The Memory System

UNIT 5 THE MEMORY SYSTEM

Structure Page Nos.
5.0 Introduction
5.1 Objectives
5.2 The Memory Hierarchy
5.3 SRAM, DRAM, ROM, Flash Memory
5.4 Secondary Memory and Characteristics
5.4.1 Hard Disk Drives
5.4.2 Optical Memories
5.4.3 Charge-coupled Devices, Bubble Memories and Solid State Devices
5.5 RAID and its Levels
5.6 Summary
5.7 Answers

5.0 INTRODUCTION
In the previous block, fundamentals of a computer system were discussed. These
fundamentals included discussion on von-Neumann architecture based machines,
instruction execution, representation of digital data and logic circuits etc. This Block
explains the most important component of memory and Input/output systems of a
computer. This unit covers the details of the Memory. This unit discusses issues
associated with various components of the memory system, the design issues of main
memory and the secondary memory. Various characteristics of secondary memory and
its types that are used in a computer system, would also be discussed. The unit also
defines how multiple disks can be used to create a redundant array of disks that can be
used to provide a faster and reliable storage.

5.1 OBJECTIVES
After going through this Unit, you will be able to:

 explain the key characteristics of various types of memories and memory hierar-
chy;
 explain and differentiate among various types of random access memories;
 explain the characteristics of secondary storage devices and technologies;
 explain the latest secondary storage technologies;
 identify the various levels of RAID technologies

5.2 THE MEMORY HIERARCHY

In computers, memory is a device used to store data in binary form. Smallest unit of
binary data is called ‘bit’. Each bit of binary data is stored in a different cell or storage
unit and collection of these cells is defined as the memory. A memory system is
composed of a memory of fixed size and procedures which tells how to access the
data stored in the memory. Based on the persistence of the stored data, memory is
classified into two categories:

 Volatile memory: which loses its data in the absence of power.

 Non-volatile memory: Do not lose data when power is switched off.

Another classification of memory devices, which is also the objective of this unit is
based on the way they interact with the CPU which can be determined from figure 5.1
Main/ Primary memory interact directly with the CPU e.g. RAM and ROM.
5
Basic Computer Organisation Auxiliary/ secondary memory need I/O interface to interact with the CPU e.g.
magnetic disks and magnetic tapes. There are other memories like cache and registers,
which directly interacts with the CPU. Such memories are used to speed up the
program execution. For execution, a program must be loaded into the main memory
and should be stored on the secondary storage when it completes its execution.
Auxiliary memory is used as a backup storage, whereas main memory contains data
and program only when it is required by the CPU.

Figure 5.1: Memory Interaction with CPU

Various memory devices in a computer system forms a hierarchy of components

which can be visualised in a pyramidal structure as shown in Figure 5.2. As you can
observe in the Figure 5.2 that at the bottom of the pyramid, you have magnetic tapes
and magnetic disks; and registers are at the top of the pyramid. Main memory lies at
the middle as it can interact directly with the CPU, cache memory and the secondary
memory. As you go up in the pyramid, the size of the memory device decreases, the
access speed, however, increases and cost per bit also increases. Different memories
have different access speeds. CPU registers or simply registers are fastest among all
and are used for holding the data being processed by the CPU temporarily but because
of very high cost per bit they are limited in size. Instruction execution speed of the
CPU is very high as compared to the data access speed of main memory. So, to
compensate the speed difference between main memory and the CPU, a very high
speed special memory known as cache is used. The cache memory stores current data
and program plus frequently accessed data which is required in ongoing instruction
execution.

You may note the following points about memory hierarchy:

 The size of the memory increases as you go down the memory hierarchy.
 The cost of per unit of memory increases as you go up in the memory hierarchy
i.e. Memory tapes and auxiliary memory are the cheapest and CPU Registers are
the costliest amongst the memory types.
 The amount of data that can be transferred between two consecutive memory
layers at a time decreases as you move up in the pyramid. For example, from
main memory to Cache transfer one or few memory words of size in Kilobytes
are accessed at a time, whereas in a hard disk to main memory transfer, a block
data of size of 1 Megabyte is transferred in a single access.
 One interesting question about the memory hierarchy is why having faster
smaller memories does not slow down the computer? This is primarily due to the
6
 fact that there is very high probability that a program may access the instructions The Memory System
and data in the closed vicinity of presently executing instruction and data. This
concept is further explained in next unit.

Figure 5.2: Memory Hierarchy

In subsequent sections and next unit, we will discuss various types of memories in
more detail.

5.3 SRAM, DRAM, ROM, FLASH MEMORY

The main memory is divided into fixed size memory blocks called words. Size of the
memory word may be limited by the communication path and the processing unit size.
As word size/ length denotes the amount of bits that can be processed by the processor
at one time. Each memory word is addressed uniquely in the memory. A 32-bit
processor uses a word size of 32 bits whereas 64-bit processor uses a word of 64 bits.
RAM (random access memory) is a volatile memory i.e. content of the RAM vanishes
when power is switched off. RAM is a major constituent of the main memory. Both
read and write operations can be performed on RAM, therefore, it is also known as
read-write memory. Access time of each memory word is constant in random access
memory. RAM can be constructed from two types of technologies - Static Random
Access Memory (SRAM) and Dynamic Random Access Memory (DRAM). The main
difference being that DRAM loses its content even if power is on, therefore requires
refreshing of stored bits in DRAM. Thus, DRAM is slower than SRAM, however, the
DRAM chips are cheaper. In general, DRAM is used as the main memory of the
computer, while SRAM is used as the Cache memory, which is discussed in details in
the next unit.
SRAM
SRAM can be constructed using flip-flops. It is a sequential circuit. A SRAM cell
using SR flip flop is shown in figure 5.3. As you can observe, this sequential circuit
has three inputs: select, read/write, and input and single output: output. When select
input is high “1” circuit is selected for read/write operation and when select input is
low “0” neither read nor write operation can be performed by the binary cell. Thus,
select input must be high in order to perform read/write operation by the binary cell.
Binary cell reads a bit when read/write input is low “0” and writes when read/write
input is high “1”. Third input input is used to write into the cell. The only caution over
here is that when read/write input is low “0” i.e. we want to perform a read operation,
then read operation must not be affected by the input input. This is ensured by
7
Basic Computer Orga
anisation inverted inpput to the firstt AND gate which
w guaranttees the inputt to both R annd S to be
low and thuus prevents anny modificatioon to the flip flop value. The
T characterristic table
of SR flip flop
f is given iin Unit 4 Blo
ock 1 for bettter understandding of the fuunctioning
of the binaryy cell.

Figure 5.3: Logic

L Diagram of
o RAM cell

Read operation: select iss high “1”, reaad/write is low

w “0” and inpput is either low “0” or
high “1” theen input to R and S will be
b 0 and flip flop
f will keepp its previouss state and
that will be the output.
Write operaation: select is
i high “1”, reead/write is high
h “1” and if
i input is loww “0” then
R will be hiigh “1” and S will be low w “0” and flipp flop will stoore “0” and if input is
high “1” theen R will go low”0” and S will go high “1” and flip flop
f will storee “1”.

A RAM chhip is composed of severral read/writee binary cellss. A block diiagram of

2mx n RAM M is shown in Figure 5.4. The T RAM shoown has a totaal capacity off 2m words
and each word is n bits long e.g. in 64 x 4 RAM M, the RAM has h 64 words and each
word is 4 biits long. To adddress 64 i.e. 26words, we need 6 address lines. So inn a 2m × n
RAM, we haveh 2m wordds where each word has n bits and RA AM has m-bbit address
which requiires___m addresss lines. The RAM
R is funcctional only when
w chip sellect (CS1)
signal =1 annd CS2 = 0. If chip sellect signal is not enabled or chip select signal is
enabled andd neither reaad nor write input is ennabled then data d bus willl in high
impedance state
s and no operation caan be perform med. During high impedaance state,
other input signals
s will be
b ignored whhich means ouutput has no lo ogical significance and
does not carrry a signal

4: Block Diaggram of 2mx n RAM

Figure 5.4
 The Memory System
DRAM
Dynamic Random Access Memory (DRAM) is a type of RAM which uses 1 transistor
and 1 capacitor (1T1C cell) for storing one bit. A block diagram of a single DRAM
cell is shown in Figure 5.5. In DRAM, transistor is used as a gate which opens and
closes the circuit and thus stops and allows the current to flow. Charging level of the
capacitor is used to represent the bit “1” and bit “0”. As capacitors tends to discharge
in a very short time period DRAM cells need to be refreshed periodically to store the
binary information despite continuous power supply. Hence they are called dynamic
random access memory. With low power consumption and very compact in size
because of 1T1C architecture DRAM offers larger storage capacity in a single chip.
Each DRAM cell in the memory is connected with Word Line (Rows) and Bit Line
(Columns) as shown in Figure 5.5. Word line (rows) controls the gates of the transfer
lines while Bit lines (columns) are connected to sense amplifiers i.e. to determine “0”
or “1”.

Figure 5.5: A DRAM cell

Figure 5.6 presents the general block diagram of 2M 2M × N DRAM, where binary
cells are arranged in a square of 2M × 2M words of N bit each. For example, 4 megabit
DRAM is represented in a square arrangement of (1024 × 1024) or (210 × 210 ) words
of 4 bit each. Thus, in the given example we have 1024 horizontal/ word lines and
1024 × 4 column/ bit lines. In other words, each element, which consists of 4 bits of
array, is connected by horizontal row lines and vertical column lines.

Figure 5.6: Block Diagram of DRAM

9
Basic Computer Organisation
Selection and role of various signals for read and write operation is as follows:

1. RAS (Row Address Strobe): On the falling edge of RAS signal, it opens or strobe
the address lines (rows) to be addressed.

2. /CAS (Column Address Strobe): Similar to /RAS, on the falling edge, this enables a
column to be selected as mentioned in the column address from the rows opened by
the /RAS to complete the read-write operation.

3. R/(/W), Write enable: This signal determines whether to perform a read operation
or a write operation. While the signal is low, write operation is enabled and data input
is also captured on falling edge of /CAS whereas high enables the read operation.

4. Sense amplifier compares the charge of the capacitor to a threshold value and
returns either logic “0” or logic “1”.

For a read operation once the address line is selected, transistor turns ON and opens
the gate for the charge of the capacitor to move to the bit line where it is sensed by the
sense amplifier. Write operation is performed by applying a voltage signal to the bit
line followed by the address line allowing a capacitor to be charged by the voltage
signal.

ROM (Read-Only Memory)

Another constituent of the main memory is ROM (read only memory). Unlike RAM,
which is read-write memory and volatile, ROM’s are read only and non-volatile
memory i.e. content of the ROM persist even if power is switched-off. Once data is
stored at the time of fabrication, it cannot be modified. This is why, ROM is used to
store the constants and the programs that are not going to change or get modified
during their lifetime and will reside permanently in the computer. For example,
bootstrap loader, which loads the part of the operating system from secondary storage
to the main memory and starts the computer system when power is switched on, is
stored in ROM.

A block diagram of 2m× n ROM looks similar to that of RAM. As ROM is a read-only
memory there is no need of explicit read and write signals. Once the chip is selected
using chip select signals a data word is read and placed on to the data bus. Hence, in
the case of ROM, you need an unidirectional data bus i.e. only in output mode as
shown in figure 5.7. Another interesting fact about ROM is that, ROM offers more
memory cells and thus, memory as compared to the RAM for same size chip.

Figure 5.7: Block Diagram of 2mx n ROM

10
 T Memory Syystem
The
m m
As shhown in Figu ure 5.7, 2 × n ROM has 2 words of n bits each foor which it haas m
addreess lines and n output datta lines. For example, in 128 × 8 ROM M, you have 128
memmory words off 8-bit each. For
F 128 × 8 ROM R i.e. 2m = 27, m = 7, yoou need 7 adddress
lines (minimum number
n of bitss required to represent
r 1288) and 8-bit ou
utput data buss.
Figurre 5.8 shows a 32×8 ROM M.

Figure 5..8: Internal diagram of 332x8 ROM

Unlikke RAMs, which

w are seequential circcuits, ROMss are combinnational circuuits.
Typiccally, to design a RAM of o specific sizze you need a decoder andd OR gates. For
exammple, to desiggn a ROM of size 32 x 8 bits
b you need a decoder off size 5×32 annd 8
OR gates.
g 5×32 decoder
d will have
h 5 input lines, which will act as 5 address linees of
the ROM,
R the deccoder will con
nvert 5-bit in
nput address tot 32 differennt outputs. Figgure
5.8 shows the connstruction of 32
3 × 8 ROM using
u 5×32 ddecoder and eiight OR gatess for
data output. ROM Ms of other sizes can be constructedd similarly. For examplee, to
consttruct a ROM of 64 × 4 RO OM, you need d a 6×64 decooder and fourr OR gates annd to
consttruct a ROM of size 256×88, you need 8×256 decoderr and 8 OR gaates.
As discussed,
d ROOMs are non-vvolatile mem mory and conteent of the ROOM once writtten,
cannnot be changedd. Therefore, ROMs are ussed to store thhe look-up tabbles for consttants
to sp
peed up the computation.
c In addition, ROM can stoore the boot loader progrrams
and gaming
g progrrams. All this requires, zeero error in writing
w of succh programs and
thereefore, ROM device
d fabricaation requires very high precision. Consstructing a ROOM,
as shhown in figurre 5.8, requires decision about
a which iinterconnectioons in the cirrcuit
shouuld be open annd which interconnections should be cloosed. There are
a four ways you
can program
p a RO
OM which aree as follows:
1. Mask
M ROM (MROM):
( Maasking of RO OM is done bby the device manufactureer in
the
t very last phase
p of the fabrication prrocess on cusstomers speciial request. Mask
M
ROMs
R are cuustomised ass per the useer requiremennts, thus, aree very costlyy as
different
d maskks are requireed for differennt specificatioons. Because of very high cost
of
o masking, thhis customizaation is generaally used in m manufacturing g of ROM at vvery
large
l scale.
2. Programmabl
P e ROM (PR ROM): MRO OMs are noot cost effeective for sm mall
productions,
p P
PROMs are preferred
p for small quantitties. PROMs are programm med
using
u a special hardware which
w blow fuses
f with a very high vooltage to prodduce
logic
l “0” andd intact fuse defines
d logic “1”. The conntent of PROM is irreverssible
once
o program
mmed.
3. Erasable
E PRO
OM (EPROM M): EPROM Ms are third type of RO OMs which are
restructured
r or med using shortwave radiaations. An ulttraviolet lightt for
o reprogramm
Basic Computer Organisation a specific duration is applied to the EPROM, which destroys/ erases the internal
information and after which EPROMs can be programmed again by the user.
4. Electrically EPROM (EEPROM) : EEPROMs are similar to EPROMs except of
using ultraviolet radiations for erasing PROM, EEPROM uses electrical signals to
erase the content. EEPROM can be erased or reprogrammed by the user without
removing them from the socket.

Flash Memory
Flash memory is a non-volatile semiconductor memory which uses the programming
method of EPROM and erases electrically like EEPROM. Flash memory was
designed in 1980s. Unlike, EEPROM where user can erase a byte using electrical
signals, a section of the memory or a set of memory words can be erasable in flash
memory and hence the name flash memory i.e. which erases a large block of memory
at once. Flash memory is easily portable and mechanically robust as there is no
mechanical movement in the memory to read-write data. Flash memory is widely used
in USB memory, SD and micro SD memory cards used in cameras and mobile phones
respectively.
There are two types of flash memory, viz. NAND flash memory, where read operation
is performed by paging the contents to the RAM i.e. only a block of data is accessed
not an individual byte or word; and NOR flash memory, which are able to read an
individual memory byte/word or cell.

The features of various semiconductor memories are summarised in the Table 1.

Erase
Write Volatile/
Memory Type Mechanism/
Mechanism Non- Volatile
Level
Random-access
Read–Write Electrical/ Byte Electrical Volatile
Memory (RAM)
Read –only
Read–Only Not Applicable Masks Non-volatile
Memory (ROM)
Programmable
Read–Only Not Applicable Electrical Non-volatile
ROM (PROM)
Erasable PROM Read-
UV light/ Chip Electrical Non-volatile
(EPROM) mostly
Electrically
Read- Electrical/
Erasable Electrical Non-volatile
mostly Byte
(EEPROM)
Read- Electrical/
Flash memory Electrical Non-volatile
mostly Block

Table 1: Features of Semiconductor Memories

Check Your Progress 1

1. Differentiate among RAM, ROM, PROM and EPROM.
……………………………………………………………………………………

……………………………………………………………………………………

2. What is a flash memory? Give a few of its typical uses.

……………………………………………………………………………………

……………………………………………………………………………………
12
3. A memory has a capacity of 16K  16 The Memory System
(a) How many data input and data output lines does it have?
(b) How many address lines does it have?
……………………………………………………………………………………

……………………………………………………………………………………
4. A DRAM that stores 4K bytes on a chip and uses a square register array. Each
array is of size 4 bits. How many address lines will be needed? If the same
configuration is used for a chip which does not use square array, then how many
address lines would be needed?
………………………………………………………………………………………
………………………………………………………………………………………
5. How many RAM chips of size 256K  4 bit are required to build 1M Byte
memory?
……………………………………………………………………………………...
……………………………………………………………………………………...

5.4 SECONDARY MEMORY AND

CHARACTERISTICS
In previous section, we have discussed various types of random access and read only
memories in detail. RAM and ROM together make the main memory of the computer
system. You know that a program is loaded into main memory to complete its
execution. Computational units or CPU can directly interact with the main memory.
Hence, faster main memory, which can match with the speed of CPU, is always
desirable. In the previous section configuration of two types of RAMs, viz SRAM and
DRAM were discussed. As you may observe the SRAM consists of flip-flop based
circuits, therefore, is quite fast in comparison to DRAM However, the cost per bit of
DRAM is much less than the SRAM. Thus, you may observe the size of main memory
is much more than cache. It is discussed in more details in the next unit. To achieve
high speed, cost per bit of main memory is generally high which also limits its size.
On the other hand, as we have discussed, RAM, which is a major constituent of the
main memory, is volatile i.e. content of the main memory is lost when power is
switched off. Because of above mentioned issues, you need a low cost and high
capacity, non-volatile memory to store program files and the data for later use.
Secondary memory devices which are at the bottom of the memory hierarchy pyramid
are ideal for the said purpose. We will discuss various secondary storage devices in
this section.

5.4.1 Hard Disk Drive

In the era of Big Data, in which variety of data is generated rapidly, large secondary
storage has become an important component of every computer system. Today, hard
disk drives (HDD) is the primary type of secondary storage. The size of hard disk
drives in modern computer system ranges from Gigabytes (GB) to Terabytes (TB).
Internal hard drives extends the internal storage of a computer system whereas
external hard drives are used for back up storage.
HDD are electro-mechanical storage devices, which store digital data in the form of
small magnetic fields induced on the surface of the magnetic disks. Data recorded on
the surface of magnetic disks is read by disks read/write head, which transforms
magnetic signal to electrical signal for reading and electrical signal to magnetic field

13
Basic Computer Orga
anisation for writing.. HDD is coomposed of manym concenntric magneticc disks mounnted on a
central shaft
ft as shown in Figure 5.8.

Figuure 5.8: Interrnal structurre of Hard diisk drives (H

HDD)

Figure 5.8 shows the innternal structture of an H HDD. An HD DD is made of o several

concentric magnetic
m diskks mounted on o a central sshaft called spindle.
s Each magnetic
disk is madee of either glaass or an alum
minium disk called platterr. Each platterr is coated
with ferrom magnetic maaterial for storing
s data. Platter itseelf is made of non-
ferromagnettic material soo that its ownn magnetic fieeld should nott interfere thee magnetic
field of the data. Generaally, both sidees of the plattter is coated with
w magneticc material
for good stoorage capacityy at low cost.
Data recordded on the disk is accessedd through a read/write headd. Each side oof the disk
has its own read write heead. Each readd/write head isi positioned at a distance of tens of
nanometer called
c flying height to thee platter so thhat it can eassily sense or detect the
polarization
n of the magneetic field.

Figuure 5.9: Read/ Write Heaad

Two motorss are used in HDD. First one o is called the spindle motor,
m which is used to
d motor is used to move
rotate the sppindle on whiich all the plattters are mouunted. Second
the read/write heads across the entire surface of the platter radially and is called The Memory System
actuator or access arm.
Magnetic Read and Write Mechanisms
During a read/ write operation, read/write head is kept stationary while platter is
rotated by the spindle motor. As you know, data on the disk is recorded in the form of
magnetic field. The current is passed through the read/write head which induces a
magnetic field on the surface of platter and thus, records a bit on the surface. Different
directions of current generates magnetic fields with different polarities and hence are
used for storing “1” and “0”. Similarly, to read a bit from the surface, the magnetic
field is sensed by the read/write head which produces an electric current of the same
polarity and hence the bit value is read.
Data Organization and Formatting
As discussed and shown in figure 5.8, hard disk drives consists of number of
concentric platters which are mounted on a spindle forming a cylindrical structure.
Data is written in the form of magnetic fields on both surfaces of these platters and is
read by read/write head which is connected to an actuator. In this section, we will
discuss structure of magnetic disk in detail.
Structure of the disk is shown in figure 5.10. As you know, each magnetic disk is a
circular disk mounted on a common spindle but entire disk space is not used for data.
Disk surface is divided in to thousands of concentric circular regions called tracks.
The width of every track is kept the same. Data is stored in these tracks. Magnetic
field of one track should not affect the magnetic region in the other track thus two
tracks are kept apart with each other by a constant distance. Further, each track is
divided into number of sectors and two sectors are kept apart using inter-sector gap.
Data is stored in these sectors. Each track forms a cylindrical structure with other
tracks on other platters below or above it. For example, an outer most cylinder will
have outer most track of all the platters. So, if we have n tracks in a platter then there
will be n concentric cylinders too.
Components of the drive are controlled by a disk controller. Now a days, disk
controllers are built in to the disk drive. A new or blank magnetic disk is divided into
sectors. Each sector has three components: header, 512 byte (or more) data area and a
trailer. This process of is called physical / low level formatting. Header and trailer
contains metadata about the sectors e.g. sector number, error correcting code etc. Disk
controller uses this information whenever it writes or reads a data item on to a sector.
Data is stored in series of logical blocks. The disk controller maps the logical blocks
on to the physical disk space and also manages sectors which have been used for
storing data and which are still free. This is done by the operating system after
partitioning the disk in to one or more groups of cylinders. Disk controller stores the
initial data structure file of every sector on to the disk. This data structure file contains
a list of used and free sectors, list of bad sectors etc. Windows uses File Allocation
Table (FAT) for the said purpose.

15
Basic Computer Orga
anisation

Figuree 5.10: Magn

netic Disk Sttructure of CAV
C

There are tw wo arrangemments with whhich platters aare divided innto tracks annd sectors.
The first arrrangement is called as connstant linear vvelocity (CLVV), in which thhe density
of bits per track
t is kept uuniform, i.e. outer tracks aare longer thaan the inner tracks
t and
hence contaains more num mber of sectoors and data. O Outermost traacks are generally 40%
longer than the innermosst track. In this arrangemennt, in order to o maintain unniform bit/
mong tracks, tthe rotation speed
data rate am s is increeased from ouutermost to innner most
track. This approach
a is used by CD-ROM and DVD D-ROM drivees.

In another approach
a callled as constaant angular velocity
v (CAV)
V), the densityy of bits /
data per track is decreassing as we move from innnermost trackk to outermost track by
keeping thee disk rotationn speed consttant. As disk is moving att a constant speed,s the
width of thee data bits inccreases in thee outer trackss, which resullts in the connstant data
rate. Figuree 5.10 shows that the widdth of sectors in outer traacks is increasing and
density of bits is decreasiing.

Disk Perforrmance

Data is readd and written on the disks by the operatting system for f usage at laater stage.
A disk storees the program ms and relateed data. How
wever, disk is a much slow wer device
and the proograms storedd on it cannoot be executeed by the prrocessing uniit directly.
Therefore, the
t programs and its related data, whicch are not in the main meemory, are
loaded in thhe main mem mory from thhe secondaryy storage. Sinnce, the speeed of disk
read/write iss very slow inn compared tot RAM, timee to read or write
w a byte from
fr or on
to the disk affects the coomputer overrall efficiencyy. Therefore, in a single read/write
r
operation on disk data ofo one or morre sectors is transferred
t too/from the meemory. An
operating syystem, in geneeral, request for
f read/writee to one or moore sectors onn the disk.
The time takken by the dissk to complette a read/ writte request of the
t operatingg system is
known as diisk access timme. There are number of faactors which affect the perrformance
of the disk. These factorss are:
1. Seek Tiime: It is deffined as a tim
me taken by tthe read/writee head, or simmply as a
head, to
o reach the desired
d track on which thee requested sector
s is locaated. Head
should reach
r the desired track in minimum tim
me. Shorter seeek time leadds to faster
I/O operration.
2. Rotationnal Latency: Since, every track consistts of a numbeer of sectors, therefore,
the readd/write operaation can beb completedd only when the desired sector is
availablle under the read/write head
h for the I/O operatio on. It dependds on the
 rotational speed of the spindle and is defined as a time taken by a particular sector The Memory System
to get underneath the read/write head.
3. Data Transfer Rate: Since, large amount of data is transferred in one read/write
operation, therefore, the data transfer rate is also a factor for I/O operation. It is
defined as the amount of data read or written by the read/write head per unit time.
4. Controller overhead: It is the time taken by the disk controller for mapping logical
blocks to physical storage and keep track of which sectors are free and which are
used.
5. Queuing Delay: time spent waiting for the disk to be free.
The disk access time is defined as the summation of seek time, rotational latency, data
transfer rate, controller overhead and queuing delay and is given by the equation.

𝑎𝑐𝑐𝑒𝑠𝑠 𝑠𝑒𝑒𝑘 𝑟𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑑𝑎𝑡𝑎 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑟

𝑞𝑢𝑒𝑢𝑖𝑛𝑔

Out of the five parameters mentioned in the above equation, most of the time of the
disk controller goes in moving the read/write to the desired location and thus seeking
the information. If the disk access requests are processed efficiently then performance
of the system can be improved. The aim of disk scheduling algorithm is to serve all
the disk access requests with least possible head movement. There are number of disk
scheduling algorithms which are presented here in brief.

First Come First Serve (FCFS) scheduling: This approach serves the disk access
request in the order they arrived in the queue.

Shortest Seek Time First (SSTF) scheduling: Shortest Seek Time First disk scheduling
algorithm selects the request from the queue which requires least movement of the
head.

SCAN scheduling: The current head position and the head direction is the necessary
input to this algorithm. Disk access requests are serviced by the disk arm as disk arm
starts from one end of the disk and moves towards the other end. On reaching the
other end the direction of the head is reversed and requests are continued to be
serviced.

C-SCAN scheduling: Unlike SCAN algorithm, C-SCAN does not serve any request in
the return trip. Instead, on reaching to the end, it reverses back to the beginning of the
disk and then serves the requests.

LOOK scheduling: LOOK is similar to SCAN algorithm with only a single difference,
after serving the last request, LOOK algorithm does not go till the end instead it
immediately reverses its direction and moves to the beginning of the other end.

5.4.2 Optical Memories

So far, the storage devices you have studied are based on either electric charge or
magnetic field. Magnetic memories are primarily used as a secondary storage device,
but they can easily be damaged. However they have lower cost per bit than solid state
devices.
First laser based memory was developed in 1982 by Phillips and Sony. Laser based
storage devices uses a laser beam to read or write data and are called as optical
memories or optical storage devices. As laser beams can be controlled more precisely
and accurately than magnetic read/write heads. Data stored on optical drives remains
unaffected by the magnetic disturbances in its surrounding.

17
Basic Computer Orga
anisation Initially, theese optical stoorage devices commonly known
k as com
mpact disk (CDD) or CD-
DA (Digitall Audio) weree used to storre only audioo data of 60 minute
m duratiion. Later,
huge comm mercial succeess of CD leead to devellopment of low cost opptical disk
technology. These CDs can be used as auxiliary storage and can store anny type of
digital data. A variety off optical-disk devices have been introduuced. We brieffly review
some of these types.

Compact Disk
D ROM (C
CD-ROM)
Compact Disk
D or CD-R ROM are mad de of a 1.2 m mm thick sheeet of a polyycarbonate
material. Eaach disk surfaace is coated with a reflecctive materiall generally alluminium.
The standarrd size of a coompact disk is 120 mm in diameter. Ann acrylic coat is applied
on top of thee reflective suurface to prottect the disk fr
from scratchess and dust.

F
Figure 5.11:: Outer Layoout of a CD

Unlike maggnetic disks, data on an optical

o disk iss recorded inn a spiral shaape tracks.
Each track is separated by a distancee of 1.6 mm.. Data in a trrack is recordded in the
form of landd and pit as shown
s in Figuure 5.13. Whhen a focused laser beam inn incident
on to the opptical disk, thhe disk is burrned as per thhe digitally reecorded data forming a
pit and landd structure. Thhe data is reaad from the suurface by meaasuring the inntensity of
the reflected
d beam. The pit area scattters the incideent beam, whhereas the lannd reflects
the incident beam, whichh are read as “0”“ and “1” reespectively.

Figure 5.112: Spiral traack of CD Figure 5.133: Land & Piit formation in CD trackk

As shown in i CD are in spiral shape. The tracks inn CDs are

n Figure 5.122, the tracks in
further divid
ded into secttors. All sectoors in CDs arre equal in leength. This means
m that
density of data
d recorded on the disk is uniform acrross all the traacks. Inner trracks have
less numberr of sectors whereas
w outer tracks
t have m
more sectors. CD-ROM devvices uses
constant lineear velocity ((CLV) methodd for reading the disk content. In this method,
m the
disk is rotated at lower velocity as we move away from the center of the disk. This The Memory System
ensures a constant linear velocity at each track of the CD. The format of a sector of
CD is show in Figure 5.14.

SYNC HEADER DATA L-ECC

12 Bytes 4 Bytes 2048 Bytes 288 Bytes

Figure 5.14: Sector format of CD

Data on the CD-ROM are stored in a track as a sequence of sectors. As shown in the
Figure 5.14 each sector has four fields viz. sync, header, user data followed by error
correcting codes. Each part of the sector is described below:

• Sync: It is the first field in every sector. The sync field is 12 byte long. The
first byte of sync field contains a sequence of 0s followed by 10 bytes of all 1s
and 1 byte of all 0s.

• Header: Header is four byte field in the sector. Three bytes are used to
represent the sector address and one byte is used to represent the mode i.e.
how subsequent fields in the sector are going to use. There are 3 modes:

• Mode Zero: Specifies a no user data i.e. blank data field.

• Mode One: Specifies an user data of 2048 bytes followed by 288 bytes of
error correcting code.
• Mode Two: No error correcting code will be used thus subsequent field
will contain 2336 bytes of user data.

• Data: Data field contains the user 2048 byte of user data when mode is 1 or
mode 2.

• L-ECC: Layered error correcting code field is 288 byte long field which is
used for error detection and correction in mode 1. In mode 2, this field is used
to carry an additional 288 bytes of user data.

Compact Disk Recordable (CD-R)

CD-Recordable are the compact disks which are capable of storing any type of digital
data. The physical structure of CD-R is same as that of CD-ROM as discussed in
previous section except that polycarbonate disk has a very thin layer of an organic dye
before the Aluminum coating. CD- R can record user data only once but user can read
the data many times thus these are also known as CD-WO (write once), or WORM
(write once read many). Many CD writers allow the users to write CD-R in multiple
session until CD is full. In each writing session, a partition is created in the CD- R.
But once written, data on CD-R cannot be changed or deleted. There are three types of
organic dyes used in CD-R.
Cyanine dyes are the most sensitive dye amongst the three types. CD-Rs have cyanine
dyes are green in color. Very sensitive to UV rays and even can lose the data if
exposed to direct sunlight for few days.
Phthalocyanine dye does not need a stabilizer as compared to cyanine dyes. They are
silver, gold or light green in color. They are very less sensitive as compared to cyanine
dyes but if exposed to direct sunlight for few weeks, it may lose the data.
Azo dye is the most stable among all types. It is most resistant to UV rays but if
exposed to direct sunlight for 3-4 weeks, the CD -R may lose the data.
19
Basic Computer Organisation Compact Disk Rewritable (CD-RW)
The CD-RW are re-writable optical disks. The data on CD-RW can be read or written
multiple times. But for writing again on the already written CD-RW, the disk data
must be erased first. There are two approaches of erasing the data written on CD-RW.
In the first approach, the entire disk data is erased completely i.e. all traces of any
previous data is erased. This is called full blanking. Whereas in another approach
called as fast blanking, only the meta data is erased. The later approach is faster and
allows rewriting the disk. The first approach is used for confidentiality purposes.
The phase change technology is used in CD-RW. The phase change disk uses a
material that has significantly different reflectivity in two different phase states. There
is an amorphous state, in which the molecules exhibit a random orientation and which
reflects light poorly; and a crystalline state, which has a smooth surface that reflects
light well. A beam of laser light can change the material from one phase to the other.
The phase change technology of CD-RW uses a 15-25 % degree of reflection whereas
CD-R works on 40-70 % degree of reflection.

Digital Versatile Disk (DVD)

Digital versatile disk commonly known as DVD is also an optical storage device like
CD, CD-R, CD-WR. Among the three DVDs have highest storage capacity ranges
from 1.4 GB to 17 GB on a single side. The higher storage capacity is enabled by the
use of laser beams of shorter wavelength as compared to compact disks. DVD uses a
laser beam of 650 nm whereas compact disk uses a laser beam of 780 nm. Shorter
wavelength laser beam creates shorter pits on the polycarbonate disk, thus offers
higher storage capacity for similar dimensions. DVD-Audio and DVD-video are a
standard format for recording audio and video data on DVDs. Like compact disks,
DVD also comes in various variants like DVD-ROM, DVD-R, DVD-WR etc.

Blue Ray Disk

A blue ray disk is a digital disk that can store several hours high definition videos.
Blue ray disks is of the same size of DVD, but can store 25 GB to 128 GB of data. A
blue ray disk is designed to replace DVD technology. It has its applications in gaming
applications, which uses very high quality animations.

5.4.3 Charge-coupled Devices, Bubble Memories and Solid State

Devices
Charge-coupled Devices CCDs (CCDs)
Charge couple devices are photo sensitive devices which are used to store digital data.
CCD is an integrated circuit of MOS-capacitors called cells, which are arranged in an
array like structure in which each cell is connected with its neighbouring cell. Each
capacitor can hold the charge which is used to represent the logic “1”. While reading
the array of capacitors, the capacitor moves its charge to the neighbouring capacitor
with next clock pulse. CCD arrays are mainly used in representing images and video
data, where presence and absence of charge in the capacitor represents the
corresponding pixel intensity.
As mentioned, CCD are highly photo-sensitive in nature and thus, produces a good
quality picture even if light is dim or in low illumination intensity. Now a days, CCDs
are widely used in digital cameras, satellite imagery, radar images and other high
resolution imagery applications.

Magnetic Bubble Memories

Working principle of magnetic bubble memory is similar to that of charge coupled
devices (CCD) discussed in the previous section. Magnetic bubble memory is an
20
arrangement of small magnetic area called bubble on a series of parallel track made of The Memory System
magnetic material. Each bubble represents a binary “1” and absence of a bubble on
magnetic material is interpreted as “0”. Binary data is read from the memory by
moving these bubbles towards the edge of a track under the influence of external
magnetic field. Magnetic field produces as bubbles remain persistent and do not
demagnetise by its own. So, magnetic bubble memories are non-volatile type
memories.

Solid State Devices (SSD)

Solid state drives also known as solid state storage devices are based on flash memory.
As discussed, flash memory, a non-volatile type memory uses semiconductor devices
to store the data. The major advantage of SSD is that it is purely an electronic device
i.e. unlike HDD, SSD does not have mechanical read/ write head other mechanical
components. Hence, reading and writing through SSD is faster than HDD. Now a
days, SSD have replaced HDD in computer systems, however, SSD disks are more
expensive than HDDs.

Check Your Progress 2

1. What will be the storage capacity of a disk, which has 8 recording surfaces, 32
tracks with each track having 64 sectors. Also, what would be the size of one
cylinder of the disk? You may assume that each sector can store 1 MB data.
……………………………………………………………………………………

……………………………………………………………………………………
2. What would be the rotation latency time for the disk specified above, if it has a
rotational speed of 6000 rpm?
……………………………………………………………………………………

……………………………………………………………………………………

3. What are the advantages and disadvantages of using SSD over HDD?
………………………………………………………………………………………………

………………………………………………………………………………

4. What are the differences between CLV and CAV disks?

……………………………………………………………………………………...
……………………………………………………………………………………...

5.5 RAID AND ITS LEVELS

Continuous efforts have been made by researchers to enhance the performance of the
secondary storage devices. As pointed out in previous sections performance of the
secondary storage is inversely affected by disk access time. Lower the disk access
time higher would be the performance. What about an idea of providing parallel
access to a group of disks? With the use of parallel access the amount of data that can
be accessed per unit time can be enhanced by a significant factor. A mechanism which
splits the data on multiple disk is known as data striping. Data access through parallel
access allows users to access data stored at multiple disks simultaneously, thus
reduces effective reading time. Does data striping ensure protection of data against
disk failure?

21
Basic Computer Organisation Another important factor for secondary storage is the reliability of data storage
system. Storing same data on more than one disks enhances reliability. If one disk
fails, then data can be accessed through another disk. Replicating data on multiple
disks is called mirroring. Mirroring brings redundancy in data. So many schemes have
been employed to enhance the performance and reliability of data and collectively
they are called as redundant arrays of inexpensive disks (RAID). Based on the trade-
off between reliability and performance RAID schemes have been categorises into
various RAID levels.
Data striping increases the data transfer speed as different data bytes are accessed in
parallel from different disks in a single disk access time. Whereas mirroring protects
data from disk failures. If one disk fails then same data is accessed from the copy of
the data stored in other disk.

RAID Levels

RAID Level-0: RAID level-0 implements block splitting of data with no protection
against disk failures. In block splitting, each block is stored in a different disk in the
array. For example, ith block of a file will be store in ( i mod n ) + 1 disk, where n is
the total number of disks in the array. In this case, a significant enhancement on the
performance can be observed as n blocks can be accessed (one each from each disk) in
a single disk access time.

(a) RAID Level 0

RAID Level-1: This level protects data by implementing mirroring. If a system has 2
disks then each block of information will be stored in both of the disks. This ensures,
if one disk fails then same copy of the block is accessed from the second disk.
Mirroring introduces redundancy unlike level-0 which increases the data transfer rate.

Strip 0 Strip 1 Strip 2 Strip 3 Strip 0 Strip 1 Strip 2 Strip 3

Strip 4 Strip 5 Strip 6 Strip 7 Strip 4 Strip 5 Strip 6 Strip 7
Strip 8 Strip 9 Strip 10 Strip 11 Strip 8 Strip 9 Strip 10 Strip 11

Strip 12 Strip 3 Strip 14 Strip 15 Strip 12 Strip 13 Strip 14 Strip 15

(b) RAID Level 1

RAID Level-2: This level uses error detection and correction bits, which are extra bits
used for detection and correction of a single bit error in a byte. This is why this level
is also known as memory-style error correction code organization. If one of the disk
fails then parity bits and remaining bits of the byte are used to recover the bit value.

22
 The Memory System

b0 b1 b2 b3 f0(b) f1(b) f2(b)

(c) RAID 2 (Redundancy through Hamming Code)

RAID Level-3: Single parity disk is used in this scheme. Parity bit for a sector is
computed and stored in a parity disk. During the access, parity bit of the sector is
computed and if computed parity bit is equal to the stored parity, the missing bit is 0
otherwise it is 1. This RAID level is also known as bit-interleaved parity organization.
Thus has an advantage over level-2 that only single parity disk is used as compare to
number of parity disks in level-2. The biggest drawback of this approach is that all the
disks are used for single I/O operation in computation of the parity bit which slows
down the disk access and also restricts parallel access.

b0 b1 b2 b3 Parity(b)

(d) RAID Level 3

RAID Level-4: This level uses block striping and one disk is used to keep parity
block. This is also called block-interleaved parity organization. The advantage of
block interleaving is that parity block along with corresponding blocks on other disks
is used to retrieve the damaged block or the blocks of the failed disk. Unlike in level-
3, block access reads one disk which allows parallel access to other blocks stored in
other disks in the array.

Block 0 Block 1 Block 2 Block 3 Parity (1-3)

Block 4 Block 5 Block 6 Block 7 Parity (4-7)

Block 8 Block 9 Block 10 Block 11 Parity (8-11)

Block 12 Block 13 Block 14 Block 15 Parity (12-15

(e) RAID 4 (Block level Parity)

23
Basic Computer Organisation RAID Level-5: This level stores block of data and parity in all the disks in the array.
One disk store the parity while data is spread out on different disks in the array. This
structure is also known as block-interleaved distributed parity.

Block 0 Block 1 Block 2 Block 3 Parity (0-3)

Block 4 Block 5 Block 6 Parity (4-7) Block 7

Block 8 Block 9 Parity (8-11) Block 10 Block 11

Block 12 Parity (12-15) Block 13 Block 14 Block 15

Parity (16-19) Block 16 Block 17 Block 18 Block 19

(f) RAID 5 (Block-level Distributed Parity)

RAID Level-6: Level-6 uses error correcting codes for recovery of damaged data
while other levels uses parity. It also provides protection against multiple disks
failures. For the recovery purposes, this arrangement is used to store redundant data
on some of the disks, hence it is also called as p + q redundancy scheme. Here, p is the
number of disks that store the error correcting codes while q is the number of disks
that store redundant data.

Block 0 Block 1 Block 2 Block 3 Parity (0-3) Parity (0-3)

Block 4 Block 5 Block 6 Parity (4-7) Q (4-7) Block 7

Block 8 Block 9 P (8-11) Q (8-11) Block 10 Block 11

Parity 12 Parity (12- Q (12-15) Block 13 Block 14 Block 15

15)

(g) RAID Level 6

Table below summarises characteristics of various RAID levels.

I/O Data
Request Transfer
RAID Typical
Category Features Rate Rate
Level Application
(Read (Read
/write) /write)
Applications
a) The disk is divided Large Small which requires
0 Striping into blocks or sectors. Blocks: Blocks: high performance
b) Non-redundant. Excellent Excellent for non-critical
data
a) Mirror disk which
contains the same data
is associated with every
disk. Good / May be used for
1 Mirroring Fair /fair
b) Data Recovery is fair critical files
simple. On failure, data
is recovered from the
mirror disk.

24
 a) All member disks The Memory System
participate in every I/O
request.
b) Synchronizes the
spindles of all the disks
to the same position. Not useful for
Parallel
2 c) The blocks are very Poor Excellent commercial
Access
small in size (Byte or purposes.
word).
d) Hamming code is
used to detect double-
bit errors and correct
single-bit error.
a) Parallel access as in
level 2, with small data
Large I/O request
blocks.
Parallel size application,
3 b) A simple parity bit is Poor Excellent
Access such as imaging
computed for the set of
CAD
data for error
correction.
a) Each member disk
operates independently,
which enables multiple
input/output requests in
parallel. Not useful for
Independent Excellent/ Fair /
4 b) Block is large and commercial
access fair poor
parity strip is created purposes.
for bits of blocks of
each disk.
c) Parity strip is stored
on a separate disk.
a) Allows independent
access as in level 4.
b) Parity strips are
distributed across all High request rate
Independent Excellent Fair /
5 disks. read intensive,
access / fair poor
b) Distribution avoids data lookup
potential input/output
bottleneck found in
level 4.
Also called the p+q
redundancy scheme, is
Application
much like level 5, but
Independent Excellent/ Fair / requiring
6 stores extra redundant
access poor poor extremely high
information to guard
availability
against multiple disk
failures.

Check Your Progress 3

1. What is the need of RAID?
……………………………………………………………………………………

……………………………………………………………………………………

2. Which RAID levels provide good data transfer rate?

……………………………………………………………………………………
25
Basic Computer Organisation ……………………………………………………………………………………
3. Which RAID level is able to fulfil large number of I/O requests?
……………………………………………………………………………………

……………………………………………………………………………………

5.6 SUMMARY

This unit introduces the concept of memory hierarchy, which is primarily required due
to the high cost per bit of high speed memory. The processing unit have register,
cache, main memory and secondary or auxiliary memory. The main memory consists
of RAM or ROM. This unit explains the logic circuit and organisation of RAM and
ROM. The unit also explains several different types of secondary storage memories.
The unit provide details on hard disk and its characteristics. It also gives details of
different kind of optical disk. The concept of access time and constant linear and
angular velocity has also been explained in details. For larger computer systems
simple hard disk is not sufficient, rather an array of disks called RAID are used for
such systems to provide good performance and reliability. The concept of RAID and
various levels of RAID has been defined in this unit. The next unit will introduce you
to the concept of high speed memories.

5.7 ANSWERS

Check Your Progress 1

1. RAM is a sequential circuit, volatile, requires refreshing (DRAM) and is a read/
write memory; ROM, PROM and EPROM are mostly non-volatile memories.
ROM is a combinational circuits. All these ROMs are written mostly once and
read many times.
2 Flash memory is a non-volatile semiconductor memory, where a section of the
memory or a set of memory words can be erased. They are portable and
mechanically robust as there is no mechanical movement in the memory to read-
write data. Flash memory is used in USB memory, SD and micro SD memory
cards used in cameras and mobile phones respectively.
3. (a) Since a word of data is 16 bits, it will have 16 data input and 16 data output
lines, if not multiplexed.
(b) The number of words are 16K, which is 214. Thus, 14 address lines would be
required.
4. The memory must select one of the 4K bytes, which is 212. In case a square array
is used (as shown in Figure 5.6), then 6 row address and 6 column address lines
would be needed, which can be multiplexed. So just 6 address lines be sufficient.
However, for a non square memory you may require all 12 address lines.
5. Two chips will be required to make 256  8 memory. 4 such combinations would
be required to make 1 MB memory. Thus, you will require 8 such chips.
Check Your Progress 2
1. Storage capacity of a disk = recording surfaces × tracks per surface × sectors per
track × size of each sector
Storage capacity of a disk = 8× 32 × 64 × 1 MB = 23×25×26×220 = 234 = 16 GB
One cylinder will have = = 8× 64 × 1 MB = 23×26×220 = 512 MB

26
2. The time of one rotation = 1/6000 min = 60/6000 sec = 1/100 sec= 10 millisec The Memory System
Rotational latency = on an average time of half rotation = 5 ms

3. SSD drives does not require any mechanical rotation, therefore are less prone to
failure. In addition, they are much faster than HDD. But they are more expensive
than HDD

4. The size of sectors on CLV disks is same on the entire disk, therefore, these disks
are rotated a different speed. Density of data is same in all the sectors. In CAV
disks the rotation speed is same, thus, sector size is more in the outer tracks.
However, reading/writing process, in general, is faster.
Check Your Progress 3
1. RAID are a set of storage devices put together for better performance and
reliability. Different kind of RAID levels have different objectives.

2. Good data transfer rate are provided by RAID level 0, 2 and 3.

3. Large number of I/O requests are fulfilled by RAID level 0, 1, 4,5 ,6.

27