HARD DISK DRIVES

A hard disk drive (HDD), commonly referred to as a hard drive, hard disk or fixed disk drive,[1] is a non-volatile storage device which stores digitally encoded data on rapidly rotating platters with magnetic surfaces. Strictly speaking, "drive" refers to a device distinct from its medium, such as a tape drive and its tape, or a floppy disk drive and its floppy disk. Early HDDs had removable media; however, an HDD today is typically a sealed unit with fixed media.[2]

HDDs were originally developed for use with computers. In the 21st century, applications for HDDs have expanded beyond computers to include digital video recorders, digital audio players, personal digital assistants, digital cameras and video game consoles. In 2005 the first mobile phones to include HDDs were introduced by Samsung and Nokia.[3] The need for large-scale, reliable storage, independent of a particular device, led to the introduction of configurations such as RAID arrays, network attached storage (NAS) systems and storage area network (SAN) systems that provide efficient and reliable access to large volumes of data.

Technology

HDDs record data by magnetizing a ferromagnetic material directionally, to represent either a 0 or a 1 binary digit. They read the data back by detecting the magnetization of the material. A typical HDD design consists of a spindle which holds one or more flat circular disks called platters, onto which the data is recorded. The platters are made from a non-magnetic material, usually glass or aluminum, and are coated with a thin layer of magnetic material. Older disks used iron(III) oxide as the magnetic material, but current disks use a cobalt-based alloy.


A hard disk drive with the disks and spindle motor hub removed. In the center, the internal structure of the spindle motor can be seen. To the left of center is the actuator arm with a read-write head under the tip of its very end (near center); the orange wires along the side of the arm are part of the path the signals take to and from the read-write head. The flexible, somewhat 'U'-shaped, ribbon cable barely visible below and to the left of the actuator arm is another part of its path connecting the head to the controller board on the opposite side.
A cross section of the magnetic surface in action. In this case the binary data encoded using frequency modulation:The platters are spun at very high speeds. Information is written to a platter as it rotates past mechanisms called read-and-write heads that operate very close over the magnetic surface. The read-and-write head is used to detect and modify the magnetization of the material immediately under it. There is one head for each magnetic platter surface on the spindle, mounted on a common arm. An actuator arm (or access arm) moves the heads on an arc (roughly radially) across the platters as they spin, allowing each head to access almost the entire surface of the platter as it spins. The arm is moved using a voice coil actuator or (in older designs) a stepper motor.

The magnetic surface of each platter is divided into many small sub-micrometre-sized magnetic regions, each of which is used to encode a single binary unit of information. In today's HDDs each of these magnetic regions is composed of a few hundred magnetic grains. Each magnetic region forms a magnetic dipole which generates a highly localized magnetic field nearby. The write head magnetizes a magnetic region by generating a strong local magnetic field nearby. Early HDDs used an electromagnet both to generate this field and to read the data by using electromagnetic induction. Later versions of inductive heads included metal in Gap (MIG) heads and thin film heads. In today's heads, the read and write elements are separate but in close proximity on the head portion of an actuator arm. The read element is typically magneto-resistive while the write element is typically thin-film inductive.[4]

In modern drives, the small size of the magnetic regions creates the danger that their magnetic state be lost because of thermal effects. To counter this, the platters are coated with two parallel magnetic layers, separated by a 3-atom-thick layer of the non-magnetic element ruthenium, and the two layers are magnetized in opposite orientation, thus reinforcing each other.[5] Another technology used to overcome thermal effects to allow greater recording densities is perpendicular recording, which has been used in some hard drives as of 2006.

Hard disk drives are sealed to prevent dust and other sources of contamination from interfering with the operation of the hard disks heads. The hard drives are not air tight, but rather utilize an extremely fine air filter, to allow for air inside the hard drive enclosure. The spinning of the disks causes the air to circulate forcing any particulates to become trapped on the filter. The same air currents also act as a gas bearing which enables the heads to float on a cushion of air above the surfaces of the disks.

Hard drives are precise devices, moving at very high speed, and a number of analogies have been made to try to describe this. One states:

“ As an analogy, a magnetic head slider flying over a disk surface with a flying height of 25 nm with a relative speed of 20 meters/second is equivalent to an aircraft flying at a physical spacing of 0.2 µm at 900 kilometers/hour. This is what a disk drive experiences during its operation.


Capacity and access speed

Using rigid disks and sealing the unit allows much tighter tolerances than in a floppy disk drive. Consequently, hard disk drives can store much more data than floppy disk drives and can access and transmit it faster. In 2007, a typical “enterprise”, i.e. workstation HDD, might store between 160 GB and 1 TB of data (as of local US market by July 2007), rotate at 7,200 or 10,000 revolutions per minute (RPM) and have a media transfer rate of over 1 Gbit/s or higher.[6] The fastest “enterprise” HDDs spin at 15,000 rpm, and can achieve sequential media transfer speeds above 1.6 Gbit/s.[7] Mobile, i.e., laptop HDDs, which are physically smaller than their desktop and enterprise counterparts, tend to be slower and have less capacity. In the 1990s, most spun at 4,200 rpm.[8] In 2007, a typical mobile HDD spins at 5,400 rpm, with 7,200 rpm models available for a slight price premium.

The exponential increases in disk space and data access speeds of HDDs have enabled the commercial viability of consumer products that require large storage capacities, such as digital video recorders and digital audio players.[9] In addition, the availability of vast amounts of cheap storage has made viable a variety of web-based services with extraordinary capacity requirements, such as free-of-charge web search and email (Google, Yahoo!, etc.).

The main way to decrease access time is to increase rotational speed, while the main way to increase throughput and storage capacity is to increase areal density. A vice president of Seagate Technology projects a future growth in disk density of 40% per year.[10] Access times have not kept up with throughput increases, which themselves have not kept up with growth in storage capacity.

As of 2006, some disk drives use perpendicular recording technology to increase recording density and throughput.[11]

The first 3.5" HDD marketed as able to store 1 TB was the Hitachi Deskstar 7K1000. It contains five platters at approximately 200 GB each, providing 935.5 GiB of usable space.[12] Hitachi has since been joined by Samsung (Samsung SpinPoint F1, which has 3 × 334 GB platters), Seagate and Western Digital in the 1 TB drive market.[13][14]

Form factor Width Largest capacity Platters (Max)

5.25" FH 146 mm 47 GB[15] (1998) 14
5.25" HH 146 mm 19.3 GB[16] (1998) 4[17]
3.5" 102 mm 1 TB[12] (2007) 5
2.5" 69.9 mm 320 GB[18] (2007) 3
1.8" (PCMCIA) 54 mm 160 GB[19] (2007)
1.8" (ATA-7 LIF) 53.8 mm


Capacity measurements

A disassembled and labeled 1997 hard drive.The capacity of an HDD can be calculated by multiplying the number of cylinders by the number of heads by the number of sectors by the number of bytes/sector (most commonly 512). Drives with ATA interface bigger and more than eight gigabytes behave as if they were structured into 16383 cylinders, 16 heads, and 63 sectors, for compatibility with older operating systems. Unlike in the 1980s, the cylinder, head, sector counts reported to the CPU by a modern ATA drive are no longer actual physical parameters since the reported numbers are constrained by historic operating-system interfaces and with zone bit recording the actual number of sectors varies by zone. Disks with SCSI interface address each sector with a unique integer number; the operating system remains ignorant of their head or cylinder count.

Hard disk drive manufacturers specify disk capacity using the SI prefixes mega-, giga- and tera-, and their abbreviations M, G and T. Byte is typically abbreviated B.

Some operating-system tools report capacity using the same abbreviations but actually use binary prefixes. For instance, the prefix mega-, which normally means 106 (1,000,000), in the context of data storage can mean 220 (1,048,576), which is nearly 5% more. Similar usage has been applied to prefixes of greater magnitude. This results in a discrepancy between the disk manufacturer's stated capacity and the apparent capacity of the drive when examined through some operating-system tools. The difference becomes with 7% even more noticeable for a gigabyte. For example, Microsoft Windows reports disk capacity both in decimal-based units to 12 or more significant digits and with binary-based units to three significant digits. Thus a disk specified by a disk manufacturer as a 30 GB disk might have its capacity reported by Windows 2000 both as "30,065,098,568 bytes" and "28.0 GB". The disk manufacturer used the SI definition of "giga", 109 to arrive at 30 GB; however, because the utilities provided by Windows define a gigabyte as 1,073,741,824 bytes (230 bytes, often referred to as a gibibyte, or GiB), the operating system reports capacity of the disk drive as (only) 28.0 GB.


Form factors

5¼" full height 110 MB HDD,
2½" 8.5 mm 6495 MB HDD,
US/UK pennies for comparisonThe earliest “form factor” hard disk drives inherited their dimensions from floppy-disk drives (FDDs), so that either could be mounted in chassis slots, and thus the HDD form factors became colloquially named after the corresponding FDD types. "Form factor" compatibility continued after the 3½ in size even though floppy disk drives with new smaller dimensions ceased to be offered.

"8 inch" drive: (9.5 in x 4.624 in x 14.25 in = 241.3 mm x 117.5 mm x 362 mm)
In 1979, Shugart Associates' SA1000 was the first form factor compatible HDD, having the same dimensions and a compatible interface to the 8" FDD. Both "full height" and "half height" (2.313 in) versions were available.
"5¼ inch" drive: (5.75 in x 1.63 in x 8 in = 146.1 mm x 41.4 mm x 203 mm)
This smaller form factor, first used in an HDD by Seagate in 1980, was the same size as full height 5¼-inch diameter FDD, i.e., 3.25 inches high. This is twice as high as commonly used today; i.e., 1.63 in = 41.4 mm (“half height”). Most desktop models of drives for optical 120 mm disks (DVD, CD) use the half height 5¼" dimension, but it fell out of fashion for HDDs. The Quantum “Bigfoot” HDD was the last to use it in the late 1990s, with “low-profile” (~25 mm) and “ultra-low-profile” (~20 mm) high versions.
"3½ inch" drive: (4 in x 1 in x 5.75 in = 101.6 mm x 25.4 mm x 146 mm)
This smaller form factor, first used in an HDD by Rodime in 1984, was the same size as the "half height" 3½ FDD, 1.e., 1.63 inches high. Today has been largely superseded by 1-inch high “slimline” or “low-profile” versions of this form factor which is used by most desktop HDDs.
"2½ inch" drive: ((2.75 in x 0.374 in x 3.945 in = 69.85 mm x 9.5 mm x 100 mm)
This smaller form factor was introduced by PrairieTek in 1988; there is no corresponding FDD. It is widely used today for hard-disk drives in mobile devices (laptops, music players, etc.). Today, the dominant height of this form factor is 9.5 mm, but there were also 19 mm, 17 mm, and 12.5 mm high variants in use.
"1.8 inch" drive: (54 mm × 8 mm × 71 mm)
This form factor, originally introduced by Integral Peripherals in 1993, has evolved into the ATA-7 LIF with dimensions as stated. It is increasingly used in digital audio players and subnotebooks. An original variant exists for 2–5 GB sized HDDs that fit directly into a PC card expansion slot.
"1 inch" drive: (42.8 mm × 5 mm × 36.4 mm)
This form factor was introduced in 1999 as IBM's Microdrive to fit inside a CF Type II slot.
"0.85 inch" drive: (24 mm × 5 mm × 32 mm)
Toshiba announced this form factor in January 2004[20] for use in mobile phones and similar applications, including SD/MMC slot compatible HDDs optimized for video storage on 4G handsets. Toshiba currently sells a 4 GB (MK4001MTD) and 8 GB (MK8003MTD) version[3] and holds the Guinness World Record for the smallest harddisk drive.[21]
Major manufacturers discontinued the development of new products for the 1-inch and 0.85 inch form factors in 2007, due to falling prices of flash memory.[22]

The inch-based nickname of all these form factors usually do not indicate any actual product dimension (which are for more recent form factors specified in millimeters), but just roughly indicate a size relative to disk diameters, in the interest of historic continuity.


Other characteristics

Capacity of a hard disk drive is usually quoted in gigabytes. Older HDDs quoted their smaller capacities in megabytes.

The data transfer rate at the inner zone ranges from 44.2 MB/s to 74.5 MB/s, while the transfer rate at the outer zone ranges from 74.0 MB/s to 111.4 MB/s. An HDD's random access time ranges from 5 ms to 15 ms.

Integrity

An IBM HDD head resting on a disk platter. Since the drive is not in operation, the head is simply pressed against the disk by the suspension.
Close-up of a hard disk head resting on a disk platter, and its suspension. A reflection of the head and suspension are visible beneath on the mirror-like disk.Due to the extremely close spacing between the heads and the disk surface, any contamination of the read-write heads or platters can lead to a head crash — a failure of the disk in which the head scrapes across the platter surface, often grinding away the thin magnetic film and causing data loss. Head crashes can be caused by electronic failure, a sudden power failure, physical shock, wear and tear, corrosion, or poorly manufactured platters and heads.

The HDD's spindle system relies on air pressure inside the enclosure to support the heads at their proper flying height while the disk rotates. An HDD requires a certain range of air pressures in order to operate properly. The connection to the external environment and pressure occurs through a small hole in the enclosure (about 0.5 mm in diameter), usually with a carbon filter on the inside (the breather filter, see below). If the air pressure is too low, then there is not enough lift for the flying head, so the head gets too close to the disk, and there is a risk of head crashes and data loss. Specially manufactured sealed and pressurized disks are needed for reliable high-altitude operation, above about 10,000 feet (3,000 m). Note that modern commercial aircraft have a pressurized cabin, whose pressure altitude does not normally exceed 8,500 feet - thus, ordinary hard drives can safely be used in flight. Modern disks include temperature sensors and adjust their operation to the operating environment. Breather holes can be seen on all disks — they usually have a sticker next to them, warning the user not to cover the holes. The air inside the operating disk is constantly moving too, being swept in motion by friction with the spinning platters. This air passes through an internal recirculation (or "recirc") filter to remove any leftover contaminants from manufacture, any particles or chemicals that may have somehow entered the enclosure, and any particles or outgassing generated internally in normal operation. Very high humidity for extended periods can corrode the heads and platters.

For giant magnetoresistive (GMR) heads in particular, a minor head crash from contamination (that does not remove the magnetic surface of the disk) still results in the head temporarily overheating, due to friction with the disk surface, and can render the data unreadable for a short period until the head temperature stabilizes (so called "thermal asperity," a problem which can partially be dealt with by proper electronic filtering of the read signal).

The hard disk's electronics control the movement of the actuator and the rotation of the disk, and perform reads and writes on demand from the disk controller. Modern disk firmware is capable of scheduling reads and writes efficiently on the platter surfaces and remapping sectors of the media which have failed.


Disk failures and their metrics

Most major hard disk and motherboard vendors now support self-monitoring, analysis and reporting technology (S.M.A.R.T.), which attempts to alert users to impending failures.

However, not all failures are predictable. Normal use eventually can lead to a breakdown in the inherently fragile device, which makes it essential for the user to periodically back up the data onto a separate storage device. Failure to do so can lead to the loss of data. While it may be possible to recover lost information, it is normally an extremely costly procedure, and it is not possible to guarantee success. A 2007 study published by Google suggested very little correlation between failure rates and either high temperature or activity level.[26] While several S.M.A.R.T. parameters have an impact on failure probability, a large fraction of failed drives do not produce predictive S.M.A.R.T. parameters.[26] S.M.A.R.T. parameters alone may not be useful for predicting individual drive failures.[26]

SCSI, SAS and FC drives are typically more expensive and are traditionally used in servers and disk arrays, whereas inexpensive ATA and SATA drives evolved in the home computer market and were perceived to be less reliable. This distinction is now becoming blurred.

The mean time between failures (MTBF) of SATA drives is usually about 600,000 hours (some drives such as Western Digital Raptor have rated 1.2 million hours MTBF), while SCSI drives are rated for upwards of 1.5 million hours.[citation needed] However, independent research indicates that MTBF is not a reliable estimate of a drive's longevity.[27] MTBF is conducted in laboratory environments in test chambers and is an important metric to determine the quality of a disk drive before it enters high volume production. Once the drive product is in production, the more valid metric is annualized failure rate (AFR). AFR is the percentage of real-world drive failures after shipping.

SAS drives are comparable to SCSI drives, with high MTBF and high reliability.

Enterprise SATA drives designed and produced for enterprise markets, unlike standard SATA drives, have reliability comparable to other enterprise class drives.

Typically enterprise drives (all enterprise drives, including SCSI, SAS, enterprise SATA and FC) experience between .70%-.78% annual failure rates from the total installed drives.

HISTOY OF COMPUTER

It is difficult to identify any one device as the earliest computer, partly because the term "computer" has been subject to varying interpretations over time.

Originally, the term "computer" referred to a person who performed numerical calculations (a human computer), often with the aid of a mechanical calculating device. Examples of early mechanical computing devices included the abacus, the slide rule and arguably the astrolabe and the Antikythera mechanism (which dates from about 150-100 BC). The end of the Middle Ages saw a re-invigoration of European mathematics and engineering, and Wilhelm Schickard's 1623 device was the first of a number of mechanical calculators constructed by European engineers.

However, none of those devices fit the modern definition of a computer because they could not be programmed. In 1801, Joseph Marie Jacquard made an improvement to the textile loom that used a series of punched paper cards as a template to allow his loom to weave intricate patterns automatically. The resulting Jacquard loom was an important step in the development of computers because the use of punched cards to define woven patterns can be viewed as an early, albeit limited, form of programmability.

In 1837, Charles Babbage was the first to conceptualize and design a fully programmable mechanical computer that he called "The Analytical Engine".[3] Due to limited finance, and an inability to resist tinkering with the design, Babbage never actually built his Analytical Engine.

Large-scale automated data processing of punched cards was performed for the U.S. Census in 1890 by tabulating machines designed by Herman Hollerith and manufactured by the Computing Tabulating Recording Corporation, which later became IBM. By the end of the 19th century a number of technologies that would later prove useful in the realization of practical computers had begun to appear: the punched card, Boolean algebra, the vacuum tube (thermionic valve) and the teleprinter.

During the first half of the 20th century, many scientific computing needs were met by increasingly sophisticated analog computers, which used a direct mechanical or electrical model of the problem as a basis for computation. However, these were not programmable and generally lacked the versatility and accuracy of modern digital computers.


A succession of steadily more powerful and flexible computing devices were constructed in the 1930s and 1940s, gradually adding the key features that are seen in modern computers. The use of digital electronics (largely invented by Claude Shannon in 1937) and more flexible programmability were vitally important steps, but defining one point along this road as "the first digital electronic computer" is difficult (Shannon 1940). Notable achievements include:


EDSAC was one of the first computers to implement the stored program (von Neumann) architecture.Konrad Zuse's electromechanical "Z machines". The Z3 (1941) was the first working machine featuring binary arithmetic, including floating point arithmetic and a measure of programmability. In 1998 the Z3 was proved to be Turing complete, therefore being the world's first operational computer.
The non-programmable Atanasoff–Berry Computer (1941) which used vacuum tube based computation, binary numbers, and regenerative capacitor memory.
The secret British Colossus computer (1944), which had limited programmability but demonstrated that a device using thousands of tubes could be reasonably reliable and electronically reprogrammable. It was used for breaking German wartime codes.
The Harvard Mark I (1944), a large-scale electromechanical computer with limited programmability.
The U.S. Army's Ballistics Research Laboratory ENIAC (1946), which used decimal arithmetic and is sometimes called the first general purpose electronic computer (since Konrad Zuse's Z3 of 1941 used electromagnets instead of electronics). Initially, however, ENIAC had an inflexible architecture which essentially required rewiring to change its programming.
Several developers of ENIAC, recognizing its flaws, came up with a far more flexible and elegant design, which came to be known as the stored program architecture or von Neumann architecture. This design was first formally described by John von Neumann in the paper "First Draft of a Report on the EDVAC", published in 1945. A number of projects to develop computers based on the stored program architecture commenced around this time, the first of these being completed in Great Britain. The first to be demonstrated working was the Manchester Small-Scale Experimental Machine (SSEM) or "Baby". However, the EDSAC, completed a year after SSEM, was perhaps the first practical implementation of the stored program design. Shortly thereafter, the machine originally described by von Neumann's paper—EDVAC—was completed but did not see full-time use for an additional two years.

Nearly all modern computers implement some form of the stored program architecture, making it the single trait by which the word "computer" is now defined. By this standard, many earlier devices would no longer be called computers by today's definition, but are usually referred to as such in their historical context. While the technologies used in computers have changed dramatically since the first electronic, general-purpose computers of the 1940s, most still use the von Neumann architecture. The design made the universal computer a practical reality.


Microprocessors are miniaturized devices that often implement stored program CPUs.Vacuum tube-based computers were in use throughout the 1950s, but were largely replaced in the 1960s by transistor-based devices, which were smaller, faster, cheaper, used less power and were more reliable. These factors allowed computers to be produced on an unprecedented commercial scale. By the 1970s, the adoption of integrated circuit technology and the subsequent creation of microprocessors such as the Intel 4004 caused another leap in size, speed, cost and reliability. By the 1980s, computers had become sufficiently small and cheap to replace simple mechanical controls in domestic appliances such as washing machines. Around the same time, computers became widely accessible for personal use by individuals in the form of home computers and the now ubiquitous personal computer. In conjunction with the widespread growth of the Internet since the 1990s, personal computers are becoming as common as the television and the telephone and almost all modern electronic devices contain a computer of some kind.


Stored program architecture

The defining feature of modern computers which distinguishes them from all other machines is that they can be programmed. That is to say that a list of instructions (the program) can be given to the computer and it will store them and carry them out at some time in the future.

In most cases, computer instructions are simple: add one number to another, move some data from one location to another, send a message to some external device, etc. These instructions are read from the computer's memory and are generally carried out (executed) in the order they were given. However, there are usually specialized instructions to tell the computer to jump ahead or backwards to some other place in the program and to carry on executing from there. These are called "jump" instructions (or branches). Furthermore, jump instructions may be made to happen conditionally so that different sequences of instructions may be used depending on the result of some previous calculation or some external event. Many computers directly support subroutines by providing a type of jump that "remembers" the location it jumped from and another instruction to return to the instruction following that jump instruction.

Program execution might be likened to reading a book. While a person will normally read each word and line in sequence, they may at times jump back to an earlier place in the text or skip sections that are not of interest. Similarly, a computer may sometimes go back and repeat the instructions in some section of the program over and over again until some internal condition is met. This is called the flow of control within the program and it is what allows the computer to perform tasks repeatedly without human intervention.

Comparatively, a person using a pocket calculator can perform a basic arithmetic operation such as adding two numbers with just a few button presses. But to add together all of the numbers from 1 to 1,000 would take thousands of button presses and a lot of time—with a near certainty of making a mistake. On the other hand, a computer may be programmed to do this with just a few simple instructions. For example:

mov #0,sum ; set sum to 0
mov #1,num ; set num to 1
loop: add num,sum ; add num to sum
add #1,num ; add 1 to num
cmp num,#1000 ; compare num to 1000
ble loop ; if num <= 1000, go back to 'loop'
halt ; end of program. stop running
Once told to run this program, the computer will perform the repetitive addition task without further human intervention. It will almost never make a mistake and a modern PC can complete the task in about a millionth of a second.[4]

However, computers cannot "think" for themselves in the sense that they only solve problems in exactly the way they are programmed to. An intelligent human faced with the above addition task might soon realize that instead of actually adding up all the numbers one can simply use the equation


and arrive at the correct answer (500,500) with little work.[5] In other words, a computer programmed to add up the numbers one by one as in the example above would do exactly that without regard to efficiency or alternative solutions.

PROGRAMS

In practical terms, a computer program might include anywhere from a dozen instructions to many millions of instructions for something like a word processor or a web browser. A typical modern computer can execute billions of instructions every second and nearly never make a mistake over years of operation.

Large computer programs may take teams of computer programmers years to write and the probability of the entire program having been written completely in the manner intended is unlikely. Errors in computer programs are called bugs. Sometimes bugs are benign and do not affect the usefulness of the program, in other cases they might cause the program to completely fail (crash), in yet other cases there may be subtle problems. Sometimes otherwise benign bugs may be used for malicious intent, creating a security exploit. Bugs are usually not the fault of the computer. Since computers merely execute the instructions they are given, bugs are nearly always the result of programmer error or an oversight made in the program's design.[6]

In most computers, individual instructions are stored as machine code with each instruction being given a unique number (its operation code or opcode for short). The command to add two numbers together would have one opcode, the command to multiply them would have a different opcode and so on. The simplest computers are able to perform any of a handful of different instructions, the more complex computers have several hundred to choose from—each with a unique numerical code. Since the computer's memory is able to store numbers, it can also store the instruction codes. This leads to the important fact that entire programs (which are just lists of instructions) can be represented as lists of numbers and can themselves be manipulated inside the computer just as if they were numeric data. The fundamental concept of storing programs in the computer's memory alongside the data they operate on is the crux of the von Neumann, or stored program, architecture. In some cases, a computer might store some or all of its program in memory that is kept separate from the data it operates on. This is called the Harvard architecture after the Harvard Mark I computer. Modern von Neumann computers display some traits of the Harvard architecture in their designs, such as in CPU caches.

While it is possible to write computer programs as long lists of numbers (machine language) and this technique was used with many early computers,[7] it is extremely tedious to do so in practice, especially for complicated programs. Instead, each basic instruction can be given a short name that is indicative of its function and easy to remember—a mnemonic such as ADD, SUB, MULT or JUMP. These mnemonics are collectively known as a computer's assembly language. Converting programs written in assembly language into something the computer can actually understand (machine language) is usually done by a computer program called an assembler. Machine languages and the assembly languages that represent them (collectively termed low-level programming languages) tend to be unique to a particular type of computer. For instance, an ARM architecture computer (such as may be found in a PDA or a hand-held videogame) cannot understand the machine language of an Intel Pentium or the AMD Athlon 64 computer that might be in a PC.[8]

Though considerably easier than in machine language, writing long programs in assembly language is often difficult and error prone. Therefore, most complicated programs are written in more abstract high-level programming languages that are able to express the needs of the computer programmer more conveniently (and thereby help reduce programmer error). High level languages are usually "compiled" into machine language (or sometimes into assembly language and then into machine language) using another computer program called a compiler.[9] Since high level languages are more abstract than assembly language, it is possible to use different compilers to translate the same high level language program into the machine language of many different types of computer. This is part of the means by which software like video games may be made available for different computer architectures such as personal computers and various video game consoles.

The task of developing large software systems is an immense intellectual effort. It has proven, historically, to be very difficult to produce software with an acceptably high reliability, on a predictable schedule and budget. The academic and professional discipline of software engineering concentrates specifically on this problem.

EXAMPLE
Suppose a computer is being employed to drive a traffic light. A simple stored program might say:

Turn off all of the lights
Turn on the red light
Wait for sixty seconds
Turn off the red light
Turn on the green light
Wait for sixty seconds
Turn off the green light
Turn on the yellow light
Wait for two seconds
Turn off the yellow light
Jump to instruction number (2)
With this set of instructions, the computer would cycle the light continually through red, green, yellow and back to red again until told to stop running the program.

However, suppose there is a simple on/off switch connected to the computer that is intended be used to make the light flash red while some maintenance operation is being performed. The program might then instruct the computer to:

Turn off all of the lights
Turn on the red light
Wait for sixty seconds
Turn off the red light
Turn on the green light
Wait for sixty seconds
Turn off the green light
Turn on the yellow light
Wait for two seconds
Turn off the yellow light
If the maintenance switch is NOT turned on then jump to instruction number 2
Turn on the red light
Wait for one second
Turn off the red light
Wait for one second
Jump to instruction number 11
In this manner, the computer is either running the instructions from number (2) to (11) over and over or its running the instructions from (11) down to (16) over and over, depending on the position of the switch.[10]


How computers work
Main articles: Central processing unit and Microprocessor
A general purpose computer has four main sections: the arithmetic and logic unit (ALU), the control unit, the memory, and the input and output devices (collectively termed I/O). These parts are interconnected by busses, often made of groups of wires.

The control unit, ALU, registers, and basic I/O (and often other hardware closely linked with these) are collectively known as a central processing unit (CPU). Early CPUs were composed of many separate components but since the mid-1970s CPUs have typically been constructed on a single integrated circuit called a microprocessor.

Control unit

The ALU is capable of performing two classes of operations: arithmetic and logic.

The set of arithmetic operations that a particular ALU supports may be limited to adding and subtracting or might include multiplying or dividing, trigonometry functions (sine, cosine, etc) and square roots. Some can only operate on whole numbers (integers) whilst others use floating point to represent real numbers—albeit with limited precision. However, any computer that is capable of performing just the simplest operations can be programmed to break down the more complex operations into simple steps that it can perform. Therefore, any computer can be programmed to perform any arithmetic operation—although it will take more time to do so if its ALU does not directly support the operation. An ALU may also compare numbers and return boolean truth values (true or false) depending on whether one is equal to, greater than or less than the other ("is 64 greater than 65?").

Logic operations involve Boolean logic: AND, OR, XOR and NOT. These can be useful both for creating complicated conditional statements and processing boolean logic.

Superscalar computers contain multiple ALUs so that they can process several instructions at the same time. Graphics processors and computers with SIMD and MIMD features often provide ALUs that can perform arithmetic on vectors and matrices.

MEMORY

A computer's memory can be viewed as a list of cells into which numbers can be placed or read. Each cell has a numbered "address" and can store a single number. The computer can be instructed to "put the number 123 into the cell numbered 1357" or to "add the number that is in cell 1357 to the number that is in cell 2468 and put the answer into cell 1595". The information stored in memory may represent practically anything. Letters, numbers, even computer instructions can be placed into memory with equal ease. Since the CPU does not differentiate between different types of information, it is up to the software to give significance to what the memory sees as nothing but a series of numbers.

In almost all modern computers, each memory cell is set up to store binary numbers in groups of eight bits (called a byte). Each byte is able to represent 256 different numbers; either from 0 to 255 or -128 to +127. To store larger numbers, several consecutive bytes may be used (typically, two, four or eight). When negative numbers are required, they are usually stored in two's complement notation. Other arrangements are possible, but are usually not seen outside of specialized applications or historical contexts. A computer can store any kind of information in memory as long as it can be somehow represented in numerical form. Modern computers have billions or even trillions of bytes of memory.

The CPU contains a special set of memory cells called registers that can be read and written to much more rapidly than the main memory area. There are typically between two and one hundred registers depending on the type of CPU. Registers are used for the most frequently needed data items to avoid having to access main memory every time data is needed. Since data is constantly being worked on, reducing the need to access main memory (which is often slow compared to the ALU and control units) greatly increases the computer's speed.

Computer main memory comes in two principal varieties: random access memory or RAM and read-only memory or ROM. RAM can be read and written to anytime the CPU commands it, but ROM is pre-loaded with data and software that never changes, so the CPU can only read from it. ROM is typically used to store the computer's initial start-up instructions. In general, the contents of RAM is erased when the power to the computer is turned off while ROM retains its data indefinitely. In a PC, the ROM contains a specialized program called the BIOS that orchestrates loading the computer's operating system from the hard disk drive into RAM whenever the computer is turned on or reset. In embedded computers, which frequently do not have disk drives, all of the software required to perform the task may be stored in ROM. Software that is stored in ROM is often called firmware because it is notionally more like hardware than software. Flash memory blurs the distinction between ROM and RAM by retaining data when turned off but being rewritable like RAM. However, flash memory is typically much slower than conventional ROM and RAM so its use is restricted to applications where high speeds are not required.[13]

In more sophisticated computers there may be one or more RAM cache memories which are slower than registers but faster than main memory. Generally computers with this sort of cache are designed to move frequently needed data into the cache automatically, often without the need for any intervention on the programmer's part.

Input/output (I/O)

I/O is the means by which a computer receives information from the outside world and sends results back. Devices that provide input or output to the computer are called peripherals. On a typical personal computer, peripherals include input devices like the keyboard and mouse, and output devices such as the display and printer. Hard disk drives, floppy disk drives and optical disc drives serve as both input and output devices. Computer networking is another form of I/O.

Often, I/O devices are complex computers in their own right with their own CPU and memory. A graphics processing unit might contain fifty or more tiny computers that perform the calculations necessary to display 3D graphics. Modern desktop computers contain many smaller computers that assist the main CPU in performing I/O.

Multitasking

While a computer may be viewed as running one gigantic program stored in its main memory, in some systems it is necessary to give the appearance of running several programs simultaneously. This is achieved by having the computer switch rapidly between running each program in turn. One means by which this is done is with a special signal called an interrupt which can periodically cause the computer to stop executing instructions where it was and do something else instead. By remembering where it was executing prior to the interrupt, the computer can return to that task later. If several programs are running "at the same time", then the interrupt generator might be causing several hundred interrupts per second, causing a program switch each time. Since modern computers typically execute instructions several orders of magnitude faster than human perception, it may appear that many programs are running at the same time even though only one is ever executing in any given instant. This method of multitasking is sometimes termed "time-sharing" since each program is allocated a "slice" of time in turn.

Before the era of cheap computers, the principle use for multitasking was to allow many people to share the same computer.

Seemingly, multitasking would cause a computer that is switching between several programs to run more slowly - in direct proportion to the number of programs it is running. However, most programs spend much of their time waiting for slow input/output devices to complete their tasks. If a program is waiting for the user to click on the mouse or press a key on the keyboard, then it will not take a "time slice" until the event it is waiting for has occurred. This frees up time for other programs to execute so that many programs may be run at the same time without unacceptable speed loss

Multiprocessing
Some computers may divide their work between one or more separate CPUs, creating a multiprocessing configuration. Traditionally, this technique was utilized only in large and powerful computers such as supercomputers, mainframe computers and servers. However, multiprocessor and multi-core (multiple CPUs on a single integrated circuit) personal and laptop computers have become widely available and are beginning to see increased usage in lower-end markets as a result.

Supercomputers in particular often have highly unique architectures that differ significantly from the basic stored-program architecture and from general purpose computers.[14] They often feature thousands of CPUs, customized high-speed interconnects, and specialized computing hardware. Such designs tend to be useful only for specialized tasks due to the large scale of program organization required to successfully utilize most of a the available resources at once. Supercomputers usually see usage in large-scale simulation, graphics rendering, and cryptography applications, as well as with other so-called "embarrassingly parallel" tasks.

Networking and the Internet

Computers have been used to coordinate information in multiple locations since the 1950s, with the U.S. military's SAGE system the first large-scale example of such a system, which led to a number of special-purpose commercial systems like Sabre.

In the 1970s, computer engineers at research institutions throughout the United States began to link their computers together using telecommunications technology. This effort was funded by ARPA (now DARPA), and the computer network that it produced was called the ARPANET. The technologies that made the Arpanet possible spread and evolved. In time, the network spread beyond academic and military institutions and became known as the Internet. The emergence of networking involved a redefinition of the nature and boundaries of the computer. Computer operating systems and applications were modified to include the ability to define and access the resources of other computers on the network, such as peripheral devices, stored information, and the like, as extensions of the resources of an individual computer. Initially these facilities were available primarily to people working in high-tech environments, but in the 1990s the spread of applications like e-mail and the World Wide Web, combined with the development of cheap, fast networking technologies like Ethernet and ADSL saw computer networking become almost ubiquitous. In fact, the number of computers that are networked is growing phenomenally. A very large proportion of personal computers regularly connect to the Internet to communicate and receive information. "Wireless" networking, often utilizing mobile phone networks, has meant networking is becoming increasingly ubiquitous even in mobile computing environments