Data Storage
Secondary Storage Media and Its
Performance
CSCI 2410 1
Physical Media
Sequential Access Storage Devices
(SASD)
Direct Access Storage Devices (DASD)
CSCI 2410 2
SASD
“tape” paradigm – linear in nature
CSCI 2410 3
SASD
“tape” paradigm – linear in nature
data is retrieved according to its
physical location in the file
CSCI 2410 4
SASD
“tape” paradigm – linear in nature
data is retrieved according to its
physical location in the file
all data preceding the desired data
must be retrieved first
CSCI 2410 5
SASD
“tape” paradigm – linear in nature
data is retrieved according to its
physical location in the file
all data preceding the desired data
must be retrieved first
file size of “n” records, requires on the
average n/2 retrievals
CSCI 2410 6
Magnetic Tape
density – amount of data record per
unit of linear measure (bits / in) ~
bytes/in or bpi
CSCI 2410 7
Magnetic Tape
density
gap – an unused length of tape;
permits tape to achieve read/write
speed or stop, whichever is the case
CSCI 2410 8
Magnetic Tape
density
gap
nominal capacity – data capacity if the
entire tape were to store data
CSCI 2410 9
Magnetic Tape
density
gap
nominal capacity
physical record – amount of data
transferred from the tape device to
memory
CSCI 2410 10
Magnetic Tape
density
gap
nominal capacity
physical record
logical record – a unit of data as viewed
by the application
CSCI 2410 11
Magnetic Tape
density
gap
nominal capacity
physical record
logical record
blocking factor – number of logical
records per block, or physical record
CSCI 2410 12
Magnetic Tape
density
gap
nominal capacity
physical record
logical record
blocking factor
nominal transfer rate – the theoretic
maximum rate for transferring data (density *
tape transport speed)
CSCI 2410 13
DASD
3-dimensional paradigm
CSCI 2410 14
Magnetic Tape Parameters
Density Gap Size Gap Size
(bpi) (inches) (bytes)
800 0.6 480
1600 0.6 960
6250 0.3 1875
CSCI 2410 15
Magnetic Tape
Logical Record(s)
BF = 6
R R R R R R G ... ^Z
Physical record
Data block
CSCI 2410 16
BF LRCL
density TTS
Magnetic Tape - Calculations
Nominal capacity = density * 12in/ft * length of reel
Transfer Rate = (density * tape transport speed)
Blocksize = BF * LCRL
Time to read a block = blocksize/(transfer rate) + SS
Time to read a file =
(nbr Recs)/BF * (time to read one block)
CSCI 2410 17
Disk Device
disk pack
individual
platters – each
platter has two
surfaces
CSCI 2410 18
Disk Platter – top view
tracks – concentric
same capacity
say ~ 200 per surface
CSCI 2410 19
Disk Pack/platter – side view
cylinders
disk surfaces:
8 of 10 are used
access arm
spindle
read/write heads
CSCI 2410 20
DASD
3D paradigm
data is retrieved according to its
physical location on the device
CSCI 2410 21
DASD
3D paradigm
data is retrieved according to its
physical location on the device
positioning algorithm enables going to
the exact address of data
CSCI 2410 22
DASD
3D paradigm
data is retrieved according to its
physical location on the device
positioning algorithm enables going to
the exact address of data
file size of “n” records, ideally requires a
single retrieval
CSCI 2410 23
Track Format
sectors
CSCI 2410 24
Direct Access
Hashing algorithm:
h (primary key) (cc, ss, bb) ~ (x, y, z)
y
x
z
CSCI 2410 25
Disk Parameters
Latency
Home address
Seek Time
CSCI 2410 26
Access Time
Latency – 3600 rpm = ~8.3ms
Seek time – device specific
Average access time – vendor supplied
~ < 10 ms
Latency + Seek time: Seeking is MUCH more
expensive than latency; a modern interleaf
factor of 1 means we can (theorectically) read
one complete track per revolution thereby
removing latency as a consideration.
CSCI 2410 27
Access Time cont’d.
Caching & buffering – assuming a
buffer of 4Kb, grab minimum of 4K per
disk read which is 8 sectors.
Sector size assumed to be 512 bytes.
Intermediate cache, if sufficiently large,
may permit reading entire file into
memory.
CSCI 2410 28
Strategies
Allocate contiguous cylinders.
Create file in cylinder-sequence (fill
vertically then horizontally).
A file consisting of 5 cylinders requires only
5 seeks when reading sequentially.
At creation, order data so that physical and
logical orders coincide (if possible).
CSCI 2410 29