50 years of hard drives!

IBM / Hitachi honors half a century of the entire magnetic storage industry

A year and a half ago, the IT community, led by Intel, celebrated the 40th anniversary of the famous "Moore's Law", which foresaw the technological pace of development of the semiconductor industry for many years to come. integrated circuits (memory, microprocessors, etc.) and to this day remains one of the cornerstones for those who move the IT business.

But today we have no less, and maybe a more weighty reason for celebrating anniversaries: September 13, 1956, that is exactly half a century ago IBM introduced its first hard disk drive (later, these devices were given the semi-official nickname "Winchester").

And this invention over time actually created a huge industry. magnetic drives, without which no more or less powerful computing device is now unthinkable, since the overwhelming majority of information with which these same computing devices operate is stored precisely on magnetic media.

In fact, for more than half a century, hard drives have become the maina means of storing and promptly issuing information rapidly accumulated by mankind, defeating both tape drives, and optical media, and semiconductor EPROM / flash. And although the future of magnetic storage devices has been questioned more than once (due to various physical limitations that scientists each time managed to successfully overcome with the help of new effects and technologies), for the coming decades they will still remain in service due to the unique combinationsimportant consumer properties. And whether the magnetic disk drives will survive their centenary, we will find out very soon - after the next 50 years. ;)

And we remembered Moore's law for a reason - the development of the magnetic storage industry has been going alongside with the semiconductor industry all these years. There is much more in common between them than it might seem at first glance. Even the rate of miniaturization (increase in the density of elements), "predetermined" by Moore's law for microcircuits, is almost exactly the same for magnetic storage media. Suffice it to say that the minimum sizes of bits of magnetic recording on platters of hard drives are now (and over the past several years) the minimum sizes of elements of silicon transistors in the most modern microprocessors and memory.

It is interesting that the first microcircuit that predetermined another, the main branch of the development of the information and computer technology (ICT), appeared even later than the first "Winchester": it earned on September 12, 1958 at Texas Instruments (Jack Kilby and one of the founders of Intel, Robert Noyce, are rightfully considered the inventors of the microcircuit). By the way, for its invention in 2000 was awarded the Nobel Prize in physics, although there was not much physics as such when creating a microcircuit. It's just that Kilby and Noyce "just" came up with the technology that revolutionized the electronics industry. Unfortunately, for the invention of the hard drive "Nobel" has not yet been given. And they are unlikely to give ...

So, the first hard drive was 2 years old olderfirst microcircuit! (By the way, get ready - in exactly 2 years, on September 12, 2008, the industry will celebrate the 50th anniversary of microcircuits.;)) What was the first magnetic disk drive? Unlike a small microcircuit (a crystal that even then, in 1958, fit on one finger), the first hard drive was a huge cabinet, which contained a package of 50 huge plates with a diameter of 24 inches (more than 60 cm) each.

The disk was named RAMAC (Random Access Method of Accounting and Control) and was developed in the IBM laboratory in San Jose, California (which later became the heart of Silicon Valley). The plates of the disc were coated with a magnetic iron oxide "paint" similar to that used to build the world famous Golden Gate Bridge in San Francisco.

The information capacity of this giant was 5 MB (5 million bytes), which by today's concepts seems like a ridiculous figure, but then it was the High-End Enterprise segment. ;) Plates were mounted on a rotating spindle, and a mechanical bracket (one!) Contained the read and write heads and moved up and down on a vertical rod, and the time for the head to reach the desired magnetic track was less than one second.

As you can see this concept in many ways served as a prototype for all subsequent hard drives - rotating hard plates ("pancakes") with a magnetic coating, concentric recording tracks, quick access to any randomly selected track (see the name RAMAC). Only now a separate pair of read-write heads is used for each magnetic surface, and not common for the entire disk. Method quick access to an arbitrary place on the media (random access) made a real revolution in storage devices, since compared to the then dominant magnetic tapes, it allowed to dramatically increase access performance. One such RAMAC weighed almost a ton (971 kg) and was rented out at a price of $ 35,000 per year (then this was equal to the cost of 17 new cars)!

These days, Hitachi Global Storage Technologies, which is the legal successor of IBM's hard drive business and has absorbed its sophisticated hard drive development specialists, is boomingly celebrating the giant's 50th anniversary at DISKCON USA 2006 in Santa Clara, California.

The next landmark step of IBM in this field was the creation of the IBM 3340 drive. This "locker" was already smaller (about a meter high),

and at the time of its appearance in June 1973 was regarded as a scientific "miracle". With a magnetic recording density of 1.7 Mbit per square inch, it was equipped with small aerodynamic heads (that is, for the first time, the heads began to "float" over a rotating magnetic surface under the action of aerodynamic forces) and a sealed "box" ("can") in which plates heads. This protected the disks from dust and dirt and made it possible to drastically reduce the working distance between the head and the plate (the "flight" height), which led to a significant increase in the magnetic recording density. The IBM 3340 is rightfully considered the father of modern hard drives, as they are built on these principles. These drives had a non-removable capacity of 30 MB plus the same (30 MB) in the removable bay.

That gave the reason to call it "Winchester" - by analogy with the famous 30-30 Winchester rifle. By the way, progress has affected not only the design and recording density, but also the access time, which the developers managed to reduce to 25 milliseconds (compare this with 10-20 ms for modern much smaller hard drives)!

Later in 1973, IBM also released the world's first small-sized hard drive FHD50, based on the principles of the IBM 3340: magnetic plates with heads were enclosed in a completely closed case, and the heads did not move between the plates.

By the way, the very introduction of the principle "one magnetic surface - one pair of heads" (that is, the refusal to move the heads betweenplates) happened a little earlier: in 1971, IBM released the 3330-1 Merlin (named after the mythical medieval wizard), where it applied this principle. The same event also includes the first introduction of servo technology for positioning heads on plates, which was later transformed into TrueTrack Servo Technology from IBM (IBM alone has more than 40 patents on it). IN modern disks Servos are located at a distance of about 240 nm from each other and allow the head to be positioned on the track with an accuracy of 7 nanometers!

Interestingly, drives such as the IBM 3340 were intended for shared use, meaning that companies could rent space on that hard drive for $ 7.81 per megabyte per month. Therefore, there was also a need for small-sized individual drives.

In 1979, IBM introduced thin-film magnetic head technology. This made it possible to bring the density of magnetic recording to 7.9 million bits per square inch.

In 1982, Hitachi, Ltd. surprised the world with the first release of the H-8598 1 GB drive, that is, breaking a psychologically significant milestone.

This 1.2 GB drive consisted of ten 14-inch platters and two sets of read / write heads in a dual-actuator configuration. With a read speed of 3 MB per second (for comparison, in desktop hard drives, this speed was reached only about a decade later), the H-8598 model ran 87% faster than products of the previous generation. Six years later, Hitachi set the record again with a 1.89 GB drive using 8 9.5-inch drives. This model, the H-6586, was the first mainframe drive that a human could carry (weighed about 80 kg).

In the 80s of the last century, there were two more significant events for the magnetic storage industry. First, compact 5.25-inch drives were released that fit into the corresponding bays personal computers IBM PC (the first IBM 5100 Portable Computer was created in 1975, and for some time the products of this 51x0 line, and later the famous IBM PC 5150, used cassette drives). And then, in the late 1980s, the American company Conner Peripherals, founded in 1986 by Seagate co-founder Finis Conner, was the first in the world to market a 3.5-inch solenoid-powered hard disk drive. This opened a new era in the magnetic storage industry - this form factor has long been considered the main one for hard drives, and larger (in size) hard drives soon ceased to be produced as unpromising.

IBM has been a pioneer in the production of both miniature hard drives for notebooks and was the first in the world to introduce a one-inch hard drive in 1999 - the famous Microdirve.

Curiously, these subminiature discs used the same rotational speed of the platters (3600 rpm) as the giant models H-8598 and H-6586, but their capacity and speed were noticeably higher! And this progress has been achieved in just some 10-15 years! If we compare Microdrive with RAMAC, then 323 thousand "microdrives" will fit into the space of the latter, and their total capacity will be 2,500 terabytes! In 2005, Hitachi GST released the 10 millionth Microdrive. And the first 2.5-inch hard drive was released by IBM in 1991 and bore the name Tanba-1 (the appearance of the Travelstar line). It had a volume of 63 MB and weighed only 215 grams (3.5-inch disks of that time weighed 3 times more). Although the shock resistance of these wearable babies was useless by today's standards - 60 times less than modern counterparts.

By the way, Hitachi GST is still a confident first place in the world in the production of small-sized hard drives.

In the mid-90s of the last century, IBM proposed at least two more revolutionary technologies that are now used by all manufacturers of hard drives. First is magnetic heads on a giant magnetoresistive effect (the so-called GMR heads, which first appeared in the Deskstar 16GP series disks in 1997), which made it possible to dramatically increase the recording density (up to 2.7 Gbit / sq. inch) and in the next decade to increase the recording density sometimes even faster, than "according to Moore's law." :) I've talked about this more than once, so I won't repeat myself. And secondly, this is the so-called No-ID sector format ( new way formatting magnetic platters), which increases the density by another 10%. It is also now used by all manufacturers.

At about the same time, the rotational speed of the magnetic platters of 3.5-inch hard drives began to increase sharply - disks for PCs began to "chirp" together to 5400, and then to 7200 rpm. (the latter has been the standard for over a decade), and the Enterprise segment drives spun up to 10,000 and then to 15,000 rpm. By the way, not without it either, although Seagate believes that it was she who made the industry's first 15K model. ;) It is interesting, however, that it was Hitachi that was the first to increase the rotation speed above 10,000 - up to 12,000 rpm. in its 9.2GB model DK3F-1 released in 1998 and breaking performance records. It used new uniquely designed 2.5-inch diameter plates (later they became standard in 15K).

In 2003, IBM introduced the so-called, which are significantly smaller than before. This allowed the company, which has already become Hitachi GST, to release several new interesting series of disks. By the way, the size of the flight of modern heads above the surface of the plates is proportional to the flight of a giant airliner at an altitude of ... 1 millimeter above the ground!

The industry has celebrated the half-century anniversary of the hard disk with another remarkable achievement - for the first time in 50 years, drives have appeared that use a different principle of magnetic recording than was used in RAMAC. Namely - perpendicular magnetic recording (PMR), when the magnetic domains are oriented not along, but across a thin magnetic film on the surface of the plate. Hitachi GST demonstrated perpendicular magnetic recording back in April 2005 on samples with a recording density of 233 Gbps. inch. The transverse orientation of magnetic domains in a thin film (although somewhat thicker than for similar models with longitudinal recording) significantly increases the stability of information storage, which is necessary to overcome the consequences of the so-called superparamagnetic effect. True, not Hitachi or Toshiba, but Seagate became the first company to put it on sale in winter 2006. But Hitachi equipped those released in summer 2006 with the second generation of PMR technology. However, paying tribute to the time, we note that both longitudinal and perpendicular magnetic recording were considered for RAMAC, and then longitudinal magnetic recording was preferred, which determined the development of the industry for half a century! :)

In theory, PMR is capable of raising the density of magnetic recording to 500 Gbps per sq. inch (this is approximately 500 GB for the capacity of a 2.5-inch hard drive). Hitachi associates further plans for increasing the magnetic recording density with the so-called patterned media technology (when the film is initially "granulated" to the required recording density level), which will increase the capacity of the media by an order of magnitude. Next comes the turn of thermally activated magnetic recording with an estimated density limit of up to 15,000 Gbps per square meter. inch, which will extend the life of magnetic disk drives until 2020, or even longer.

According to research by scientists at the University of California at Berkeley, about 400,000 terabytes are now created every year. new information only at the expense email... A population of 6.3 billion people annually create 800 MB of information each, that is, about 5,000,000 terabytes of new data per year, 92% of which is stored on hard drives... This, of course, does not include information that is copied and replicated many times. Industry analysts predict an annual increase in hard drive sales from 409 million drives in 2006 to more than 650 million drives in 2010, or 12-15% annually.

Much of this growth will come from the burgeoning consumer electronics market, meaning the hard drive will soon become an indispensable feature of typical households. electronic devices... And demand, as you know, gives rise to supply. Therefore, there is no reason to doubt the prospects and viability of the magnetic disk storage industry in the foreseeable future.

The industry has come close to breaking another psychologically important milestone - this time 1 Terabyte for the capacity of a single drive (3.5-inch form factor). Who will be the first to release such a disc? Seagate, which has already released a 750 GB hard drive, and Hitachi came closest to this. Both of these manufacturers already sell hard drives in 160GB platters (up to 187.5 for Seagate). However, Hitachi has long mastered the five-platter design, while Seagate is still limited to the 4-platter design (and up to 250 GB platters are still far away). Therefore, Hitachi came closest to the terabyte device. What's more, Hitachi GST says they will begin shipping the 1TB drive later this year! When will this model be officially announced? Could it be the 50th anniversary of the Winchester? ..;)

We already use various kinds of computers without "hard drives": smartphones, tablets, laptops - any devices in which, instead of boxes with rotating plates inside, drives based on flash memory chips are installed. And despite the fact that, in terms of 1 GB, solid-state drives are not yet able to compete in price with classic HDDs, the outcome of this confrontation seems to be predetermined: high speed, low power consumption, high resistance to mechanical stress, miniature size - everything suggests that sooner or later SSD will finish off the mechanics.
To understand how we got there, let's take a look at how storage history has evolved over the past 50+ years.

IBM's first 350 hard drive Disk storage Unit was shown to the world on September 4, 1956. It was a huge cabinet 1.5 m wide, 1.7 m high, 0.74 m thick, weighed almost a ton and cost a fortune. Its spindle had 50 discs measuring 24 "(61 cm) covered with a paint containing a ferromagnetic material. D

the suits rotated at a speed of 1200 rpm, and the total amount of information stored on them was equal to the fantastic 4.4 MB for those times. The drive on which the heads were attached weighed almost 1.5 kg, but it took less than a second to move from the inner track of the upper disc to the inner track of the lower one. Imagine how fast this not at all light mechanism had to move.

Invented by a small group of engineers, the IBM 350 Disk Storage Unit was part of the computing system IBM 305 RAMAC. Such systems in the 50s and 60s were used exclusively in large corporations and government organizations. It is interesting that all the ideas embodied in the very first hard disk, which appeared in the era of lamp computers, have survived to this day: in modern drives, the same set of disks covered with a ferromagnetic layer, on which tracks with data are written, and a block of read and records posted on the "arm" with an electromechanical drive. By the way, the idea of \u200b\u200bheads that rise above the surface of the disk due to the air flow created by the rotation of the disks themselves was also proposed by IBM engineers, and this happened back in 1961. And almost until the end of the 60s everything about hard drives came from IBM in one way or another.

Disc race

In 1979, Alan Shugart, who previously worked at IBM and took part in the development of the IBM 350 Disk Storage Unit, announced the creation of Seagate Technology, and, perhaps, from that moment the history of the hard disk as a mass product began.
In the same 1979, Seagate created the first 5.25 ″ ST-506 5 MB drive, and a year later it went into production. A year later, the 10 MB model ST-412 was released. These discs were used in the legendary personal iBM computers PC / AT and IBM PC / XT.
Western digital, which later became the main competitor of Seagate, was founded nine years earlier and at the time of its foundation was called General Digital Corporation (it was renamed in 1971, a year after its foundation). She was engaged in the production of single-chip controllers and various electronics. The first controller for hard drives Seagate ST-506 / ST-412 on a single chip was made by Western Digital in 1981, and it was called WD1010. For the next seven years, WD participated in the joint development of the ATA standard, was engaged in the development of chips for SCSI and ATA disks, and in 1988 it acquired the disk division of Tandon Corporation and in 1990 introduced its own hard drives of the Caviar series. More details about this technique are discussed on the electronics forum - http://www.tehnari.ru/f30/.
In general, in the 20-year interval, from 1985 to 2005, there was a real boom in disk production, and a huge number of companies appeared, most of which by now have either become part of the major giants Seagate and Western Digital, or simply ceased to exist. Just remember the once well-known disk brands - Conner, Fuji, IBM, Quantum, Maxtor, Fujitsu, Hitachi, Toshiba, which have established themselves as manufacturers of good technology. All of them, one way or another, took part in the “disk race”, which started from the moment when the HDD became an integral part of the personal computer.

Parallel universe

Almost from the very beginning, computers have used several different types of memory, but only because a perfect memory device has not yet been invented. If we imagine that we managed to get chips that work as fast as random access memory, non-volatile like flash, but with a large rewriting resource and as large as modern hard drives, then we would not need to divide this memory into separate devices ... Each of the currently existing types of storage devices is imperfect, and in connection with the total miniaturization, hard drives are especially imperfect due to their mechanical nature. They emerged from the idea of \u200b\u200brelatively inexpensively obtaining a large amount of memory, and therefore, initially, requirements for other parameters, such as, for example, speed and reliability, somehow faded into the background. Therefore, it is not surprising that they have always been looking for an alternative to HDD.
Back in the 70s and 80s, attempts were made to create solid state drives (Solid State Drive, SSD) based on dynamic memory, which were equipped with a special controller and rechargeable battery in case of power outage. Back then, these were almost insane projects that cost a lot of money, and they were implemented exclusively in supercomputers (IBM, Cray) and in systems used for processing data in real time (for example, at seismic stations). Later when the volumes of the chips random access memory increased significantly and their cost decreased, such drives appeared as solutions for personal computers (for example, the well-known i-RAM manufactured by Gigabyte), but still remained the lot of geeks, and did not receive mass distribution due to the relative high cost and small volume.

Another direction of SSD was born from the idea of \u200b\u200bcreating a large volume electrically rewritable read only memory (EEPROM) chip. The problem was that the recorded cells can be placed on the crystal quite tightly, but if you need not only to write, but also to erase, and then write again, then you need a chain responsible for erasing, which greatly increases the size of the memory cell.
A way out of the situation in the early 80s was found by a scientist who worked at Toshiba - Dr. Fujio Masuoka. He suggested crossing two ways to erase cells permanent memory, and instead of clearing the entire chip or, on the contrary, only one cell, erase the memory in large enough blocks. In 1984, Masuoka presented his development at the IEEE 1984 International Electron Devices Meeting (IEDM), and in 1989 at the International Solid-State Circuits Conference, Toshiba showed the developed concept nAND Flash... Then, even in the wildest dreams, hardly anyone would have thought that a small-volume chip with a complex data access scheme would be able to compete with hard drives that were already gaining momentum.
Founded in the same 1989, the Israeli company M-Systems was the first to work on the idea of \u200b\u200ba flash drive, and in 1995 released DiskOnChip - a drive on a single chip. It contained both flash memory and a controller. Moreover, this 8.16 and 32 MB single-chip disk even then contained algorithms for monitoring cell wear and detecting and redistributing damaged blocks in its firmware. By the way, it is M-Systems in 1999 that will be the first to release USB flash drives - DiskOnKey, and IBM will sign a contract with the company and sell them in the US under its own brand.
But for flash-based SSDs to become mainstream, it took another 10 years or so. In 2006, Samsung, by then the largest manufacturer of memory chips, released the world's first laptop with a 32GB SSD. Within two years of Apple showed MacBook Air, in which an optional SSD could be installed, and in 2010 this laptop began to be produced exclusively with solid state drives.
Modern SSDs certainly have drawbacks. Although, if you look closely, there are not so many of them: but by and large there is only one - the high cost of 1 GB in comparison with classic hard drives. But the semiconductor industry is developing very quickly, new types of memory are being developed, controllers' algorithms are being improved, volumes are rapidly increasing, and the cost is gradually decreasing. But that's not all.
There is another important argument that creates strong competition and prices quickly become attractive because of the ease of manufacture of solid state drives. Basically, assembling an SSD is like assembling just a controller board for a hard drive, and you only need an assembly line of surface mount boards. This is, of course, very simplistic, but generally true. Assembling a classic hard drive is much more complicated and therefore expensive. That is why no one has any doubts that until the moment when SSDs begin to actively displace "hard drives", there is very little left. The process has already begun.