Hardware vs. SoftwareGreat Example: Data Compression

This lecture has a theme:– Hardware/Devices are getting bigger, meaner,

and faster, but– Sometimes clever software can make even more

of a difference.– Designing clever software is often much more

cost effective and turn around time is faster.– We’ll look at one clever piece of software that

beat the hardware

Storage Units

One character is represented using 8 bitsExample: ‘A’ = 1010 0001

8 bits = 1 byte Kilobyte (KB) 1,000 bytes Megabyte (MB) 1,000,000 bytes Gigabyte (GB) 1,000,000,000 bytes Terabyte (TB) 1,000,000,000,000 bytes (trillion) Message: Just know the differences in magnitude

Storage Unit (Examples)

A high resolution image file (jpg format) is 500 KB– The highest quality images can be as large as 10 MB

Your average 100 page word document would take up about 2 or 3 MB

– My 143 page Ph.D thesis takes up 6.4 MB (includes embedded pictures)

Walmart’s central database is approximately 2 TB (terabytes) or so they say…

Message: Just be aware of the striking difference between 2TB and 500KB. One is insignificant compared to the other

Storage Unit (Examples)

A 4 minute song (mp3 format) is about 5 MB In an uncompressed format (wav), it would be about 60 MB (4

minute song) Message: Quality is the same. They seem close in size, but

imagine waiting 5 minutes vs. 1 hour to download.

CD quality songs 50 minutes (600MB). A 22 minute “South Park” video (mpg format) is 40 MB (very

poor quality) A higher quality video (mpg format) can take up as much as 10

MB per minute. A 2.5 hour movie (highest quality video) 4000 MB (4 GB); Can’t

fit on a regular CD

Storage Unit (Examples)

CD quality songs 50 minutes (600MB). A 22 minute “South Park” video (mpg format) is 40

MB (very poor quality) A higher quality video (mpg format) can take up as

much as 10 MB per minute. A 2.5 hour movie (highest quality video) 4000 MB (4

GB); Can’t fit on a regular CD Message: DVD’s emerged out of necessity to store

video. If there was a way to compress it on a CD…?

Storage Devices

3.5 Floppy (soon to be obsolete) 1.4 MB Zip disk: 250MB + CD ROM 700MB+ DVD 5GB + Hard disks 80GB+

Special disk racks that can hold up to 500GB in a standard tower case.

Message: There is a correspondence between the size of data and where its stored.

The Real Problem: Transmission not Storage

Storage devices have kept up with the demand to store larger and larger databases.

and to store larger and larger media formats The real problem is transmitting these large files

over an Internet that has severe limitations. The solution comes in two forms:1. Compression algorithms for files that can not be

split.2. Fancy software and new protocols for delivering

only the parts of the data that need to be sent.

Types of ComputersHistorically, there are have been 5 different types of computers1. Supercomputers (once a hot area, now 90% of the

companies are out of business)– Custom multiple processors, complex architecture, Cray/SGI was

the big player2. Mainframe (still going strong, go figure?)

– Highly fault tolerant, supports multiple users, never shuts off3. Minicomputers are really just smaller mainframes4. Microcomputers are PC’s and laptops

– Generic components, cheap (in many ways), prone to crash5. Workstations are PC’s but with expensive components

– The buzz in the early 90’s, much better memory architecture than PC’s (Sun, SGI, HP, DEC, were the players), 1 year, 244 days and counting….

Message: Nowadays, there is very little difference between the average home computer and the average computer used in business.

Processors

1. 8086 Intel Processor (1980) 5 MHz 8 bit2. 80286 Intel Processor (1984) 8 MHz

20 time faster 16 bit3. 386 Processor (1988) 20 MHz 32 bit4. 486 Processor (1991) 66 MHz 32 bit5. Pentium (would have been 586) 200 MHz 64 bit

(bought one in 1994 for $2400)6. Pentium4 (2002) 2 GHz $15007. Pentium4 (2003) 3.2 GHz $1000 Message: There are a lot of messages here.

Processors

Is Intel the only manufacturer of processors? No! IBM probably sells more processors but they are for other

devices: cell phones, lab instruments, cars, etc. Other processor manufacturers:

– Apple (Mac)– Motorola– AMD– Sun Microsystems– Patriot Scientific

Message: Computing is not just about PC’s

Processor Speed

1 MHz means that the processor cycles (i.e., processed an instruction 1,000,000 times per second). However, this is a very basic instruction.

Adding two numbers sometimes can require as many as 4 or 5 basic instructions

In theory, a 2 GHz processor should be able to perform 2 billion instructions in just one second.

In practice, it’ll do around 400 to 500 million Why?

Memory

Two kinds of memory: Storage and RAM Storage: Hard Disk, CDRW, Zip disk, floppy RAM (Random Access Memory): Memory chips

used to store working programs and data. RAM is much much faster than storage Faster means the stored data can be delivered to

the processor in less time.Message: Memory is currently the processing

bottleneck. The memory can not deliver instructions to the processor fast enough.

Miniaturization

The key to making processors and memory faster is to make them smaller

Smaller things are closer together. Thus, it takes less time for the signal to travel from one component to another.

Also, if your “wires” are very small you can connect thousands or millions of wires in a small space (bandwidth)

In about 6-7 years, we will reach a critical limit where it will be very very hard to make things smaller.

Currently, the big buzz is nanocell computers; the processor and its connections are built with units that are as small as a molecule (several atoms).

Where are things going

Mobile computing Computers in everything: beepers, phones, cars,

even the fridge How about a mini computer in your credit card?

– Its already in development

Computers will be part of everything– X10– Part of a new house, just like plumbing

Theme

Even though processors are getting so fast and so cheap, we can NOT count on this to continue indefinitely.

There are two main issues:1. We might hit a physical brick wall in terms of physical

limitations on miniaturization

2. There is a need to put computer-like processors in devices that need to be very small very cheap

Theme

Thus, the industry is eventually going to have to look into clever ways of making programs– more efficient– more reliable

Thus, there is a need for better– software engineering– more efficient integration of hardware and

software.

Homework

Skim through pages 38-60. but concentrate on 41-53 (read these pages

more carefully) Read 110-114. Next, we are going to study one detailed

example of compression.