Sie sind auf Seite 1von 5

History

CS 102 1898 – Valdemar Poulsen invented magnetic


File Structures & wire recording
File Organizations 1928 – Fritz Pfleumer invented the magnetic
tape for recording sound
1951 – Magnetic tape was first used to record
computer data on Eckert-Mauchly
Chapter 03 - A UNIVAC I
Magnetic Tapes 1980s – Compact audio cassettes were used
with home computers for backup
purposes.

CJD

Sequential Access Usage


 a sequential-access storage device  used for backup/archiving purposes
 blocks of data are stored serially along the length  used to represent computers in TV and movies
of a tape  modern versions of the tape still relied upon by
 can only be accessed in a serial manner US national research labs, NASA and human
 for sequential file organization only genome project

 must read the first n records to read the (n+1)st  $2.3 Billion Industry (2004)
record

a record a block of 3 records


CJD CJD

CPU – I/O Device Communication Cost


 The CPU and I/O devices communicate via the  Cheaper than magnetic disks: Tapes cost less
I/O channel (processor) and the device controllers than 1 U.S. cent per gigabyte vs 25 U.S. cents to
$3 on disk per gigabyte
CPU  Costs change over time
Main
Memory Tape Tape
Controller Drive
Buffer I/O
Channel
Disk Disk
Controller Drive
System
Bus I/O
Bus
CJD CJD

1
New Research Physical Description
 new research by IBM and Fuji makes tapes still A thin strip of plastic about 0.002 in thick and
viable for the next decade 0.5 in wide but varies.
29.5 billion bits per square inch – cartridges Lengths vary: 300- 600- 1200- and 2400-ft long
can hold 35 TB of data which is several times Coated with magnetic oxide
bigger than the capacity of hard disks of the
same size Wound onto reels

uses thinner tape with barrium ferrite as


magnetic medium that protrude perpendicularly
from the tape allowing denser storage
spooling the tape provides more data per
volume – disks cannot be spooled
source : http://www.technologyreview.com/computing/24406
CJD CJD

How Bytes are Stored Illustration


Data are organized into spots or bits in nine Bytes on a Tape :
tracks parallel to the edges of the tape
One character (byte) for every nine bits
across the tape’s width
Formats : EBCDIC (IBM) or ASCII encoding
schemes
One bit is the parity bit. Method is fixed per
tape drive.
odd parity – the number of bits to represent a
character (including the parity bit) must be odd
even parity – the number of bits to represent a
character must be even
CJD CJD

Recording Density Transfer Speed


number of characters per inch  Tape drive speed or transfer speed

typical densities are 200, 556, 800, 1600, data can be read when tape is moving at a
3200, 6500 and beyond constant speed past the read/write heads

higher densities mean greater capacity typical speeds are 75, 125, 200 inches or
more per second
the tape movement is accelerated to get to
this speed and decelerated after each access.
the amount of data transferred per access is a
physical record which is a “block” of data
typical transfer rates range from 60 to 1250
KB/sec or more
CJD CJD

2
A Tape Drive Interblock Gaps (IBG)
Source Destination  Interblock Gaps – unoccupied space on the tape
Tape Tape between blocks
Reel Reel
provided for the acceleration and deceleration

Read/Write
usually 0.75 in, 0.6 in or less
Head
decreases the amount of tape available for
data

IBG IBG
CJD CJD

Markers Labels
 End-of tape markers are at beginning and end volume label - after the load-point marker that
beginning of tape has a load-point-marker – a contains a serial number and other identifying
reflective aluminum strip to align the tape head information
in preparation to start accessing the tape. file header label at beginning of each file
end-of-tape marker – another aluminum strip to contains file identifiers
prevent tape to warn tape drive and avoid the file trailer label at end of each file contains
tape unthreading from the reel block counts

Volume File Header Block Block Block File Trailer


Label
File 1 File 2
Label
IBG
1
IBG
2
IBG ... IBG
n
IBG
Label

CJD CJD

Blocking Space Calculations


Problem:
large blocks reduce the number of IBGs to If a tape drive has a density of 1600 bpi and IBG length of
store a file and 0.5 in, how much space is needed by a file with 100,000
records, 100 bytes each, with a blocking factor of 32 ?
allows more data to be stored on tape
results in less time to read a file Solution:
number of blocks = 100,000 records / (32 records / block)
= 3,125 blocks
block length :
in bytes : 100 bytes/record x 32 records/block
= 3,200 bytes/block
in inches : 3,200 bytes/block / 1,600 bytes/in
= 2.0 in/block

CJD CJD

3
Space Calculations Blocking Factor vs. Tape Length
Problem: If a tape drive has a density of 1600 bpi and IBG
length of 0.5 in, how much space is needed by a file with Question :
100,000 records, 100 bytes each, with a blocking factor of 32 ?
If you increase the blocking factor, will the
Solution (continued): tape length required increase or decrease ?
number of blocks = 3,125 blocks
block length = 2.0 in/block Answer :
number of IBGs = 3,125 blocks + 1 = 3,126 IBGs Higher blocking factor results in less IBGs,
tape length = 3,125 blocks x 2.0 in/block therefore shorter tape is required.
+ 3,126 IBG x 0.5 in/IBG
= 7,813 inches
= 651.083 ft (excluding labels and markers)

CJD CJD

Time Calculations Time Calculations


Problem : Solution :
block length = (80 bytes/record x 7 records/block)
Compute the time required to read a file on a tape with the
following characteristics: / (1600 bytes/in)
records = 237162 records, 80 bytes each = 0.35 in / block
blocking factor = 7 number of blocks = 237162 / 7 = 33880.29
density = 1600 bpi = 33881 (rounded up)
assume tape stops completely after reading each block = 33,881 blocks
start time = stop time = 10 ms/block time to read 1 block = start time + transfer time + stop time
start/stop time includes the time to advance the IBG = 0.1 sec/block
transfer speed = 50 in/sec + block length / transfer speed
+ 0.1 sec/block
= 0.2 sec/block + (0.35 in / block) / (50 in / sec)
= 0.027 sec/block
time to read file = 0.027 sec/block x 33,881 blocks
= 914.787 sec
= 15.2 min
CJD CJD

Blocking Factor vs Access Time Buffering Calculations


Problem :
Given 33,881 blocks
Question : I/O time per block = 0.027 sec/block
If you increase the blocking factor, will the CPU processing speed = 25 ms/block
time required to read the file increase or Find the time to read the file with
(a) single buffering and (b) double buffering
decrease ? Solution :
Answer : (a) single buffering
total time = (0.027 + 0.025) sec/block * 33,881 blocks
Higher blocking factor reduces the number = 1,761.812 sec = 29.36 min
of IBGs and results in less starts/stops, (b) double buffering
resulting in less time required to read the file. observe that processing is slightly I/O bound
total time = time to do I/O for all blocks
+ time for CPU to process last block
= 0.027 * 33,881 + 0.025
= 914.812 sec = 15.25 min
CJD CJD

4
Blocking and Buffering Bibliography
Conclusions :
 http://www.ibm.com
 Blocking reduces space and time needed.
 http://www.usbyte.com
 Buffering further reduces total processing
time by overlapping I/O time with CPU time.  http://computer.howstuffworks.com
 http://en.wikipedia.org
 among others

CJD CJD

End

CJD

Das könnte Ihnen auch gefallen