Input / Output Subsystem: Introduction - Computer Science & Engineering Video Lecture - Computer Science Engineering (CSE)

Today we are going to begin with the last
topic of this course namely input/output subsystem.
We have so far talked about processor and
memory. Processor is the central component
of the entire system where the programs are
executed. Memory is essential in order to
hold the program and data on which some instructions
operate and input output is important to connect
the computer to the rest of the world. If
you do not have any means of input and output
no information can enter the memory or the
processor and no information can come out.
So this is essential to make it useful because
after all computation involves that you have
some data, you process and produce the results.
So ultimately you need input and output. We
are going to look at various aspects of input/output
subsystem.
We will talk about performance considerations
particularly of I/O system. We will talk of
peripheral devices which are the extremities
of a computer system where the data gets transformed
into a form which is understandable by others;
it could be humans, it could be the environment
or the other information is brought in the
form which is understood outside into the
computer in form of 1s and 0s. Then we will
talk about interfaces and the buses. These
are the means to connect the input output
devices or peripherals to rest of the system
and finally we will look at the operational
aspect how I/O or input or output is actually
carried out. So our focus today would be on
the first two aspects.
I/O performance is contributed to the overall
performance but it often gets neglected. One
would talk of speeding up the processor, executing
more and more instruction anytime, making
access to memory fast. But if you neglect
I/O your benefit of performance improvements
may not really be fully effective. So consider
for example that you have some benchmark program
which takes hundred seconds and all out of
which the CPU consume 90 percent of the time
and 10 percent of the time is taken for I/O
that means the data is brought in, results
are sent out and we know that by technological
innovations the speed of processor is continuously
improving and if we have 50 percent improvement
every year so over a period of five years
we could see that this 90 seconds which the
CPU consume would get reduced to 60 40 27
18 and 12 leading to an overall improvement
by factor of seven and a half times.
Now, on the other hand, suppose I/O performance
remains unchanged then what is the effect
of this on the overall performance. So the
total time is what we for the CPU and 10 seconds
of the I/O remain unchanged so the total figure
would be 100, next state goes to 70, next
state goes to 50 and so on and finally we
have 22 after five years. So now if you look
at the ratio of 122 it is only a 4.5 times
improvement. So, not improving I/O performance
could bring down your expected performance
improvements. This was the case where the
program is computation bound. Computation
bound means that majority of the time is taken
in performing computations: arithmetic, logic,
decision and so on and there is a comparatively
much small amount of I/O.
On the other hand, there are situations which
are I/O bound where processing is very little,
it is essentially input and output. If you
if you look at scientific computation, let
us say weather prediction or nuclear modeling
and so on, here I/O is very little and most
of the time it is number crunching or heavy
computation. But in business data processing
or database oriented application it is often
most of the time gets spent in input and output.
So if the let us say ratios were reversed
and it is only 10 percent part which you are
improving then obviously there is not much
improvement which gets ramped. So it is important
to take care of I/O performance improvement
also in order to improve the overall performance.
The next question is how do you define I/O
performance; what do you mean by I/O performance?
As we saw in case of CPU, depending upon your
perspective your requirement there may be
different definition of what is I/O performance.
It could be throughput oriented measure that
means how much of input output gets done,
how much of data transfer gets done per unit
time. One could talk of amount of data transfer
in unit time like kilobytes per second or
megabyte per second or you could talk of number
of I/O operations per unit time. Each I/O
operation may involve some transfer: it could
be small, could be few bytes, could be few
kilobytes and we want to let us say maximize
how many such transfers take place.
On the other hand, it may be the response
time which is important. So, from the time
you take some action, you give some command
and it takes to respond to that so there may
be environments where that may be more important.
There could also be cases were both are important.
For example, if you take reservation system
for railways or airlines as a as a user as
an individual user you want your job to be
done as fast as possible. Once your requirements
are fed in you want the system to respond
immediately. So, response of course would
involve computation as well as input/output.
On the other hand, if you look at the whole
system, then there is a central database somewhere
where requests are being sent in and it is
responding. So at that point, at the point
of the central database it is a throughput
of I/O which might become bottleneck and you
may be trying to improve that. So, at different
in different parts of the system or with different
perspectives the consideration for I/O performance
may be different.
So if you are interested in let us say total
time the program takes to execute as we had
seen for processor then clearly you can divide
the total time between the time which is taken
for execution for computation and the time
for taking in the data or sending out the
data. Now, based on these considerations suppose
you look at a supercomputing application or
typically scientific application there it
may be typically a computation bound application
but there also may be at times lots of results
or simulation results which may have to be
printed out. So here it is not the response
time but again the data throughput which is
important which may be measured in kilobytes
or megabytes per second.
Transaction processing, for example, let us
say ATM, bank ATM. So here transaction per
second is important from the overall system
point of view and again for a person, when
you feed in a request for drawing some money
you want it to respond immediately so response
time and transactions per second both are
important.
Take example of a file server. Suppose you
have a cluster of computers as we have in
our department and there is one of the machines
which is acting as a file server maybe in
your homes are there so this file server would
receive request for creating files, opening
files, reading, writing and so on each such
request is essentially an I/O operation. so
your concern would be how many I/O operations
the system is able to sustain and that could
be one thing you would like to look at. so
depending upon the situation you may describe
your I/O consideration, I/O performance in
terms of some operations or some data transfer
per second or in terms of time it takes to
respond to you or some combination of both.
While talking of I/O performance one must
keep in mind that often there could be a discrepancy
in terms of units. When we have been talking
of memory we talk of KB, MB and so on. Once
again here we are talking of kilobytes per
second, megabytes per second. But typically
when you talk of memory size you mean 1kilobyte
as 1024 bytes or 2 raised to the power 10.
Similarly, 1 megabyte is 1024 kilobytes or
1048576 bytes which are basically which is
a power of 2/10th power of 2 and this should
have been 20th, it is 20th power of 2 whereas
in I/O you are talking of thousands and millions
in the real sense so 1 KB per second here
is thousand bytes per second or 10 is to the
power 3 and similarly 1 megabyte per second
is 1000 kilobytes per second or 1 million
bytes per second.
Approximately they are same but somewhere
if you want to be little more accurate one
must keep in mind that there may be a difference;
you know the language one is talking in memory
domain or in processor domain may be different
from what you are talking of in I/O domain
so this is just a word of caution.
Now let us move towards peripheral devices.
Before we do that let us look at the overall
system, how we conceive a stored program computed
typically.
This is a classical block diagram showing
five units of a computer which probably you
would have seen in your school days also.
You have control, arithmetic, memory, input
and output and they are connected to each
other so control and arithmetic are..... the
data path has been mentioned here, these form
the central processing unit or the central
processor. So we are talking of this part
which is input and output.
Elaborating it further this is how the picture
looks like. I am also bringing in two levels
of memory hierarchy here. There is a cache
memory and main memory and the interconnection
I am not showing point to point here unlike
in the previous diagram. Here we put it as
something in between the processor, memory
and the I/O. what is shown at the bottom is
the overall I/O subsystem and somehow they
are interconnected. We will elaborate on this
in subsequent lectures. But notice here that
there are two things involved here in I/O
subsystem; they are I/O devices or peripheral
devices and there are I/O controllers which
which sit between these devices and the rest
of the system.
So a controller may be taking care of multiple
devices also, that possibility is also shown
or a controller may take care of a single
device. So we need to understand what these
devices and controllers are and we will have
some discussion on this today. And in subsequent
lectures, we will talk of how these get connected.
So we will talk of buses and other means of
connection and then finally we talk of how
information goes across between processor
and these devices or memory and these devices.
We will focus now, our attention on devices
and controller.
You may often not see these as separate things.
Physically the controller and device may be
all packaged together. So roughly speaking,
the devices, the actual transducers so as
to say which converts information from one
form to another form; let us say printer,
you send in some bits and bytes, at some point
it is getting converted to a printed page
but there is something between the processor
and buses on one hand and this printing mechanism,
so there is a controller which takes care
of movement of data, sometimes even conversion
of data in some form.
So there is a wide variety of peripheral devices
and you can try to classify them in many different
ways. You could, for example, talk about their
behavior, in terms of whether say input that
means bring information into processor in
memory combination or take information out
or play both the roles so that is one aspect.
So, for example, keyboard is an input device,
printer is an output device.
The second issue is who is the partner, with
whom are these devices trying to communicate
or with whom do these devices make the processor
talk to. So it could be, at the other end
it could be a human programmer, human operator
or there could be another computer, another
machine or the devices could let a computer
talk to the environment in general, neither
a machine nor a human being or they could
be simply some kind of internal devices in
the sense that they only store information,
disk drive for example. So let me give examples
of each of these.
Keyboard is for example for human operators.
Network controller will connect one machine
to another machine. Environment that is typically
in an embedded domain, it could have a sensor
for example sensing temperature or computer
trying to turn a motor. Storage example is
a disk drive. then thirdly they could be characterized
in terms of the kind of data which is moving
across; in terms of either its speed, some
devices are very slow, some are very fast,
some could be medium; of course this is a
very coarse classification, one could specify
the rate of transfer; how many bytes or kilobytes
or megabytes getting the device is capable
of transferring. So, depending upon how you
organize the whole system that capability
may not be met, may be met or may not be met.
So suppose there is a disk which is capable
at the rate of 10 megabytes per second, you
have to organize the rest of your system so
that this capacity gets utilized; if you do
not the disk may be ready to transfer data
but you may not be willing to take it.
How much of data gets transferred in in one
chunk one quantum?
It could be serial data which means information
goes bit by bit as it goes on the network
for example or it could be parallel data which
may be 8 bits, 16 bits or it could be a larger
size.
How it is encoded?
You have 8 bits going, what does one group
of 8 bits mean; so different devices may have
different kind of encoding and a whole variety
exists there.
Now, before we talk of the current situation
let me recall how a set of peripheral devices
looked like years ago. So, this is the story
of this is the picture of IIT Delhi which
I am trying to paint when I was exposed to
the computer for the first time.
We had not in the present position there was
a computer called ICL 1909, it was somewhere
on E or F floor, and the only device, the
only peripheral device it had was a paper
tape reader and paper tape punch. The paper
tape basically was just a strip of paper with
holes punched on it and group of holes form
one character. So the way one could input
information was to use a device called............
a typewriter like device, a typewriter like
device called flexolator. Basically you could
think of a typewriter with a sort of spool
of tape attached at the end at one end. So
as you type, information could be punched
on that. And similarly, if you run a punch
tape it could also print out on a sheet of
paper.
So, as far as computer is concerned that directly
connected peripherals are paper tape reader
and paper tape punch. So everything everything
was in the form of paper tape; the operating
system, the compiler, the user program, the
library so everything had to be fed one by
one and finally you will your results will
be punched on paper tape so there was a larger
room roughly of this size and a smaller room.
In the smaller room there were a couple of
these, typewriter like devices, where you
could prepare your programs and when results
come back you print them out here.
So, of course, couple of years later the additional
devices were introduced, magnetic tape
where program and data could be stored for
repeated use and this was of course much faster.
The paper tape was just.... the reader was
a little faster; something like a 300 character
per second; punch was only some 10 or 20 character
per second and a line printer. Line printer
was something like I think 300 lines per minute
so it could print one line at a time so that
makes life a little easier. And subsequently
tape got replaced by punched cards and at
that time the system also moved to the current
location.
The punch card's added advantage is it has
like a deck of cards; each card basically
is one line of your program, one statement.
So if you want to add it, you could take out
a few cards and you put back new cards; you
could you could add it to your program like
that. You cannot change your card, but in
the deck you can take out a few cards and
you can insert a few cards. Of course, editing
in paper tape environment was much much more
painful, you will have to cut the tape and
patch it with something, another piece. So
obviously it is not surprising that people
some people could look at the tape and read
it, look at the punched holes.
Disk drive has again added convenience which
was of course much faster than tape; the tape
goes absolutely serially whereas in disk you
can reach a particular track and then of course
within a track you are sequential but you
can you do not have to go from one track to
another track by going through one track completely;
tape is entirely sequential. But the disk
capacity of those days were something like
20 megabytes was there, was a big figure obviously
because memory was 32 kilobytes or 64 kilobytes
of that order.
So now we fast forward into the modern era,
we have whole variety of peripherals and let
me list some of these with the kind of classification
we did earlier. Let us look at the peripherals
you have for introduction with human beings,
human programmer or operator. So the important
thing here is to.... I mean you are......
I am sure you are familiar with all the names
which I am putting here but it is important
at this stage to get a feel of what kind of
data rate or throughput rate these peripherals
are dealing with. Because when you design
a complete I/O system you should have a feel
of these things.
So devices like keyboard and mouse are limited
by the rate at which you can type so it is
only a fraction of kilobytes or basically
a few bytes per second is the rate at which
you can feed information and that is what
you require. I mean these are designs to take
care of let us say the fastest typist; we
do not need anything faster then.
You see here voice input and voice output
so the trend is towards developing these type
of capabilities; voice input voice output,
pictorial input pictorial output which are
which are more natural to human being and
instead of typing if you can talk to computer
it is definitely much nicer. But it is not
just an issue of peripheral device, it is
also a matter of the recognition of what is
being spoken or generating speech outputs
so we are not worrying about we are not concentrating
on that part but the voice input voice would
be digitized and brought to the computer in
the form of a sequence of samples.
So, of course the figures given here these
are from your textbooks; they are 0.02 bytes
per second or 20 bytes per second this is
a this is the rate of information if it is
compressed. But typically the speech inputs
are sampled at least at a rate of 10 kilo
ten samples per second and if each sample
is 1 byte so it has to be at least 10 kilobytes
per second and not 20 bytes per second.
Similarly, a plain speech would be again typically
sampled 8 to10 kilobytes per second.
If you are talking of music output, that is
sampled at a much higher rate, typically in
the range of 40 50 1000 samples per second
or so many kilobytes per second. Then another
input device is scanner which can basically
scan text or image whatever is there in a
sheet of paper and depending upon the size
and the resolution you may have different
data rates. So the resolution or the accuracy
here is measured in terms of Dots Per Inch
or DPI. You have scanners which are 600 DPI
or 1200 DPI and so on, that means how closely
or how finely the information can be resolved.
So if this DPI value is lower, then you will
get a very sort of coarse picture, if you
have higher DPI then you get a finer, shaper
picture.
We have variety of output devices, line printers,
laser printers, graphic display. As you would
notice graphic display is a highly demanding
peripheral. So it is shown as 60000 kilobytes
per second or 60 megabytes per second. Why
is it so high is because it again depends
upon the resolution. On a screen suppose roughly
speaking suppose you have 1000 by 1000 pixels,
it could be more accurately 1024 by 760 or
whatever, so as an approximate calculation
suppose you have 1 million pixels on the screen
then the question is how much how many bits
are required to represent each pixel so you
would have seen color settings, you could
use 1 byte to represent one pixel or you have
24 bits if you are going for a good quality
that means 3 bytes per pixel and 1 million
pixel means 3 million bytes is the information
present on screen and for a persistent display
this needs to be repeated refreshed so that
you get a persistent picture. So the rate
of repetition could be something like twenty-five
frames per second or fifty frames per second
or sixty frames per second. So if you if you
take that into account you realize that it
is a really very highly demanding peripheral
device.
Then you have a variety of peripheral devices
which connect a computer to a network of computers.
You have various kinds of modems; it could
be fax modem, cable modem; cable modem connects
to a TV, a video cable and fax modem connects
to a telephone line. ADSL modem also connects
to a telephone line which is capable of carrying
digital information.
On the other hand, you could have LAN adapters
Local Area Network adapters wired or wireless.
So these could be, these are typically input
and output both and the data rates for LANs
are higher depending upon whether it is a
10 megabytes per second LAN or 100 megabytes
per second LAN or 1 gigabytes per second LAN
so these are the three common standards. So
many bits per second can be easily translated
into bytes per second by reducing one order,
for example. Similarly, you have modems in
a wide variety of speeds. The slowest one
would be the fax modem which works on telephone
lines. So typically they are 56 kilobytes
56 kilobits per second. You might have heard
the term baud b a u d so baud is same thing
as bits per second; it is a unit used in terms
of communication specifying the rates in terms
of bits per second. So it is 56 kilo bauds
or 56 kilobits per second.
Example of peripheral devices for storage:
floppy disk drive, optical disk or CD ROM
and DVDs, their storage is in the form of
optical so on a disk you have portions which
are opaque which are portions which are transparent
and depending upon how finely you can resolve,
how closely you can space the opaque region
and transparent regions you will get more
bits per unit area and that will also determine
how much of data can be read out as the medium
moves.
Migrated tape: As I mentioned earlier they
are sequential devices, so information is
stored from one end to the other end in a
serial form and they are for archival purpose,
for long time storage where you are not reading
and writing very often. Among these disks
are the fastest.
As you can see that the data rates could be
varying from something like 100 kilobytes
per second of floppy drive to as much as 1000
kilobytes per second or 10 megabytes per second
of hard disk drive.
Another memory you can add to this list is
these flash memories or the pen drive USB
drive, the kind of which I am carrying in
my lecture and plugged into the USB port.
So these are...... there is no moving part
in that, it is basically a semiconductor memory
which is a non-volatile type called flash
memory and it is comparatively......................
what would be the transfer rate, I think it
will be, it is slightly slower than the migrated
disk, I do not have the exact figure. Let
us look little bit deeper into some of the
peripherals.
This is a schematic of the hard disk drive.
So basically it is a group of rotating media
or platters as they are called. This picture
shows two platters. So these are rotating
disks where on the surface, information is
stored. Each platter typically is capable
of storing information on both sides so you
can there is some coating which is magnetically
sensitive and it is on that 110 gets stored
in the form of polarity or magnetization.
So there is a there is a set of read write
heads which are mounted on an arm, here it
is pivoted on this and it rotates around this.
By that rotation this end which is the head
can move closer to the center or move towards
the circumference towards the peripheral.
You have circular tracks on the disk, very
closely spaced circular tracks and these heads
get positioned over some tracks. So all these
heads move together. so at one point of time
let us say all the heads are positioned on
track one of that particular platter then
you move them they could be positioned on
and all of them could be positioned on their
track number 10 and so on so there is a movement
of head from track to track and the movement
of disk is rotary.
So, if you need to reach some point in the
some some information which is stored somewhere
here, then this part has to rotate and come
in contact with the head, and head also effectively
this moves radially and it has to get positioned
at that point so once that is done by information
can be recorded or read out.
This is a real picture of a hard disk drive.
Here you can see four platters, so there will
be eight surfaces eight recording surfaces
and this is the head, this is the head assembly
and this is sorry this is the arm and the
head is actually close to the tip. So it is
pivoted around this point so this head will
move outside outwards or inwards, move from
track to track.
The rotational speeds are typically.......
they follow certain standards. so standards,
these keep changing, 3600 revolutions per
minute or from that it changes to 5400 revolutions
per minute or 7200 revolutions per minute,
this example is of 15.3000 or 15300 revolutions
per second. So, example of disk from Sea Gate.
So 15 point 15K is the highest speed you find
these days commercially and that decides how
fast you can get the data. It influences both
the time it takes to reach the initial point
and the time to transfer. When the disk is
rotating once you have positioned the head
at the right point the rate at which it rotates
determines the rate at which you are encountering
the data and that is the rate of transfer.
This has total capacity of 73 GB, speed is
15K rpm, seek time seek time is the time to
reach the desired point which could be more
or which could be small. So if you are already
close to that it may take less time;
if you are sitting far away then it might
take more time, so 3.6 is the average. The
interface also follows certain standards and
this standard is called ultra 320 SCSI. We
will talk of these little more a little more
later and this is characterized by the rate
of transfer 320 megabytes per second. So this
is the rate sustained by the interface that
interface, on that interface depending upon
what disk you connect and what speed it is
rotating at you will have actual data generated
at a different rate. So this is not an example
of a very high capacity disk but a fast disk.
The next one is somewhat slower, about half
the speed but capacity is larger, so capacity
I think it will come on the next part.
This is the 250 GB disk. 8 MB buffer size
which means the data does not go to the processor
or memory directly, it is first brought into
a local memory which is of size 8 megabytes
and then from there it is transferred. Average
latency: this is a rotational latency so the
time to reach a particular point is determined
by two things: One is the time taken for rotation
and second is time for the head to go from
one track to other track.
So the rotational latency in this case 4.2
millisecond and this can be determined if
you take the RPM figure how many revolutions
per minute, you can find the time for one
revolution and if you take half of that, that
is typically taken as average rotational latency.
So, if you do that arithmetic you will get
4.2 milliseconds.
The seek time is..... we can look at this
here; track to track seek time is moving from
one track to other track, is 2 milliseconds
and this is if you were to move one track
only. But on the average there is some average
figure given, it is for reading 8.9 milliseconds
and writing 10.9 milliseconds. So basically
the time involved to reach the first piece
of data is of the order of several milliseconds
and the time to transfer the data is much
faster; once you reach that point then things
are faster so this is the buffer to disk transfer
time, between the disk and the buffer; this
is 737 megabits per second maximum and this
is dependent upon the rotational speed. And
buffer to host here it is a different interface
EIDE. In the previous stage there was a different
interface, this takes 100 megabyte per second.
So here the peak of the transfer between disk
and buffer is much higher. But what happens
is that the disk is not always busy in transferring,
it is at times busy in seeking also. Therefore
this is this figure is okay. Initially one
might think that disk is faster but you have
arranged for a slower interface but disk cannot
be continuously seeking, it has to.......
disk is not continuously transferring, in-between
it spends time in moving to the desired point.
So, continuing with that there are some parameters
which define the storage layout. Cylinder
is actually collection of all the tracks which
are sort of coinciding on different platters.
So suppose you have four platters, you talk
of tenth track on each of these so all this
together will form what is called a cylinder.
often you do not write in terms of number
of tracks, you talk of total number of cylinders,
so this is one figure, the total number of
heads each surface is accessed by different
head, then number of each track is divided
into number of sectors. Well, this is from
the western digital internet site but I think
there is a there is something wrong in this
figure because it does not tell you the rest.
Number of sectors per track typically is few
hundreds, total capacity is 250 GB, bytes
per sector is 512 so half kilobyte in each
sector.
So this was one of the very important peripheral
devices that disk......... well, sometimes
it is not considered as an I/O device, some
people try to characterize this as a storage
device, okay but all the same. The way a processor
memory have to communicate with disk is something
like an I/O device. This is an example of
an LCD monitor from Samsung, different views
are shown. Interesting thing is this view
which shows how thin it is unlike, well, this
is also a LCD monitor. You have CRT monitors
which are very bulky. So you can get some
idea of the resolution here. The one shown
in the picture is of a really very high resolution.
1920 into 1200 and the viewing angle.........
so, often LCD monitors may have limited viewing
angle. If you look from side you may have
problem in seeing, but this one is 170 degrees
so you can say almost 180.
Then these are wireless LAN adapters. So this
one is the one which connects to a USB. USB
is a serial bus. We will talk of these terms
little more later. So this connects externally
to a computer.
This goes inside the computer this connects
to a PCI bus. Again we will talk of this term
little later.
Graphics card: so these are the cards which
drive really high resolution monitors and
they are able to show you changing scenes
very conveniently.
Scanners: Scanners are characterized by their
resolution in DPI as I mentioned, so 1200
DPI. The number of colors they can resolve
so 48-bit color, 6 bytes. These scanners have
a document feeder attached. That means you
can put a bunch of papers and they can automatically
feed like a photocopy machine if you have
seen one with a feeder.
Inkjet printers: They are very cost effective
and quality is continuously improving. They
are also characterized by the number of dots
per inch that is the printing quality and
the number of colors they can resolve. I do
not have those figures here. These have direct
network interface so that it can be shared
over in a network.
Well, I am sorry for this picture. Idea was
to show you a little close-up of an inkjet
printing mech- printer mechanism. If you can
see here in this one, the paper gets set from
the top and goes through this area where printing
gets done and then moves out here. There are
ink cartridges here and the key thing here
is this ink cartridge, the crucial mechanism
is sitting here, rest of it is essentially
paper feed and movement of this and interface.
But the new technological breakthrough which
has happened over last couple of years are
in this. So how very fine drops of ink can
be thrown on paper to form high quality pictures.
So they are typically....... there are two
techniques: one is piezoelectric mechanism
and other is thermal system. So, to understand
this, we will just quickly take a minute here.
So this is ink cartridge where in this chamber
there is ink and here is a plate quartz plate
which is piezoelectric and it can vibrate
and by that vibration it can throw out a very
tiny very fine finely controlled droplet of
ink. So this vibration would be controlled
by the electrical signal which it receives
depending upon the information which is to
be printed.
Another view is that again the same thing.
This is the chamber where ink and a droplet
gets thrown by this vibration of piezoelectric
disk here.
This is a thermal mechanism where the ink
droplet is thrown out by a small, by very
localized heating process. So here there is
a heating register which gets which gets heated
by electrical current and it is a very controlled
heating, so by heating there is an expansion
and the droplet goes out.
So, you remember that we are talking of resolution
of nature of 1200 dots per inch so you can
imagine the amount of the level of control
which you have to have on this ink droplet
to make it printed. So just contrast it with
the technology of peripheral devices which
existed couple of decades back and where we
are now. Okay let me close at this point.
Just to summarize, we talked of I/O performance
consideration, their importance. Briefly we
talked about throughput type of I/O, definition
I/O performance definition or response time
type of definition. We looked at variety of
peripherals, tried to classify from different
counts, tried to get a feel of their data
transfer rates and we looked at a few examples
of peripherals in little more detail. In the
next few lectures we will talk of how they
get interfaced to the rest of the system.
Thank you.