accept commands)refuses to accept commands)

38

Common Disk Errors Common Disk Errors (Cont.)(Cont.)

Seek error (e.g., the arm is directed to Seek error (e.g., the arm is directed to cylinder 6 but it goes to 7) cylinder 6 but it goes to 7) The disk arm is positioned on The disk arm is positioned on

cylinders by pulses (one pulse per cylinders by pulses (one pulse per cylinder). When the arm reaches its cylinder). When the arm reaches its destination the cylinder number is destination the cylinder number is checked (written when the drive checked (written when the drive was formatted). If the arm is in a was formatted). If the arm is in a wrong position then a seek error wrong position then a seek error occursoccurs

39

Common Disk Errors Common Disk Errors (Cont.)(Cont.)

Some controllers can correct the Some controllers can correct the seek error by issuing a seek error by issuing a RECALIBRATE commandRECALIBRATE command

This command moves the arm as This command moves the arm as far out as it will go to reset the far out as it will go to reset the arm on cylinder 0. If this does not arm on cylinder 0. If this does not solve the problem then the drive solve the problem then the drive has to be repaired (replaced with has to be repaired (replaced with a new one?)a new one?)

40

RAID (Redundant Array RAID (Redundant Array of Inexpensive Disks)of Inexpensive Disks)

Security (fault tolerance)Security (fault tolerance) PerformancePerformance 0-5 levels are defined0-5 levels are defined

41

RAID 0 LevelRAID 0 Level A file is written (distributed) over several A file is written (distributed) over several

disksdisks This permits multiple reads and writesThis permits multiple reads and writes Consequently speed is improvedConsequently speed is improved But no error correctionBut no error correction

42

RAID 1 (Mirroring)RAID 1 (Mirroring) A file is written on at least two drivesA file is written on at least two drives The other drive becomes a mirror The other drive becomes a mirror

image of the first driveimage of the first drive

43

RAID 1 (Mirroring)RAID 1 (Mirroring)

Reading is improved because of two Reading is improved because of two pathspaths

Writing is slower as the same data Writing is slower as the same data has to be written twicehas to be written twice

Fault tolerance is improved as the Fault tolerance is improved as the failure of two disks at the same time failure of two disks at the same time is lowis low

44

RAID RAID 22 A file is distributed on several disks as in RA file is distributed on several disks as in Raidaid 0 0, ,

but Raid level 2 works on words or bytes basisbut Raid level 2 works on words or bytes basis AAdditional drivedditional drives s contains the parity information contains the parity information

which may be used to reconstruct the file if a which may be used to reconstruct the file if a drive fails (drive fails (See “Hamming Codes” for error See “Hamming Codes” for error correctioncorrection)), ,

45

RAID RAID 22

Raid level 2 requires synchronized Raid level 2 requires synchronized drivesdrives

Reading is Reading is very very fast as all drives can fast as all drives can transfer data (portions of the file)transfer data (portions of the file). In . In one sector reading time n (number one sector reading time n (number of drives) sectors are read inof drives) sectors are read in

Writing is slower Writing is slower because of because of parity parity information (to write parity all information (to write parity all requests must access this drive)requests must access this drive)

46

RAID RAID 33 Raid 3 is a simplifed version of Raid 2Raid 3 is a simplifed version of Raid 2 It uses an additional drive for a parity bit It uses an additional drive for a parity bit

(word/byte)(word/byte) Drives must be synchronizedDrives must be synchronized Provides 1 bit (word/byte) error correctionProvides 1 bit (word/byte) error correction

47

RAID RAID 44 A file is distributed A file is distributed as strips as strips on several on several

disks as in RAID 0disks as in RAID 0 Another drive holds the parity stripAnother drive holds the parity strip Reading is fast as all drives can transfer Reading is fast as all drives can transfer

data (portions of the file) independentlydata (portions of the file) independently Parity drive is heavily usedParity drive is heavily used

48

RAID RAID 55

Raid level 5 is Raid level 4 with Raid level 5 is Raid level 4 with parity information distributed on parity information distributed on all drivesall drives

49

RAID 5RAID 5 Similar to RAID Similar to RAID 44 but parity is but parity is

distributed to all disksdistributed to all disks Fast read and writesFast read and writes Suitable for transaction oriented Suitable for transaction oriented

processing such as on-line banking, processing such as on-line banking, hotel reservations etc.hotel reservations etc.

Total capacity for a RAID 5 system Total capacity for a RAID 5 system with with NN disks = capacity of one disk disks = capacity of one disk * (N-1)* (N-1)

50

Hamming Code Example Hamming Code Example (One Bit Correction)(One Bit Correction)

4 data bits (D3, D5, D6, 4 data bits (D3, D5, D6, D7) – 1 0 1 0 and 3 D7) – 1 0 1 0 and 3 parity bits (P1, P2, P4)- parity bits (P1, P2, P4)- 1 0 11 0 1

P1 makes D3, D5, D7 P1 makes D3, D5, D7 even; P2 makes D3, D6, even; P2 makes D3, D6, D7 even; P4 makes D5, D7 even; P4 makes D5, D6, D7 evenD6, D7 even

Check parities after Check parities after transmission. 0= parity transmission. 0= parity check is correct; 1= check is correct; 1= parity check failedparity check failed

1 10110 0P1 D3 D5P4 D7D6P2

P4P4 P2P2 P1P1 Error Error inin

00 00 00 No No errorerror

00 00 11 P1P1

00 11 00 P2P2

00 11 11 D3D3

11 00 00 P4P4

11 00 11 D5D5

11 11 00 D6D6

11 11 11 D7D7