Hubbry Logo
History of IBM magnetic disk drivesHistory of IBM magnetic disk drivesMain
Open search
History of IBM magnetic disk drives
Community hub
History of IBM magnetic disk drives
logo
7 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
History of IBM magnetic disk drives
History of IBM magnetic disk drives
from Wikipedia

IBM manufactured magnetic disk storage devices from 1956 to 2003, when it sold its hard disk drive business to Hitachi.[1][2] Both the hard disk drive (HDD) and floppy disk drive (FDD) were invented by IBM and as such IBM's employees were responsible for many of the innovations in these products and their technologies.[3] The basic mechanical arrangement of hard disk drives has not changed since the IBM 1301. Disk drive performance and characteristics are measured by the same standards now as they were in the 1950s. Few products in history have enjoyed such spectacular declines in cost and physical size along with equally dramatic improvements in capacity and performance.

IBM manufactured 8-inch floppy disk drives from 1969 until the mid-1980s, but did not become a significant manufacturer of smaller-sized, 5.25- or 3.5-inch floppy disk drives (the dimension refers to the diameter of the floppy disk, not the size of the drive).[4] IBM always offered its magnetic disk drives for sale but did not offer them with original equipment manufacturer (OEM) terms until 1981.[5] By 1996, IBM had stopped making hard disk drives unique to its systems and was offering all its HDDs as an OEM.[6][7]

IBM uses many terms to describe its various magnetic disk drives, such as direct-access storage device (DASD),[a] disk file and diskette file. Here, the current industry standard terms, hard disk drive (HDD) and floppy disk drive (FDD), are used.

Early IBM HDDs

[edit]

IBM 350

[edit]
IBM 305 at U.S. Army Red River Arsenal, with two IBM 350 disk drives in the foreground
RAMAC mechanism at Computer History Museum

The IBM 350 disk storage unit, the first disk drive, was announced by IBM as a component of the IBM 305 RAMAC computer system on September 14, 1956.[8][9][10][11] Simultaneously a very similar product, the IBM 355, was announced for the IBM 650 RAMAC computer system. RAMAC stood for "Random Access Method of Accounting and Control". The first engineering prototype 350 disk storage shipped to Zellerbach Paper Company, San Francisco, in June 1956,[12] with production shipment beginning in November 1957 with the shipment of a unit to United Airlines in Denver, Colorado.[13]

Its design was motivated by the need for real time accounting in business.[14] The 350 stores 5 million 6-bit characters (3.75 MB).[15] It has fifty-two 24-inch (610 mm) diameter disks of which 100 recording surfaces are used, omitting the top surface of the top disk and the bottom surface of the bottom disk. Each surface has 100 tracks. The disks spin at 1200 rpm. Data transfer rate is 8,800 characters per second. An access mechanism moves a pair of heads up and down to select a disk pair (one down surface and one up surface) and in and out to select a recording track of a surface pair. Several improved models were added in the 1950s. The IBM RAMAC 305 system with 350 disk storage leased for $3,200 per month. The 350 was officially withdrawn in 1969.

U.S. patent 3,503,060 from the RAMAC program is generally considered to be the fundamental patent for disk drives.[16] This first-ever disk drive was initially cancelled by the IBM Board of Directors because of its threat to the IBM punch card business but the IBM San Jose laboratory continued development until the project was approved by IBM's president.[17]

The 350's cabinet is 60 inches (150 cm) long, 68 inches (170 cm) high and 29 inches (74 cm) wide.

The RAMAC unit weighs about one ton, has to be moved around with forklifts, and was frequently transported via large cargo airplanes.[18] According to Currie Munce, research vice president for Hitachi Global Storage Technologies (which acquired IBM's storage business), the storage capacity of the drive could have been increased beyond five million characters, but IBM's marketing department at that time was against a larger capacity drive, because they didn't know how to sell a product with more storage. Nonetheless, double capacity versions of the 350 were announced[8] in January 1959 and shipped later the same year.

In 1984, the RAMAC 350 Disk File was designated an International Historic Landmark by The American Society of Mechanical Engineers.[19] In 2002 at the Magnetic Disk Heritage Center, a team[20] led by Al Hoagland began restoration of an IBM 350 RAMAC in collaboration with Santa Clara University.[21] In 2005, the RAMAC restoration project relocated to the Computer History Museum, Mountain View, California and is now demonstrated to the public in the museum's Revolution exhibition.[22]

IBM 353

[edit]

The IBM 353, used on the IBM 7030, was similar to the IBM 1301, but with a faster transfer rate. It has a capacity of 2,097,152 (221) 64-bit words or 134,217,728 (227) bits and transferred 125,000 words per second.[23] A prototype unit shipped in late 1960 was the first disk drive to use one head per surface flying on a layer of compressed air as in the older head design of the IBM 350 disk storage (RAMAC). Production 353s used self-flying heads essentially the same as those of the 1301.

IBM 355

[edit]

The IBM 355 was announced on September 14, 1956, as an addition to the popular IBM 650.[24] It used the mechanism of the IBM 350 with up to three access arms[b] and stored 6 million decimal digits and 600,000 signs.[24] It transferred a full track to and from the magnetic core memory of the 653, an IBM 650 option that included just sixty signed 10-digit words, enough for a single track of disk or a tape record, along with two unrelated features.

IBM 1405

[edit]

The IBM 1405 Disk Storage Unit was announced in 1961 and was designed for use with the IBM 1400 series, medium scale business computers.[25] The 1405 Model 1 has a storage capacity of 10 million alphanumeric characters (60,000,000 bits) on 25 disks. Model 2 has a storage capacity of 20 million alphanumeric characters (120,000,000 bits) on 50 disks. In both models the disks are stacked vertically on a shaft rotating at 1200 rpm.

Each side of each disk has 200 tracks divided into five sectors. Sectors 0–4 are on the top surface and 5–9 are on the bottom surface. Each sector holds either 178 or 200 characters. One to three forked-shaped access arms each contains two read/write heads, one for the top of the disk and the other for the bottom of the same disk. The access arms are mounted on a carriage alongside the disk array. During a seek operation an access arm moved, under electronic control, vertically to seek a disk 0–49 and then horizontally to seek a track 0–199. Ten sectors are available at each track. It takes about 10 ms to read or write a sector.

The access time ranges from 100ms to a maximum access time for model 2 of 800ms and 700ms for model 1. The 1405 model 2 disk storage unit has 100,000 sectors containing either 200 characters in move mode or 178 characters in load mode, which adds a word mark bit to each character. The Model 1 contains 50,000 sectors.[26]

IBM 7300

[edit]

The IBM 7300 Disk Storage Unit was designed for use with the IBM 7070; IBM announced a model 2 in 1959, but when IBM announced the 1301 on June 5, 1961, 7070 and 7074 customers found it to be more attractive than the 7300. The 7300 uses the same technology as the IBM 350, IBM 355 and IBM 1405

IBM 1301

[edit]

The IBM 1301 Disk Storage Unit was announced on June 2, 1961[27][28] with two models. It was designed for use with the IBM 7000 series mainframe computers and the IBM 1410. The 1301 stores 28 million characters (168,000,000 bits) per module (25 million characters with the 1410). Each module has 25 large disks and 40[c] user recording surfaces, with 250 tracks per surface. The 1301 Model 1 has one module, the Model 2 has two modules, stacked vertically. The disks spin at 1800 rpm. Data is transferred at 90,000 characters per second.

A major advance over the IBM 350 and IBM 1405 is the use of a separate arm and head for each recording surface, with all the arms moving in and out together like a big comb. This eliminates the time needed for the arm to pull the head out of one disk and move up or down to a new disk. Seeking the desired track is also faster since, with the new design, the head will usually be somewhere in the middle of the disk, not starting on the outer edge. Maximum access time is reduced to 180 milliseconds.

The 1301 is the first disk drive to use heads that are aerodynamically designed to fly over the surface of the disk on a thin layer of air.[3] This allows them to be much closer to the recording surface, which greatly improves performance.

The 1301 connects to the computer via the IBM 7631 File Control. Different models of the 7631 allow the 1301 to be used with a 1410 or 7000 series computer, or shared between two such computers.[29]

The IBM 1301 Model 1 leased for $2,100 per month or could be purchased for $115,500. Prices for the Model 2 were $3,500 per month or $185,000 to purchase. The IBM 7631 controller cost an additional $1,185 per month or $56,000 to purchase. All models were withdrawn in 1970.[27]

IBM 1302

[edit]

The IBM 1302 Disk Storage Unit was introduced in September 1963.[30] Improved recording quadrupled its capacity over that of the 1301, to 117 million 6-bit characters per module. Average access time is 165 ms and data can be transferred at 180 K characters/second, more than double the speed of the 1301. There are two access mechanisms per module, one for the inner 250 cylinders and the other for the outer 250 cylinders.[31] As with the 1301, there is a Model 2 which doubles the capacity by stacking two modules. The IBM 1302 Model 1 leased for $5,600 per month or could be purchased for $252,000. Prices for the Model 2 were $7,900 per month or $355,500 to purchase. The IBM 7631 controller cost an additional $1,185 per month or $56,000 to purchase. The 1302 was withdrawn in February 1965.

IBM 1311

[edit]
IBM 1311 Disk Drives – Model 2 (slave) & Model 3 (master)
IBM 1311 disk drive with IBM 1316 removable disk pack at the Computer History Museum

The IBM 1311 Disk Storage Drive was announced on October 11, 1962, and was designed for use with several medium-scale business and scientific computers.[32] The 1311 is about the size and shape of a top-loading washing machine and stores 2 million characters (12,000,000 bits) (or, in so-called "Load Mode" on an IBM 1401, a sector can hold 90 7-bit characters, or 12,600,000 bits total [33]) on a removable IBM 1316 disk pack.[34] Seven models of the 1311 were introduced during the 1960s. They were withdrawn during the early 1970s.

Each IBM 1316 Disk Pack is 4 inches (100 mm) high, weighs 10 pounds (4.5 kg) and contains six 14-inch (360 mm) diameter disks, yielding 10 recording surfaces (the outer surfaces are not used). The 10 individual read/write heads are mounted on a common actuator within the disk drive which moves in and out hydraulically and is mechanically detented at the desired track before reading or writing occurs. The disks spin at 1500 rpm. Each recording surface has 100 tracks with 20 sectors per track. Each sector stores 100 characters. The disk pack is covered with a clear plastic shell and a bottom cover when not in use. A lifting handle in the top center of the cover is rotated to release the bottom cover. Then the top of the 1311 drive is opened and the plastic shell lowered into the disk-drive opening (assuming it is empty). The handle is turned again to lock the disks in place and release the plastic shell, which is then removed and the drive cover closed. The process is reversed to remove a disk pack. The same methods are used for many later disk packs.

There are seven models of the 1311 disk drive. The first drive attached to a system is a "master drive" which contains the controller and can control a number of Model 2 "slave drives."[34]

  1. Master drive on an IBM 1440, IBM 1460, or IBM 1240 system and can control up to four Model 2 drives. Introduced October 11, 1962. Withdrawn February 8, 1971.
  2. Slave drive to a master drive. Can have any special feature incorporated that the master drive has incorporated. Introduced October 11, 1962. Withdrawn January 6, 1975.
  3. Master drive on IBM 1620 or IBM 1710 system and can control up to three Model 2 drives. Does not support any special features. Introduced October 11, 1962. Withdrawn May 12, 1971.
  4. Master drive on an IBM 1401 system and can control up to four Model 2 drives. Introduced October 11, 1962. Withdrawn February 8, 1971.
  5. Master drive on an IBM 1410, IBM 7010, or IBM 7740 system and can control up to four Model 2 drives. Direct Seek comes as standard on this model. Introduced January 7, 1963. Withdrawn May 12, 1971.
  6. No information available, probably a master drive. Introduced March 5, 1968. Withdrawn February 2, 1971.[32]
  7. No information available, probably a slave drive to Model 6. Introduced March 5, 1968. Withdrawn February 2, 1971.[32]

The optional special features are:

  • Direct Seek: Without this option every seek returns to track zero first.
  • Scan Disk: Automatic rapid search for identifier or condition.
  • Seek Overlap: Allows a seek to overlap a single read or write, and any number of other seeks, when multiple drives are in use.
  • Track Record: Increases the capacity of the disk by writing a single large record per track instead of using 20 separate sectors. A track can hold 2,980 characters in 6-bit 'Move Mode", and 2,682 7-bit characters in "Load Mode", giving the drive a total capacity of 17,880,000 bits in 6-bit mode, and 18,774,000 bits in 7-bit mode.[35]

The master drives, Models 1, 3, 4, and 5, which contain extra power supplies and the control logic, are about a foot wider than the Model 2 slave drive.

IBM System/360 and other IBM mainframe HDDs

[edit]

IBM 2302

[edit]

The IBM 2302 is the System/360 version of the 1302, with track formatting in accordance with S/360 DASD architecture rather than 7000 series architecture.

It uses a non-removable module of 25 platters, of which 46 surfaces are used for recording. The 2302 Model 3 contains one module and the Model 4 two. There are two independent access mechanisms per module, one for the innermost 250 cylinders, and one for the outermost 250, tracks available to each access mechanism are called an access group. The access mechanism provides one read/write head per track. Average rotational delay is 17 milliseconds (msec), and maximum is 34 msec. Maximum seek time per access group is 180 msec. The track size is 4985 bytes; with formatting information and alternate tracks, module capacity is stated as 112 MB. The 2302 attaches to IBM mainframes via a IBM 2841 Storage Control Unit.[36]

IBM 2305

[edit]

The IBM 2305 fixed head storage (a fixed-head disk drive sometimes incorrectly called a drum) and associated IBM 2835 Storage Control[37] were announced in 1970,[38] initially to connect to the 360/85 and 360/195 using the IBM 2880 Block Multiplexor Channel.

The 2305 Drive was in much demand when the System 370 offered Virtual Storage, and these 2305s were often used for paging devices. They were used in this way on 3155, 3165, 3158, 3168, 3033, 4341, and 3081 (with special feature microcode.) The 2305 was also used for high activity small data sets such as catalogs and job queues.

The 2305-1 has a capacity of 5.4 MB and runs at 3.0 MB/second when attached using the 2-byte channel interface. Average access time is 2.5 ms. The larger 2305-2 has a capacity of 11.2 MB and runs at 1.5 MB/second with an average access time of 5 ms.[38]

The 2305 provides large-scale IBM computers with fast, continuous access to small-sized quantities of information. Its capacity and high data rate make it ideal for some systems residence functions, work files, job queues, indices and data sets that are used repeatedly.[38] Its fast response time makes it attractive as a paging device in a heavily loaded systems, where there are 1.5 or more transactions per second.[39]

IBM 2311

[edit]
IBM 2311 Disk Storage Drive, with its six platters

The IBM 2311 Disk Storage Drive was introduced with the 2841 Control Unit in 1964 for use throughout the IBM System/360; the combination was also available on the IBM 1130 and the IBM 1800. The drive also directly attaches to the IBM System/360 Model 20 and the IBM System/360 Model 25. All drives used the IBM 1316 Disk Pack introduced with the IBM 1311.

The 2311 Model 1 attaches to most IBM mainframes through a 2841 Control Unit; it attaches to the System/360 Model 25 thru a Disk Attachment Control which provides the function of the control unit. Disk packs are written in these attachments in IBM's count key data variable record length format.

The 2311 models 11 or 12 are used when attached to an integrated control of the System/360 Model 20 and the disk packs are written with a fixed sector format. The disk packs are not interchangeable between those written on the Model 1 and those written on the Models 11 or 12.

The 2311 mechanism is largely identical to the 1311, but recording improvements allow higher data density. The 2311 stores 7.25 megabytes on a single removable IBM 1316 disk pack (the same type used on the IBM 1311) consisting of six platters that rotate as a single unit. The 2311 has ten individual read/write (R/W) heads mounted on a common actuator which moves in and out hydraulically and is mechanically detented at the desired track before reading or writing occurred. Each recording surface has 200 tracks plus three optional tracks which can be used as alternatives in case faulty tracks are discovered. Average seek time is 85 ms. Data transfer rate is 156 kB/s.

Because the 2311 was to be used with a wide variety of computers within the 360 product line, its electrical interconnection was standardized. This created an opportunity for other manufacturers to sell plug compatible disk drives for use with IBM computers and an entire industry was born.

IBM 2314/2319

[edit]
IBM 2314s at the University of Michigan. Note removable disk packs and empty covers on top of the drives.

IBM 2314 Disk Access Storage Facility Model 1

[edit]

The IBM 2314 Disk Access Storage Facility Model 1 was introduced on April 22, 1965, one year after the System/360 introduction.[40] It was used with the System/360 and the System/370 lines. With the Two Channel Switch feature it could interface with two 360/370 channels. The 2314 Disk access mechanism was similar to the 2311, but further recording improvements allowed higher data density. The 2314 stored 29,176,000 characters (200×20×7294 bytes per track) on a single removable IBM 2316 disk pack which was similar in design to the 1316 but was taller as a result of increasing the number of disks from six to eleven. The 2316 disk pack containing the eleven 14-inch (360 mm) diameter disks yielded 20 recording surfaces. The drive access consisted of 20 individual R/W heads mounted on a common actuator which was moved in and out hydraulically and mechanically detented at the desired track before reading or writing occurred. Each recording surface has 200 tracks. Access time was initially the same as the 2311, but later models were faster as a result of improvements made in the hydraulic actuator. Data transfer rate was doubled to 310 kB/s.

The original Model 1 consists nine disk drives bundled together with one price; separately shipped was a storage control unit, a single drive module, and two four drive modules for a total of nine drives. The drives are mounted in individual drawers that are unlatched and pulled out to access the disk pack. Because of their appearance they acquired the nickname of "Pizza Ovens". Only eight drives of the nine are available to the computer at any one time. The ninth drive is there for a spare for the user and can also be worked on "offline" by a Field Engineer while the other drives are in use by the customer. Each drive's system address is determined in part by a user-swappable plug, one such plug denoting a spare drive not system accessible. This permits physically changing the address of a drive by changing the plug.

IBM 2314

A 2844 Control Unit can be added to the 2314 Control Unit which allows two S/360 Channels simultaneous access to two separate disk drives in the Storage Facility.

Other 2314 models came later:

IBM 2314 direct access storage facility - A series

[edit]

In 1969 IBM unbundled the facility into separate models allowing up to nine drives (eight on line) attached to a 2314 Storage Control:

  • 2312 Disk Storage, a one drive module.
  • 2313 Disk Storage, a four drive module.
  • 2318 Disk Storage, a two drive module.

IBM 2319

[edit]
  • In a response to competition from plug compatible manufacturers of 2314 equivalent storage subsystems, IBM beginning 1970 introduced a series of low priced three drive module 2319s which were manufactured by removing one module from the four drive module 2313, rebranding it as a 2319 A1 and offering it at a substantially reduced rental price. This had the effect of lowering the rental price to new customers while keeping the high rental price on existing customers. The 2319-A1 attaches to integrated controllers for only the System/370 Models 135 and 145. Conventional 2314 DASD such as the 2312 or 2318 can attach to the 2319-A1
  • 2319 B series of three disk drives modules allow three, six or nine drive attachment to a new 2314 Model B Storage Control Unit.

IBM 3310

[edit]

IBM introduced the IBM 3310 Direct Access Storage Device on January 30, 1979, for IBM 4331 midrange computers.[41] Each drive had a capacity of 64.5 MB. The 3310 was a fixed-block architecture device, used on DOS/VSE and VM, the only S/370 operating systems that supported FBA devices.

IBM 3330

[edit]
3D artist's concept of an IBM 3330 Direct Access Storage Facility. Shown are three 3330s and one 3333 (on the right).

The IBM 3330 Direct Access Storage Facility, code-named Merlin, was introduced in June 1970 for use with the IBM System/370 and the IBM System 360/195. The original announcement included the 3330 Model 1, with two drives, and the 3330 Model 2, with only one drive. The 3330 has removable disk packs, similar to its predecessors, and the packs hold 100 MB (404×19×13,030 bytes). Access time is 30 ms and data transfers at 806 kB/s. A major advance introduced with the 3330 is the use of error correction, which makes the drives more reliable and reduces costs because small imperfections in the disk surface can be tolerated. The circuitry can correct error bursts up to 11 bits long through use of fire codes.

The initial configuration consists of one storage control unit (3830 Model 1) bolted to a 3330, with optionally three more 3330's bolted together. This is known as a string, making a maximum of eight drives in a string.

In August 1972 IBM announced the 3830 Model 2 Storage Control and the 3333 Disk Storage and Control, separating the control unit from the string. The 3830 became a director type of storage control, controlling one or more strings. The now first unit of the string, the 3333 contains a controller and two drives and it can control up to three attached 3330's for a maximum of eight drives in the string as shown in the illustration. The 3830 Model 2 can connect two 3333's for a maximum of 16 drives per storage control and the 3333 optionally has a string switch that enables it to be connected to two different storage controls.

In 1973 IBM announced double density versions (-11 models) of the 3330 product line: the 3333–11, 3330-11 and the 3336–11; the 3336-11 Disk Packs hold up to 200 MB (808x19x13,030 bytes). It is not possible to mix single and double density drives within a string. It is possible to field upgrade existing 3330 Models to Model 11, but this is a major task, as the drives had to be converted, and all the existing data had to be copied to the new media.

The 3330 was withdrawn in 1983.[42]

IBM 3340 and 3344

[edit]

The IBM 3340 and 3344 have similar characteristics. However, only a 3340 can serve as head of string; there are no A model 3344 drives, and a 3344 must be attached to a 3340 A model as head of string.

IBM 3340

[edit]

The IBM 3340 Direct Access Storage Facility, code-named Winchester, was introduced in March 1973 for use with IBM System/370.[43] Three models were announced, the 3340-A2 with two drives and a controller, the models B2 (two drives) and B1 (one drive). B-units can connect to the model A2 to a maximum of eight drives.

It uses removable data modules that included the head and arm assembly; an access door of the data module opens or closes during a mechanical load/unload process to connect the data module to the drive; unlike previous disk packs and cartridges there is no cover to remove during the insertion process. Access time is 25 millisecond and data transfers at 885 kB/s. Three versions of the removable IBM 3348 Data Module were sold, one with 35 megabyte capacity, another with 70 megabytes, the third also has 70 megabytes, of which 500 kilobytes were accessible with fixed heads for faster access. The 3340 also uses error correction. It was withdrawn in 1984.

The 3340 was developed in San Jose under the leadership of Ken Haughton. Early on, the design was focused on two removable 30 megabyte modules. Because of this 30/30 configuration, the code name Winchester was selected after the famous Winchester .30-30 rifle;[44] subsequently the capacities were increased, but the code name stuck.

One significant aspect of this product, and the reason that disk drives in general became known as "Winchester technology", was that this head design was very low cost and did not require the heads to be unloaded from the media. Winchester technology allowed the head to land and take off from the disk media as the disk spun up and down. This resulted in very significant savings and a large reduction of complexity of the head and arm actuating mechanism. This head design rapidly became a standard design within the disk drive manufacturing community.

Up into the early 1990s the term Winchester or Winnie was used for hard disk drives in general long after the introduction of the 3340, but is no longer in common use in most parts of the world.

IBM 3344

[edit]

The IBM 3344 is similar to the 3340, except that it uses fixed media rather than removable 3348 data modules, each spindle has four logical drives each with the capacity of a 3348–70, there is no A (head of string) model and it is only available in dual drive models. The 3344-B2F is identical to the 3344-B2 except that both drives have fixed heads over some cylinders. Both 3344-B2 and 3344-B2F require a 3340-A2 or 3340-A2F as head of string. Inside, the 3344 is exactly the same as IBM 3350, the difference is only in the microcode in the control unit.

IBM 3350

[edit]
IBM 3350 at the Enter Museum

The IBM 3350 Direct Access Storage Facility, code-named Madrid, was introduced in 1975 for use with IBM System/370.[45] Its non-removable head-disk assemblies (HDAs) are sealed and included the head and arm assembly. The 3350 disk geometry is 555 cylinders, 30 heads, and 19,069 bytes per track, which give each HDA a storage capacity of 317,498,850 bytes. Sealed HDAs were standard practice on all IBM DASD hereafter.

Disk units are identified as Models A2, A2F, B2, B2F, C2, and C2F with each model containing two HDAs. Model A2 and A2F has one additional electronic board, allowing it to be connected to the control unit. They are referred to as controllers, and also sometimes head-of-string. The models are installed in strings of units with an A2 or A2F unit, and then up to three B2 units or up to two B2s and a C2. The A2 unit usually has a string switch, allowing it to be connected to two different storage control units. This allows two I/O operations simultaneously take place to two different HDAs in the string. The storage control unit can be a 3830 Model 2, or the ISC (Integrated Storage Control) found in the 3148, 3158 or 3168 cpu's. Also later control units (3880) are backwards compatible and can be used. The C2s unit also contains a controller, that can be connected to a storage control unit and serves as a secondary path to itself and the A2 and B2 units. The C2 controller is a spare, it can only be used when the controller in the A unit is broken, and subsequently powered off. It has also limited connections, usually the A unit has a string switch, but the C unit only can be connected to one storage control unit. The valid 3350 strings are: -A, -AB, -ABB, -ABBB, -AC-, -ABC-, or -ABBC- configurations.

The "x2F", as in Model A2F, unit is a normal x2 unit, but its two HDAs also have a Fixed Head area over the first five cylinders, thereby reducing[d] seek time to zero for these five cylinders. This fixed head area is intended to be allocated to the frequently accessed HASP or JES2 checkpoint area and thus greatly reduce head motion on the SPOOL device. The fixed head area can also be utilized for TSO swap data (MVT and SVS) and system swap data (MVS) wherein the swap data for SVS and MVS consist of blocks of pages that have been in memory when an address space is selected for swap-out; those pages need not be contiguous and in general do not include pages that have not been modified since their last page-in. This system architecture greatly improves context switches between TSO users or batch regions.

The IBM 3350 family was withdrawn in September 1994.

IBM 3370 and 3375

[edit]

IBM 3370

[edit]

IBM introduced the IBM 3370 Direct Access Storage Device in January 1979 for IBM 4331, 4341, and System/38 midrange computers.[46] It has seven fixed 14-inch (360 mm) disks, and each unit has a capacity of 571 MB. It was the first HDD to use thin-film head technology; research on that technology started at Thomas J. Watson Research Center in the late 1960s.[46] The 3370 was a fixed-block architecture device, used on DOS/VSE and VM, the only S/370 operating systems that supported FBA devices.

IBM 3375

[edit]

The sister unit was called the IBM 3375 and used count key data architecture, which was required for OS/360 and successor operating systems.

IBM 3380

[edit]
IBM 3380 disk drive module

The IBM 3380 Direct Access Storage Device was introduced on June 11, 1980.[47] It uses film head technology and has a unit capacity of 2.52 gigabytes (two hard disk assemblies each with two independent actuators each accessing 630 MB within one 3380 unit) with a data transfer rate of 3 megabytes per second. Average access time was 16 ms. Purchase price at time of introduction ranged from $81,000 to $142,200. Due to tribology problems encountered between heads and media, the first units did not ship until October 1981.[47]

Similar to its predecessor (3350) the standard configuration is one A unit and up to three B units, but because each 3380 contains four devices each string now can contain up to 16 devices. Usually it is connected to a 3880 storage control units with two paths, allowing two simultaneous I/O operations, however, the operations must target different HDAs.

In February 1985, IBM announced a double density version – the Extended Capability Models of the 3380 (3380 E) having 5.04 gigabytes per 3380 unit, that is, two 1.26 gigabyte actuators on two hard disk assemblies.

A triple capacity version, the 3380 K was announced in August 1987 having 7.562 gigabytes per 3380 unit, that is, two 1.89 gigabyte actuators on two hard disk assemblies. The new Model K and Model J can optionally run in four-path mode. In this mode, the string has two A units located in the middle, and up to three B units connected on each side, giving a maximum of 32 devices in a string. This requires they are connected to a 3990 storage control unit, and allows four simultaneous I/O operation in the string.

There are twelve models of the IBM 3380 family: six A-units, five B-units and one C-unit. A-units (heads of string) contain additional logic to perform string controller functions and connect to IBM storage control units (3880 or 3990). The C-units connect directly to an IBM channel. B-units connects to A-units or C-units.

The last models were withdrawn by IBM in May 1996 representing a production run of 15 years; a run longer than most disk drives[48]

IBM 3390

[edit]

The IBM 3390 Direct Access Storage Device series was introduced November 1989, offering a maximum storage of up to 22 gigabytes in a string of multiple drives.[49] Cost of a storage system varied by configuration and capacity, between $90,000 and $795,000.

A 3390 string consists of an A unit placed in the middle, and optionally one or two B units bolted to its sides. The A unit can have four or eight devices, each B unit can have up to 12 devices. The 3390 is always running in four-path mode, connected to a 3990 storage control unit. Optionally the 3990 can have a second 3390 string attached, giving a maximum of 64 devices in the subsystem.

Packaged in Hard Disk Assemblies with two actuator-head units and one set of platters, a model 1 HDA provides 1.89 GB before formatting and a model 2 provides 3.78 GB/HDA. The Model 3 enhancement to the drive family, announced September 11, 1991, increased capacity 1.5 times to 5.67 GB/HDA and the Model 9, announced May 20, 1993, further increased capacity 3 times to 11.3 GB/HDA.[50][51]

All the preceding DASD models are equipped with a large AC motor, driving the HDA with a belt; however, the 3390 HDA is directly driven by a DC motor that is included in the enclosure.

The 3390 Model 9 was the last Single Large Expensive Disk (sometimes called SLEDs) drive announced by IBM.

IBM 9340 and 9345

[edit]

IBM 9345

[edit]

The IBM 9345 HDD first shipped in Nov 1990 as an RPQ on IBMs SCSE (SuperComputing Systems Extensions). Developed at IBM's San Jose, California laboratory under the code name Sawmill. It was an up to 1.5 GB full height 514-inch HDD using up to 8 130 mm disks. It was the first HDD to use MR (Magneto Resistive) heads.[52]

IBM 9340

[edit]

In October 1991 the 9345 DASD was announced as part of the IBM 9340 channel-attached, count key data (CKD) DASD subsystem family which attached to IBM mainframes including the ES/9000 processor family.[53] The 9345 DASD Model 1 had two 1.0 GB HDDs while the Model 2 had two 1.5 GB HDDs.

For most practical applications, the 9340/9345 was functionally equivalent to a 3990/3390, although without non-volatile RAM cache of the 3990 and with a somewhat shorter maximum block length than the 3390.

The OS's IOS component learned of this device's characteristics through a special initializer, IECCINIT, which also serviced other DASD device types, and for the same purpose. It was at initialization-time that the OS learned that the 9340 has no non-volatile cache and the 9345 has a shorter than expected track capacity. The initializer, therefore, assigned a different device type than the 3990/3390.

9330 family of disk drives

[edit]
9335 drive
  1. 9331 Diskette Unit models 1 and 11 contained one 8-inch FDD while the models 2 and 12 contained one 514-inch FDD.[54]
  2. 9332 Direct Access Storage Device used the IBM 0667 HDD.[55]
  3. 9333 High Performance Disk Drive Subsystem used the IBM 0664 or IBM 0681 HDDs depending upon subsystem model
  4. 9334 Disk Expansion Unit attaches from one to four SCSI HDDs to the RS/6000 system.[56]
  5. 9335 Direct Access Storage Subsystem This HDD used in this subsystem was developed under the code name "Kestrel" at IBM Hursley, UK, and was an 850 MB HDD using three 14-inch disks with dual rotary actuators, each actuator accessing three surfaces with two heads per surface.[57] The HDD was in the rack mountable 9335 announced as a part of the October 1986 IBM 9370 Information System announcement.[58] There is no known OEM version of this HDD.
  6. 9336 Disk Unit used the IBM 0681 HDD (Redwing)[59]
  7. 9337 Disk Array Subsystem used the IBM 0662 (Spitfire) or 0663 (Corsair) HDDs.[59]

HDDs offered for IBM small systems

[edit]

IBM 2310

[edit]
IBM 2315 disk cartridge

The IBM 2310 Removable Cartridge Drive was announced in 1964 with the IBM 1800,[60] and then in 1965 with the IBM 1130; it likely first shipped with the 1130 in late 1965.[61] It could store 512,000 16-bit words (1,024,000 bytes) on an IBM 2315 cartridge. A single 14-inch (360 mm) oxide-coated aluminum disk spun in a plastic shell with openings for the read/write arm and two heads.

IBM 5444

[edit]

The IBM 5444 was announced September 1969 as part of System/3. Developed at IBM's Hursley, England, laboratory under code name Dolphin[57] it used the 5440 disk cartridge. The cartridge in turn contained one 14-inch disk. There were three models:[62]

  • Model 1 has one fixed disk and one removable disk each with 100 tracks per surface for a disk cartridge capacity of 1.23 MB
  • Model 2 has one fixed disk and one removable disk each with 200 tracks per surface for a disk cartridge capacity of 2.46 MB
  • Model 3 has only one removable disk with 200 tracks per surface for a disk cartridge capacity of 2.46 MB

IBM 62GV

[edit]
IBM 62TM

The 62GV first shipped in May 1974. Developed at IBM's Hursley, UK, laboratory under the code name Gulliver with an initial capacity of 5 MB. Subsequent models have 10 MB (62TM) and 14 MB capacities. It used a Swinging Arm actuator with one 14-inch disk. The simple design of the actuator,[63] invented at IBM's UK Hursley Labs, became IBM's most licensed electro-mechanical invention[64] of all time, the actuator and filtration system being adopted in the 1980s eventually for all HDDs, and still universal nearly 40 years and 10 Billion arms later. During its production life the IBM 62GV shipped 177,000 units making it the first HDD known to have shipped in excess of 100,000 units.[57]

OEM and Small Systems HDDs

[edit]

This section lists IBM manufactured HDDs offered both as an OEM product and for attachment to IBMs small systems such as the System/3, System/32, /34 and /36 and the AS/400. HDDs are identified by their OEM model number and listed chronologically by date of first customer shipment.

IBM 0680

[edit]

The 0680 first shipped in 1979 on most IBM small systems[65] and the low end of the System/370 as the 3310 direct access storage.[41] The OEM version was announced as the 0680 in September 1981.[5] Developed at IBM's Hursley, UK, laboratory under the code name Piccolo with an initial capacity of up to 65MB, it used six 8-inch disks (210 mm) and had an improved rotary actuator.[66][57]

A double capacity version, the 62SW, shipped in June 1984 but very few units were sold because its price per megabyte was the same as the 62GV.[57]

IBM 0676

[edit]

The 0676 first shipped in November 1982 as a 5247 Disk Storage Unit for the IBM System/23 Datamaster.[67] Developed at the IBM Rochester, Minnesota, laboratory as the 21ED it was an 8-inch HDD with an initial capacity of 15 or 30 MB in two or four 210 mm disks. In 1983 it shipped as the HDD in the 5360 System Unit of the S/36. In 1984 its capacity was doubled by doubling the number of tracks per surface and it was incorporated into the 5362 System Unit of the System/36.[68]

IBM 0665

[edit]
A British IBM 0665-30 hard disk exposed, possibly manufactured in 1985. A head crash has occurred.

The 0665 first shipped in October 1985 in the system unit for the PC AT (5170).[69] Developed under the code name "Pixie" at IBM Rochester, Minnesota, it was a 514-inch HDD with capacities of 20, 30 and 44 MB.

IBM 0667

[edit]

The 0667 first shipped in August 1986.[67] Developed at IBM Rochester, Minnesota, under the code name "Grant", it was a 70 MB ESDI full height 514-inch HDD with up to four 130 mm disks.[70] It was offered as a feature on certain models of the PC RT (6150, 6151, 6152) and in System/36 Model System Units (5363, 5364).[71]

IBM 0669

[edit]

The 0669 first shipped in 1987.[67] Developed at IBM Rochester, Minnesota, under the code name "Grant-Prime", it was a full-height 512-inch HDD with a capacity of up to 115 MB on up to four 130 mm disks.[72] It was the HDD internal to the System/36 5363 System Unit and Series 1 4956 System Unit.[72]

IBM 0671

[edit]

The 0671 first shipped in 1987.[67] Developed under the code name "Lee" at IBM Rochester, Minnesota, it was an up to 316 MB ESDI full height 514-inch HDD with up to eight 130 mm disks depending upon model.[72] This was IBM's first usage of a thin metal film as the disk's recording surface.[72] In 1988 it shipped as part of the 9404 System Unit of the IBM AS/400 system which contained two, or optionally three of these HDDs.[73]

IBM 0661

[edit]

The 0661 first shipped in 1989 as the model 371.[74] Developed initially under the code name "Lightning" at IBM Rochester Minnesota (and IBM Hursley, UK) as a 320 MB SCSI HDD with up to eight 95 mm disks (14 heads), it was followed in 1990 with a 400 MB version, code named "Turbo". During 1990 it was added as a standard drive on several major IBM systems,[74] e.g., IBM AS/400 System Unit Model CXX.[75]

IBM 0681

[edit]

The 0681 first shipped in April 1990. Developed at IBM's Hursley, UK, laboratory under the code name Redwing, it was the last HDD product developed at Hursley.[76] It was an up to 857 MB full-height 514-inch HDD using up to 12 130 mm disks. It was the first HDD to use PRML decoding of data.[3] It was the drive component of the 9333 Disk Drive Subsystem which first shipped in early 1992.

A higher density, 1.07 GB, version was incorporated into the 9333 subsystem in May 1992.[77]

IBM 0663

[edit]

The 0663 first shipped in late 1991. Developed under the code name "Corsair", it was a 312-inch HDD with the height of a 512-inch half-height device (1.6-inch high) and up to 1 GB on up to 8 95 mm disks.[78][79] It was offered as a feature on certain models of the PS/2 and RS/6000.[79] It was the first OEM disk drive to use MR Heads.[3]

IBM 0664

[edit]

The 0664 first shipped in November 1992. Developed under the code name "Allicat" at IBM Rochester, Minnesota, it was a full-height 512-inch HDD (3.25-inch high) that combined two 312-inch devices in one, with up to 2.013 GB capacity on up to 8 95 mm disks.[80][81]

IBM 0662

[edit]

The 0662 first shipped in June 1993. Developed under the code name "Spitfire" at IBM Rochester, Minnesota, it was a full-height (1-inch high) 312-inch HDD with 1.05 GB on three disks or five disk surfaces.[81][82][83] It was the HDD internal to the 9336 Disk Unit and the 9337 Disk Array.[81]

Floppy disk drive

[edit]

Another important IBM innovation is the floppy disk drive. IBM first introduced the 8-inch FDD in 1971 as a read only program load device. In 1973 IBM shipped its first read/write floppy disk drive as a part of the 3740 Data Entry System. IBM established early standards in 8" FDDs but never sold such products separately so that the industry then developed separate from IBM.

IBM was at one point was the world's largest purchaser of OEM 514-inch FDDs; its selection of the two-sided, 48 tracks-per-inch model helped establish the model as the de facto industry standard. IBM made extensive preparations to manufacture such models and smaller form factors but cancelled all such efforts in 1985.[84] IBM's 1983 attempt to OEM its 4-inch DemiDisk failed.[85]

"Star" series of HDDs

[edit]

On October 17, 1994, IBM's Storage Systems division announced three new families of hard disk drives, the Travelstar 212-inch family for notebooks, the Deskstar 312-inch family for desktop applications and the Ultrastar 312-inch family for high performance computer system applications.[7]

IBM's first HDD versus its last HDDs

[edit]

The following table compares IBM's first HDD, the RAMAC 350, with the last three models it manufactured in each of its "Star" series of OEM HDDs. It illustrates HDD's spectacular decline in cost and size along with corresponding improvement in capacity and performance.

Parameter (units) RAMAC 350[8] Ultrastar 146Z10[86][87] Deskstar 180GXP[88][89] Travelstar 80GN[90][91]    46-year improvement (maximum)   
Announced Sep 1956 Jul 2002 Oct 2002 Nov 2002
Capacity (gigabytes) 0.004 146 180 80 48,000
Dimensions (inches) 60×68×29 4×1×5.75 4×1×5.75 2.75×0.38×3.95
Dimensions (mm) 1500×1700×700 102×25×146 102×25×146 70×9.5×100
Volume (in³) 118,320 23 23 4 29,161
Volume (litres) 1,939 0.4 0.4 0.1
Weight (lbs) 2,140 1.7 1.4 0.2 1,244
Weight (kg) 971 0.8 0.64 0.095
Power (watts) 8100 BTUs/hour (i.e., 2374 watts),[92] up to 5500 VA depending upon model[93] 16 10.3 1.85 1,283
Power density (megabytes/watt) 0.0016 9,125 17,476 43,243 27,375,856
List price (US$)   $34,500[94]
($234,000 in 2002 dollars)[95]
1200 360 420
Price/megabyte (US$)   $9,200
($68,000 in 2002 dollars)
0.0082 0.0020 0.0053    4,600,000
(34,000,000 in 2002 dollars)
Density (megabits/in²) 0.002 26,263 46,300 70,000 35,000,000
Density (kilobits/mm²) 0.003 40,708 71,765 108,500
Volume density (gigabytes/in³) 0.00000003 6 8 20 622,100,131
Volume density (megabytes/cm³) 0.000002 388 478 1,203
Latency (ms) 25 3 4 7 8
Average seek time (ms) 600 5.9 10.2 12 102
Data rate (megabytes/s) 0.001 103 29.4 43.75 11,719

DASD devices not HDDs or FDDs

[edit]

IBM in some of its operating systems classifies HDDs and FDDs as DASDs, direct access storage devices. Other technologies so classified include:

IBM 7320 drum

[edit]

The IBM 7320 is a magnetic-drum storage unit announced in 1962.

IBM 2301 drum

[edit]

The IBM 2301 is a magnetic-drum storage device introduced in the late 1960s

IBM 2303 drum

[edit]

The IBM 2303 is a magnetic-drum storage device introduced in 1964.

IBM 2321 Data Cell

[edit]

The IBM 2321 Data Cell announced in 1964 is a device that uses short strips of magnetic tape to store data. It holds 10 40 MB removable cells, for a total capacity of 400 MB.

IBM 3850 Mass Storage System

[edit]

The IBM 3850 Mass Storage System, announced in 1974, is a library system of tape cartridges that staged data from the cartridges onto physical IBM 3330 or 3350 disk drives which then appeared to the system as virtual 3330 drives.

See also

[edit]

Notes

[edit]

References

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia
from Grokipedia
The history of IBM magnetic disk drives encompasses the development and commercialization of technology from the mid-20th century through the early 2000s, marking IBM's pivotal role in transforming data access from sequential tape-based systems to random-access that enabled modern applications. IBM's journey began in 1956 with the announcement of the (Random Access Method of Accounting and Control), which incorporated the IBM 350 Disk Storage Unit—the world's first commercial —featuring 50 stacked 24-inch aluminum platters coated in , providing a total capacity of approximately 5 million characters (about 3.75 MB) and enabling in seconds rather than hours or days. The system, developed at IBM's San Jose laboratory under , weighed over a ton and was first shipped to Zellerbach Paper Company in in June 1956, revolutionizing industries like , reservations, and scientific by supporting processing. Throughout the 1960s, IBM advanced disk technology with innovations such as the 1961 IBM 1301 Disk Storage, which introduced aerodynamic head sliders for higher areal density of 26,000 bits per square inch, and the 1962 IBM 1311 Disk Storage Drive, the first with removable disk packs holding 2 million characters across six disks. The 1964 launch of the mainframe family further integrated , with models like the 1965 IBM 2314 Direct Access Storage Facility offering 29 MB packs, enhancing reliability and scalability for enterprise use. In the 1970s and , IBM pioneered sealed, non-removable designs to combat contamination, exemplified by the 1973 IBM 3340 "Winchester" drive, which provided up to 70 MB per drive (280 MB per string of drives) using lubricated platters and low-mass heads, setting a standard for future hard drives. This era culminated in the 1980 IBM 3380, a high-capacity drive that generated billions in annual revenue and supported the growing demands of computing. By the 1990s, IBM led in head technology with the 1990 IBM 9345, the first drive using magneto-resistive heads for capacities of 1–1.5 GB, dramatically improving read performance and density. Innovations continued into the late 1990s, including the 1999 , the smallest hard drive at the time with 170–340 MB capacities, targeting mobile devices. IBM's disk drive , which had dominated the market for decades, effectively ended in 2003 when the company sold its hard disk drive operations to for $2.05 billion, transferring assets including and manufacturing to form what became . This divestiture allowed IBM to focus on services and software while preserving its legacy in storage innovation.

Early Developments (1956–1960)

IBM 350 RAMAC

The IBM 350, part of the (Random Access Method of Accounting and Control) system, represented 's pioneering effort in magnetic , initiated in 1952 by a team of engineers in , led by . The project aimed to develop a random-access storage device to supplant slower punched-card and magnetic-tape systems for business data processing, particularly accounting tasks. After four years of development, engineering prototypes began shipping in June 1956, with the first unit delivered to Zellerbach Paper Company in . officially announced the system on September 14, 1956, marking the debut of commercial hard disk technology. Technically, the IBM 350 featured 50 aluminum platters, each 24 inches in diameter and coated with magnetic oxide, stacked vertically and spinning at 1,200 RPM to store a total of approximately 5 million characters (equivalent to 3.75 MB in modern terms, using encoding). Data transfer occurred at 8,800 characters per second, with an average access time of 600 milliseconds, enabling faster retrieval than tape-based alternatives. The unit weighed over 2,000 pounds (including associated air compressor and power supply) and was leased as part of the full 305 RAMAC system for $3,200 per month, reflecting its high cost and complexity for the era. Key architectural innovations included the first implementation of read/write heads that floated on a thin cushion of pressurized air (a hydrostatic ) just 800 microinches above the disk surfaces, preventing contact and wear while allowing reliable data access across 100 recording surfaces. A hydraulic mechanism positioned the head assembly linearly across tracks, providing precise control over the 100 tracks per surface without mechanical issues common in earlier designs. The drive integrated seamlessly with the 305 computer, a vacuum-tube-based tailored for real-time , where the disk served as the primary non-volatile storage. Early deployments focused on business applications like inventory tracking and payroll processing, with the Zellerbach installation handling paper company records efficiently. A demonstration at the 1956 National Association of Accountants convention showcased its speed in random-access queries, impressing industry observers and accelerating adoption in commercial computing. Over 1,000 units were produced before withdrawal in 1961, establishing the foundational principles of magnetic that influenced all subsequent hard disk drives, including later models like the 353 for enhanced reliability.

IBM 353

The IBM 353 Disk Storage unit, developed from 1956 to 1960 at 's San Jose laboratory as part of the supercomputer project and introduced in 1961, marked the first production-scale implementation of 's magnetic disk technology following the prototype 350 RAMAC. It drew directly from field trial experiences with the RAMAC to refine the design for greater reliability and performance in demanding computational environments. Unlike the RAMAC's experimental nature, the 353 was optimized for integration into high-end systems, transitioning from a novel concept to a practical component for advanced . This evolution helped position magnetic disks as reliable alternatives to slower magnetic drums and sequential tape systems in early . Key technical advancements in the IBM 353 addressed the RAMAC's limitations in speed and durability, including the adoption of flying magnetic heads—one per recording surface—based on a 1954 aerodynamic concept by IBM engineer Jake Hagopian. These heads hovered on a thin air cushion over the disk surfaces, reducing physical contact and minimizing wear compared to the RAMAC's contact-style heads, while enabling higher areal densities and faster data access suitable for real-time scientific simulations. The unit featured a stack of disks with 80 read/write heads across multiple surfaces, supporting a total capacity of 16 MB (2,097,152 64-bit words) and a data transfer rate of up to 1 MB/s, a significant leap that facilitated burst transfers directly to the Stretch's core memory. Enhanced error correction mechanisms, including parity checks, further improved over the RAMAC's basic schemes. Production of the IBM 353 was limited due to the niche focus on the Stretch program, with only a small number of units manufactured at IBM San Jose for deployment in specialized supercomputing installations, such as the Los Alamos National Laboratory's system for nuclear research. These drives supported both scientific simulations and emerging requiring rapid , demonstrating disk storage's versatility beyond the RAMAC's accounting-focused applications. By overcoming early prototype challenges like frequent head crashes and disk media degradation—issues prevalent in the RAMAC's hydraulic positioning and contact mechanisms—the 353 validated multi-platter disk architectures as scalable and dependable, paving the way for broader commercialization in subsequent models.

IBM 355

The IBM 355 Disk Storage Unit was announced on September 14, 1956, as a rack-mounted addition to the Magnetic Drum Data Processing Machine, providing random access storage as an alternative to for mid-1950s computing systems. It represented IBM's early effort to expand capacity for business , enabling faster retrieval compared to sequential tape access in applications like inventory management and accounting. The system supported up to four disk storage units per control unit, with each Model 1 unit offering a capacity of 6 million digits (approximately 3 MB) using 50 magnetic platters of 24 inches in diameter. The platters spun at 1,200 rpm, and the design featured one read/write head per track, eliminating radial seek time and achieving an average access time of about 50 ms, dominated by rotational latency. This head-per-track configuration marked an improvement over the moving-head assembly in the prior IBM 350, prioritizing speed for frequently accessed data. Integration was facilitated by a shared control unit that managed multiple drives and included basic error detection and logging capabilities to support reliable operation in attached systems. Deployed in sectors such as airlines for reservation systems and banks for transaction records, it enhanced real-time data handling on the 650. Later compatibility extended to systems like the 1410 and 7010, where it served as the required system residence drive. The IBM 355 bridged the transition from single-unit disk designs like the IBM 350 to more scalable, modular direct access storage device (DASD) architectures, laying groundwork for subsequent innovations in multi-drive configurations. Its emphasis on fixed heads influenced removable media approaches in later models such as the IBM 1301.

Removable Disk Pack Era (1961–1969)

IBM 1301

The IBM 1301 Disk Storage Unit represented a significant advancement in magnetic disk technology, introducing aerodynamic flying heads to early computing systems. Developed by IBM's San Jose laboratory as part of the Advanced Disk File project starting in 1955, it aimed to dramatically improve upon the limitations of the IBM 350 RAMAC by achieving ten times the storage capacity and one-tenth the access time. Announced on June 2, 1961, the device began shipping in 1962 and was initially compatible with IBM 7000-series mainframes such as the 7090 and 7094, as well as the 1410 and 7010 systems. This launch marked the first commercial use of flying head technology on a large scale. Key specifications of the IBM 1301 included a storage module comprising 25 disks of 24-inch diameter, spinning at 1,800 RPM, which provided up to 50 recording surfaces for . Each module offered a capacity of 28 million 6-bit characters in the context of 7000-series systems (equivalent to approximately 21 MB), or 25 million 8-bit characters for the (about 25 MB), with the Model 1 unit supporting one module and the Model 2 supporting two for up to 56 million characters total. Data transfer rates reached 90,100 characters per second (roughly 67 KB/s or 4 MB/min), while maximum access time was reduced to 180 milliseconds through parallel head operations across surfaces. These metrics established a new benchmark for random-access storage, supporting real-time applications that were impractical with prior fixed-media drives. Innovations in the IBM 1301 centered on its aerodynamic flying heads, one per recording surface, mounted on a comb-like arm that allowed simultaneous access to multiple tracks in a "string-of-pearls" configuration for reduced latency. The heads "flew" at a height of just 250 microinches above the disk surface—compared to 800 microinches in earlier designs—enabling higher areal density of 26,000 bits per and eliminating the need for an external air filtration system. Powered by a hydraulic positioning mechanism, this approach achieved finer of 50 tracks per inch and recording densities up to 520 bits per inch. The adoption of the 1301 transformed in mid-1960s by providing higher capacity and faster access, supporting real-time applications that were impractical with prior drives. It found widespread use in commercial applications, notably powering ' reservation system with 16 units handling real-time , as well as in military data centers for high-reliability archival needs. This performance paved the way for subsequent removable pack designs, such as the smaller IBM 1311.

IBM 1302

The IBM 1302 Disk Storage Unit was introduced in September 1963 as a higher-capacity evolution of the 1301, designed to meet the expanding data requirements of 1960s mainframe systems such as the 7000 series and early System/360 models. It represented a significant step in magnetic disk technology by quadrupling storage capacity through advancements in recording techniques, enabling more efficient handling of large-scale tasks. The 1302 employed a with each module containing a stack of 25 disks (24 inches in ), of which 20 were usable to provide 40 recording surfaces, organized into 500 with 500 tracks per surface. This configuration delivered a capacity of 117 million 6-bit characters (approximately 87.9 MB) per single-module unit (Model 1), or double that for the dual-module Model 2, achieved via doubled linear and track densities of 1050 bits per inch and 100 tracks per inch, respectively, facilitated by finer read/write head gaps. Access performance remained comparable to the 1301, with an average total access time of 165 milliseconds—including seek motion and rotational latency at 1800 RPM—allowing full scans in 1.3 seconds and a data transfer rate of around 180 KB per second. Compatible with existing 1301 controllers, the 1302 integrated seamlessly into prior IBM data processing systems like the and 7090/7094, while its architecture supported adaptation for the System/360 via the related 2302 model, broadening its applicability in emerging compatible computing environments. This minimized migration costs for users upgrading storage. The 1302's density improvements facilitated larger datasets for scientific simulations and business applications on mainframes, marking a pivotal shift toward higher areal densities in disk modules and paving the way for standardized packs in successors like the IBM 1311. By enabling up to five units per system for total capacities exceeding 500 MB, it underscored the rapid evolution of direct-access storage during the early mainframe era.

IBM 1311

The IBM 1311 Disk Storage Drive, announced in October 1962 and first shipped in July 1963, introduced the industry's first removable , marking a significant advancement in portable and secure data storage for mid-range computing systems. This drive established the 14-inch as a , enabling users to interchange media between drives and systems while providing offline capacity limited only by the number of packs available. Designed primarily for IBM's medium-scale business and scientific computers, such as the 1401, 1440, 1410, 1620, and 7010, the 1311 reduced reliance on punched cards and tapes by offering storage in a compact form factor roughly the size of a top-loading . The drive utilized the , a sealed cartridge containing six 14-inch disks spaced 0.5 inches apart, providing 10 recording surfaces (the innermost surfaces of the top and bottom disks were unused to minimize ). Each pack offered a formatted capacity of 2 million alphanumeric characters (approximately 2.5 MB using 6-bit BCD encoding) in standard sector mode, expandable to 2.98 million characters with the optional track-record feature that eliminated sector overhead. The disks spun at 1,500 RPM, resulting in a rotational latency of up to 40 ms. Access was managed via a comb-like assembly of 10 read/write heads, with 100 tracks per surface and 20 sectors per track at a linear density of about 1,025 bits per inch and 50 tracks per inch. Key features distinguished the 1311 models for cost-effective deployment. Model 1 functioned as the master drive, interfacing directly with the host system and supporting up to four additional Model 2 satellite drives in a daisy-chain configuration, allowing up to five drives per string. The standard seek mechanism used a hydraulic for head positioning, delivering an average time of 250 ms (maximum 400 ms, including 150 ms seek plus rotational delay). Optional enhancements included direct seek, which bypassed the home position to reduce average access to 150 ms (maximum 250 ms) and was standard on higher-end models; seek overlap for concurrent operations across drives; scan disk for sequential searching; and track record for denser data packing. Each pack weighed about 10 pounds, with a change time of under 2 minutes, and the drive's closed-loop design with vented hubs minimized dust contamination. Data transfer occurred at rates supporting the era's systems, typically around 15,000 characters per second. The 1311's design proved highly influential, with over 6,000 units and 41,000 disk packs shipped by the end of , finding widespread use in banking, inventory management, and processing applications. Non-IBM vendors, including , quickly adopted the 1316 pack format to produce compatible drives, fostering an ecosystem of interchangeable media that persisted into the . This removable pack concept laid the groundwork for subsequent storage solutions, such as the 2311 used with System/360 systems.

IBM 2310

The IBM 2310 Disk Storage Drive was introduced in 1964 as part of IBM's effort to provide affordable (DASD) options for smaller computing systems, initially announced alongside the IBM 1800 Data Acquisition and Control System on November 30, 1964. It was later integrated with the IBM 1130 Computing System, which was announced in 1965 and targeted scientific, , and applications requiring storage beyond the limitations of . Unlike larger mainframe drives, the 2310 was designed for environments, offering a compact, cost-effective alternative that expanded disk technology to non-mainframe markets. The 2310 featured a removable 14-inch oxide-coated disk cartridge (IBM 2315), marking the first use of such a cartridge-based in HDDs and enabling easy media interchange for offline storage. Each cartridge provided a capacity of approximately 1.02 MB (512,000 sixteen-bit words) on its two recording surfaces, organized into 200 with two tracks per and four sectors per track, using double-frequency recording at 1,080 bits per inch and 100 tracks per inch. The drive employed a voice-coil actuator for head positioning, achieving seek times of 15 ms to an adjacent and up to 150 ms for 20- movements, with an additional 20 ms stabilization delay before read/write operations; rotational speed was 1,500 RPM, contributing to an average latency of about 20 ms. Models included the B1 (single drive) and B2 (dual drive), with up to four drives attachable to a single system for a maximum online capacity of 4 MB, and transfer rates of 720,000 bits per second. This integrated seamlessly with the host system's channel, often alongside tape units for hybrid sequential and workflows. In applications, the 2310 supported engineering computations, for small businesses, and process control tasks on the 1130 and , where its capabilities—contrasting the sequential nature of tape—enabled faster retrieval of program overlays, data files, and results in resource-constrained environments. For instance, on the 1130, it facilitated Disk Monitor operations requiring at least one drive for core storage extensions up to 32K words. The drive's significance lay in democratizing magnetic disk technology for midrange systems, reducing reliance on punched cards or tapes and paving the way for subsequent small-system storage solutions like the IBM 5440.

System/360 Mainframe Drives (1964–1974)

IBM 2311

The IBM 2311 Disk Storage Drive was introduced in 1964 as the standard direct-access storage device for IBM's System/360 mainframe family, marking a key advancement in modular, removable media storage. It utilized the same IBM 1316 disk packs as the earlier 1311 drive, consisting of six 14-inch platters in a 4-inch high stack weighing 9.4 pounds, which allowed for easy interchangeability and offline data transport across systems. This compatibility facilitated a smooth transition for users upgrading to the System/360 architecture, while the drive's design supported up to eight units per 2841 Storage Control Unit, enabling configurations of up to 58 MB total capacity per channel. Key specifications included a storage capacity of 7.25 million 8-bit bytes per in the Model 1 configuration, with data organized across 200 cylinders and 2,000 tracks using double-frequency (NRZ) encoding for improved density and reliability. The packs rotated at 2,400 , yielding an average rotational latency of 12.5 milliseconds, while the hydraulic provided an average seek time of 75 milliseconds and a maximum of 135 milliseconds, resulting in typical full access times around 87.5 milliseconds. Data transfer rates reached 156,000 bytes per second, and the heads operated on an at 125-160 microinches above the surface, with retractable assemblies to protect media during pack removal. Innovations in the 2311 centered on its tight integration with the System/360's (I/O) channel architecture, allowing seamless attachment to models like the System/360 Model 40 and supporting block-multiplexed operations for efficient data handling. were enhanced through track error-flagging mechanisms, which flagged defective sectors and automatically shifted to three spare cylinders per pack, alongside check digits for during read/write operations. The drive's standardized interface also spurred the plug-compatible manufacturers (PCM) industry, with competitors like and producing compatible units by the late 1960s, fostering market competition. The 2311 became the core storage solution for IBM's OS/360 operating system, powering and in enterprise environments worldwide, including major centers for banking, , and scientific applications. Its modular design enabled scalable expansion in System/360 installations, handling critical workloads until it was eventually superseded by the higher-capacity IBM 3330 in the early 1970s.

IBM 2314 and 2319

The 2314 Direct Access Storage Facility marked a major evolution in modular for mainframes, with significant enhancements announced in 1969 for the A-series models. These models supported configurations of up to eight removable 2316 disk packs in Model 1, each providing 29 million bytes of capacity, while Model 2 incorporated fixed heads and a removable disk, achieving an average access time of 60 ms for the facility in high-speed operations. The facility employed a string architecture, where a single managed multiple drives with shared actuators, enabling efficient parallel access and reducing overhead in multi-drive setups. Introduced alongside System/360 in 1965 but refined in 1969, the 2314 facilitated scalable storage growth by allowing hot-swappable disk packs, which contained 11 platters spinning at 2,400 RPM with 20 read/write heads per actuator assembly using technology for improved reliability over prior metallic cores. This design supported average access times of 60 ms in standard configurations, prioritizing direct-access performance for batch and interactive workloads on System/360. The IBM 2319, announced in 1971 as a compatible variant for the newly introduced System/370, built on the 2314 foundation with upgraded electronics that maintained the 312,000 bytes per second data transfer rate—doubled from the original 2314—for compatibility with System/370's enhanced channel architecture and faster I/O throughput. Available as a three-drive module attachable to 2314 control units, the 2319 maintained the 29.17 million byte capacity per drive while integrating seamlessly with System/370's enhanced channel architecture, supporting configurations from 87 to 233 million bytes total. This string-based approach with shared control and actuators optimized resource utilization across drives, making the 2314 and 2319 ideal for high-volume in enterprise environments, such as banking and systems. Their impact extended to OS/370 implementations, where the drives served as high-throughput paging devices, improving system responsiveness by accelerating data swapping between main memory and secondary storage. The 2314 and 2319 paved the way for subsequent advances, such as the transition to larger-capacity packs in the 3330 series.

IBM 3330

The IBM 3330 Direct Access Storage Facility, code-named , was introduced in June 1970 as a high-performance removable system designed specifically for IBM's System/370 mainframes. It represented a major leap in storage capacity and reliability for enterprise environments, quadrupling the density of its predecessor, the IBM 2314, through innovative engineering that addressed limitations in head positioning and . The system consisted of the 3333 and , individual 3330 disk storage units, and 3336 disk packs, enabling scalable configurations that supported the growing demands of virtual storage computing. Key specifications included a capacity of 100 MB per 14-inch for the Model 1 (3330-1), which utilized 19 recording surfaces and 404 tracks per surface recorded at 192 tracks per inch (TPI). In 1973, IBM released double-density versions, including the 3330-11, which increased capacity to 200 MB per pack via (MFM) encoding and a higher track density of 384 TPI. The drives spun at 3,600 rpm, delivering an average access time of 30 ms (comprising 25 ms average seek time and 8.3 ms rotational latency) and a data transfer rate of 806 kB/s. Up to 8 drives could be daisy-chained in a single string via the 3333 , providing up to 800 MB (Model 1) or 1.6 GB (Model 11) of online storage while maintaining compatibility with System/360 Model 195. Advancements in the 3330 focused on precision and durability, including the first implementation of a track-following servo system that used a dedicated servo surface to dynamically adjust head position, reducing errors from spindle wobble and . This was complemented by voice-coil actuators for faster, more reliable seeking compared to hydraulic mechanisms, and glass-bonded ferrite read/write heads that flew at a reduced 50 microinch height for better signal quality. Integrated error correction codes (ECC) further enhanced reliability by detecting and correcting single-bit errors on the fly, minimizing data loss and maintenance needs in mission-critical operations. In usage, the 3330 became integral to enterprise computing, powering database management systems like IBM's Information Management System (IMS), which relied on its high-capacity packs for hierarchical and rapid in banking, airlines, and government applications. Its removable packs facilitated data transport and backup, while the system's modularity allowed seamless integration into data centers, supporting the transition to architectures. The 3330's servo technology and head improvements directly influenced the development of sealed, fixed-pack designs in the subsequent IBM 3340 drive.

Winchester and Fixed-Pack Advances (1973–1980)

IBM 3340

The IBM 3340, code-named after the 30-30 rifle to reflect its initial 30 MB capacity per spindle and 30 ms access time target, was announced on March 13, 1973, as the first direct access storage facility for mainframes. Developed over four years starting in 1969, it introduced the sealed head/disk assembly concept, enclosing read/write heads, disks, and actuator in a removable 3348 data module to maintain a contamination-free environment. Shipping began in November 1973, marking a shift from open disk packs to protected modules that minimized dust-related failures. Available in Model 1 (35 MB formatted capacity) and Model 2 (70 MB), the 3340 utilized 14-inch platters rotating at 3,000 RPM within each sealed module, achieving an average access time of 25 ms and a data transfer rate of 885 KB/s. Each module contained 11 platters with 20 recording surfaces, formatted at 300 tracks per inch, and supported up to eight drives per string via a 3344 . Optional fixed heads reduced access to 5 ms for select tracks, enhancing performance for high-demand applications. Key innovations included lubricated disk surfaces to enable low-mass, low-load landing heads that operated via contact start/stop on a dedicated while flying on an over data areas, drastically cutting wear and maintenance needs compared to prior removable packs. The sealed, removable design prevented particle contamination that plagued earlier drives, allowing reliable operation without frequent cleaning or head adjustments. These features established the Winchester architecture as a foundational technology for subsequent sealed drives. The 3340 saw rapid adoption in System/370 installations, particularly for low-end models like the 115, reducing downtime and service costs in data-intensive environments such as banking and government operations reliant on mainframe reliability. Its success, with over 100,000 units shipped by the late 1970s, prompted plug-compatible manufacturers to emulate the design, accelerating an industry-wide transition to sealed, high-reliability disk systems. This influence extended to later fixed-pack advancements like the IBM 3350.

IBM 3350

The IBM 3350 Direct Access Storage Facility, code-named during development, was introduced in 1975 and first shipped in 1976 as a high-performance, non-removable system designed for mainframes. It featured sealed head-disk assemblies (HDAs) with fixed platters, building on the Winchester technology pioneered in the IBM 3340 by enclosing the read/write heads and lubricated disks in a contamination-free environment to enable lower head flying heights and higher recording densities. This design eliminated the need for removable disk packs, reducing handling errors and improving reliability for enterprise environments, while supporting compatibility with System/370 models such as the 135, 138, 145, 148, 155-II, 158, 165-II, 168, and their variants. The 3350 was available in models A2, A2F, B2, B2F, C2, and C2F, with the A2 and B2 offering standard configurations of up to 317 million bytes per spindle in native mode (or approximately 200 million bytes in 3330 compatibility mode), achieved using 16 recording surfaces across 11 platters accessed by 30 heads. Key technical attributes included an average seek time of 25 milliseconds (with a maximum of 50 ms and minimum of 10 ms), an average rotational latency of 8.4 ms at 3,600 RPM, and a data transfer rate of 1.198 MB per second, powered by a actuator for precise head positioning. Models with the "F" suffix (A2F, B2F, C2F) incorporated fixed-head features providing up to 1.144 million bytes of zero-seek-time storage per drive, enhancing performance for frequently accessed data. Notable features included a track density of 478 tracks per inch and a linear recording density of 6,425 bits per inch, which supported efficient packing through defect-skipping mechanisms and rotational position sensing for precise . The used tri-bit encoding for servo to maintain accurate track following, contributing to its overall reliability in closed-loop operation. Subsystems could scale to four modules housing eight drives, delivering up to 2.5 GB of total capacity when integrated with controllers like the IBM 3830 Storage Control. The 3350 played a pivotal role in advancing (OLTP) applications on System/370, such as database management and , by providing faster access times and higher throughput compared to prior removable-pack drives, often paired with cache controllers for optimized I/O performance. Its fixed-platter design influenced subsequent developments, serving as a precursor to smaller form factors like those in the IBM 3370 half-inch drives.

IBM 3370 and 3375

The 3370 , introduced in 1979, represented a significant advancement in midrange for 's System/38 and 4300 series processors, such as the 4331 and 4341 models. It featured seven fixed 14-inch platters in a sealed head-disk assembly (HDA), providing non-removable storage with capacities of 571 MB per unit across models A1 (controller-integrated) and B1 (drive-only, supporting up to three per string). This design emphasized reliability through a sealed environment that protected against contamination, combined with dual independent actuators per HDA for concurrent access to different data areas, enabling average seek times of 20 ms and data transfer rates up to 1.4 MB/s. The 3370 was the first drive to incorporate thin-film head technology, which used photolithographic to create more precise and durable read/write heads, improving recording and signal quality over traditional ferrite heads. Building on the 3370's architecture, the IBM 3375, announced in 1980 as a count-key-data (CKD) compatible evolution, targeted broader System/370 and 4300 compatibility while enhancing capacity and performance for flexible configurations in space-constrained environments. It retained the sealed HDA with fixed 14-inch platters but increased unformatted capacity to 819.7 MB per unit (models A1, B1, and D1 with dual controllers), allowing strings up to 3.279 GB, with average seek times reduced to 19 ms and transfer rates of 1.859 MB/s. Key improvements included rotational position sensing for better synchronization and defect management via surface skipping, supporting up to four units per 3880 storage control for shared access in midrange setups like those used in departmental computing or telecommunications switching systems. The half-height rack-mountable design of the 3375 facilitated easier integration into compact mainframe cabinets, contrasting with bulkier predecessors and enabling transportable configurations for distributed processing needs. These drives prioritized Winchester-style fixed-media reliability in sealed modules, avoiding the handling risks of removable packs while offering hot-pluggable string switching in supported controllers for minimal during reconfiguration. Their adoption in systems underscored IBM's shift toward higher-density, lower-maintenance storage for non-enterprise applications, paving the way for subsequent fixed-pack advances like the 3380.

IBM 3380

The IBM 3380 direct access storage device was announced by on June 11, 1980, and first shipped in October 1981, designed for compatibility with System/370 and 4300 mainframe processors. It represented a significant advancement in mainframe storage, featuring oversized 14-inch platters to achieve higher capacities than predecessors like the 3350. The drive utilized two head/disk assemblies (HDAs) per unit, each containing seven platters, enabling sealed, contamination-free operation. Key specifications included capacities ranging from 2.52 gigabytes per unit in early models (equivalent to approximately 2.52 billion characters) to up to 4 gigabytes in later variants like the 3380K, with strings supporting up to 10.08 gigabytes across four units. Average seek time was 16 milliseconds, complemented by an average rotational latency of 8.3 milliseconds at 3,600 RPM, allowing for efficient data access in demanding environments. Each access mechanism handled 630.2 megabytes, with 15 read/write heads per assembly using thin-film inductive technology for improved density and reliability. The device attached via the IBM 3880 storage to block multiplexer channels, supporting data rates up to 3 megabytes per second. The 3380's design emphasized modularity, with field-installable upgrades allowing users to convert between models (e.g., from single to dual controllers) without full replacement, facilitating seamless expansion. It incorporated symmetrical dual actuators per HDA for balanced performance and employed a 27-run-length-limited (RLL) encoding scheme, which optimized flux changes to boost linear density and served as a precursor to partial-response maximum-likelihood (PRML) methods. Error correction via embedded codes protected against burst errors up to four bytes, enhancing data integrity. This drive proved instrumental in managing massive datasets for financial and transaction-processing applications, offering cost-effective high-capacity storage that reduced reliance on tape for archival needs. Its robust performance contributed to 's revenue, with production continuing until 1991 and installations remaining operational into the 1990s. The 3380 was later superseded by the IBM 3390, which offered higher capacities and improved performance for mainframe storage.

High-End Mainframe Drives (1980–1994)

IBM 3390

The IBM 3390 direct access storage device (DASD) was announced on November 14, 1989, as a high-capacity disk storage solution for IBM mainframe systems, including the System/370 and the flagship 3090 series. Originally slated for a July 1989 release, the introduction was delayed four months due to technical challenges with the disk platter material, which affected surface reliability. Designed for enterprise environments requiring robust, high-performance storage, the 3390 marked a key evolution in IBM's mainframe DASD lineup, emphasizing scalability and data integrity for large-scale data processing. It interfaced through IBM's proprietary storage control units, such as the 3990, connected via parallel channel attachments like bus-and-tag or later ESCON fiber optics. The 3390 series offered a range of models with varying configurations to meet different capacity needs, from the single-spindle Model 1 at 1.89 GB to the high-density Model 9 at approximately 11 GB unformatted capacity, based on 10,017 cylinders and leveraging advanced recording techniques. Multi-spindle models, such as the Model 3 with three spindles, provided up to about 5.7 GB, while a full subsystem could house up to six drives for a total of 22.7 GB. Performance specifications included an average seek time of 12.5 milliseconds, a spindle speed of 4,260 RPM, and a sustained data transfer rate of 4.2 MB/s over 15 data surfaces per spindle using 10.8-inch platters. Redundancy was enhanced through the 3990 control unit's dual-copy feature, which mirrored data across paths for fault tolerance and continuous availability in mission-critical setups. The architecture supported up to 32 devices per control unit, enabling expansive storage arrays for complex workloads. Key technological advancements in the 3390 included the use of thin-film inductive heads, which improved signal strength and allowed for higher linear recording densities compared to earlier ferrite-core designs introduced in prior IBM drives. This head technology, refined from 1979 developments, enabled more precise data writing on the thin-film disk media. The drive also employed zoned constant angular velocity (CAV) recording, varying the number of bits per zone across the platter radius to optimize areal density and overall capacity without sacrificing performance uniformity. These innovations contributed to a 20% improvement in data storage and handling efficiency for the 3090 mainframe ecosystem. The IBM 3390's legacy endured well into the 2000s, particularly in sectors demanding high reliability, such as banking, where early customer testing involved institutions like the Bank of Boston. Production began in 1989, with initial shipments reaching nearly 200 units by announcement month, and the model became a cornerstone of mainframe storage, influencing modular RAID-integrated systems like the IBM 9340. Modern mainframe storage subsystems, such as the IBM 2105, continue to emulate the 3390's count-key-data (CKD) format for backward compatibility, supporting capacities up to 11 TB per emulated volume while preserving its architectural principles.

IBM 9340 and 9345

The 9345, introduced in 1991 as part of the channel-attached count key data (CKD) DASD subsystem family, served as the primary controller for high-end mainframe environments, offering compatibility with the 3390 interface to facilitate seamless integration with existing systems. This controller supported configurations ranging from 2 GB to a maximum of 486 GB in array setups, utilizing 5.25-inch form factor drives with magneto-resistive heads for improved data density and access times of 10-11 ms. The 9340 encompassed the broader subsystem architecture, including the and associated power and cooling components such as the 9344 unit, which provided reliable power distribution to support the 9345 controllers and attached storage drawers. The design emphasized redundancy through implementations in its cached storage drawers, achieving RAID-5 equivalents via parity-based protection and techniques to ensure availability during failures. Key features included dual-cluster configurations with up to four system adapters and two device adapters per cluster, enabling support for up to eight host channels and 64 logical DASDs, along with cache memory for up to four simultaneous writes and eight reads. High-availability aspects were central to the 9340/9345 design, incorporating hot-swappable components in the storage drawers to minimize downtime, comprehensive data integrity verification through mechanisms, and tight integration with /ESA for optimized I/O operations in enterprise workloads. These subsystems were particularly vital for mission-critical, 24/7 applications in industries such as and airlines, where uninterrupted access to large-scale data arrays was essential for and operational continuity. The 9340 family complemented standalone high-end drives like the 9330 series by providing scalable, controller-managed arrays for demanding mainframe installations.

9330 Family

The 3390 series extensions represented IBM's advancements in high-end mainframe during the early 1990s, building on the original 3390 design for the ES/9000 processor line and maintaining compatibility with System/370 architectures to support growing enterprise storage demands. Later models, such as the 1993 Model 9 extension, increased capacity to up to 27 GB per unit through additional cylinders (up to 32,760), enabling larger data repositories for applications like financial processing. Performance improvements included average seek times around 10 ms, optimizing retrieval in high-throughput mainframe environments. A significant advancement was the integration of magnetoresistive (MR) read heads in compatible subsystems, enhancing signal detection for higher densities while maintaining reliability. Interfaces utilized ESCON channels for transfer rates up to 17 MB/s, with support for parallel attachments in earlier configurations. These developments achieved areal densities around 0.5 Gbit per using MR technology and error-correcting codes, representing a substantial increase over prior generations. Fault-tolerant features, including dual power supplies, mirrored controllers, and dynamic path switching, supported no-single-point-of-failure operations essential for continuous mainframe use. The extended 3390 models bridged traditional mainframe systems with emerging requirements, deployed in environments like supercomputing for parallel processing.

Midrange and Small System Drives (1960s–1980s)

IBM 5444

The IBM 5444 Disk Storage Drive was introduced on July 30, 1969, alongside the , a system designed specifically for applications such as , inventory management, and tasks previously handled manually or with punch cards. Developed at IBM's Hursley Laboratory in under the code name , the 5444 marked an early effort to bring affordable, integrated to non-technical users in small and medium-sized businesses (SMBs), enabling to without the need for larger mainframe systems. The 5444 featured a compact design with two 8-inch magnetic platters—one fixed and one removable via the lightweight 5440 Disk Cartridge weighing about 6 pounds—mounted in a sliding drawer for easy desktop integration within the System/3 cabinet. It provided a storage capacity of up to 4.92 million bytes (approximately 5 MB) in higher configurations with 200 tracks per surface, using four read/write heads (one per surface) and an integrated controller that handled data formatting, error checking, and interface to the System/3 processor. The drive spun at 1,500 RPM, yielding an average rotational delay of 20 milliseconds, while average seek time was 126 milliseconds, supporting efficient access for business files organized sequentially, sequentially indexed, or directly. Its low power consumption and small footprint—fitting alongside printers and card readers in a single unit—made it suitable for office environments without dedicated computer rooms, and it supported RPG II programming for straightforward application development. The 5444 played a pivotal role in popularizing magnetic among SMBs by offering reliable, for master files, customer records, and inventories, which accelerated speeds compared to tape or cards and reduced operational costs. Integrated into over 20,000 System/3 installations by the mid-1970s, it helped capture a significant share of the emerging market for business automation, paving the way for later midrange evolutions like the IBM 62GV.

IBM 62GV

The IBM 62GV, code-named Gulliver, was a pioneering small-capacity hard disk drive developed by IBM at its Hursley Laboratory in the United Kingdom and first shipped to customers in May 1974. It was integrated into midrange systems such as the IBM System/32, announced in 1975, and the IBM System/34, introduced in 1977, providing non-removable direct access storage device (DASD) capabilities for small business computing. The drive featured a single 14-inch platter and introduced the rotary actuator design, which improved positioning accuracy and became a foundational technology for subsequent hard disk drives in the 1980s. Initial models offered 5 MB of storage, with later variants increasing to 10 MB and 14 MB through enhancements in recording density, achieving 167 tracks per inch and 5,650 bits per inch. By the time of its integration into the System/32, capacities aligned with system options of 5 MB, 9 MB, or 13 MB, supporting efficient data handling for entry-level configurations. The drive's design emphasized reliability for standalone or multi-station setups, with over 177,000 units shipped through early 1990. In applications for retail and environments, the 62GV enabled processing of , , and tasks on the System/32 and System/34, where data from input devices like MICR readers could be directly loaded for batch operations and reporting. These systems incorporated automatic backup mechanisms to diskettes, ensuring for critical daily operations without requiring manual intervention. The 62GV's compact integration made DASD accessible beyond large mainframes, facilitating scalable storage for growing midrange workloads. The significance of the 62GV lay in democratizing magnetic for , allowing smaller organizations to adopt reliable, expandable DASD solutions that influenced the evolution toward later systems like the AS/400. It served as the basis for OEM adaptations, such as the STC 8800 Super Disk, extending its technology to third-party markets.

OEM and PIC Drives (1970s–1990s)

066x and 067x Series

The 066x and 067x series marked IBM's expansion into the OEM market for magnetic disk drives, targeting non-IBM systems during the 1980s. These drives were designed as component products to enable third-party manufacturers to integrate reliable storage without purchasing complete systems, thereby broadening IBM's influence in the emerging midrange computing sector. The series emphasized compact, cost-effective designs suitable for embedded applications and workstations, leveraging advancements in technology to seal the head and disk assembly for reduced contamination and improved . The initial models, the 0665 and 0667, were introduced in 1985 and 1986, respectively, with formatted capacities of 20 MB (0665-30), 30 MB (0665-38), and up to 44 MB in other 0665 variants, in a 5.25-inch form factor. These drives utilized the ST-506 interface for the 0665 (MFM) and ESDI for the 0667 (around 70 MB), standards for early Winchester-based units that facilitated compatibility with diverse host systems. Targeted at minicomputers from vendors like DEC and HP, the 0665/0667 provided essential storage for data-intensive tasks in engineering and scientific computing environments, offering seek times around 50 ms and transfer rates of approximately 0.625 MB/s (5 Mbit/s formatted) to meet the performance needs of the era. In the , the series evolved with the 0669 model, launched in 1987 with 115 MB ESDI capacity, specifically oriented toward embedded systems and early workstations where space and power efficiency were critical. This model featured enhanced error correction and thermal compensation to ensure reliability in non-controlled environments. The OEM focus of the 066x and 067x series allowed to capture in the burgeoning peripheral component industry without direct competition in full system sales. The lineup was later extended by the 068x models to address demands for even higher capacities.

068x Models

The 068x models represented 's targeted OEM offerings for small systems, computers, and emerging PC/ markets during the late to early , emphasizing compact designs and improved performance for third-party integration. The series debuted with the 0680 Piccolo in 1979, a 64.5 MB drive utilizing six 8-inch platters and marking one of the first uses of a swing-arm mechanism, which allowed for faster seek times and more reliable head positioning compared to earlier voice-coil designs. This model was optimized for 's 4300 series processors but was available under OEM terms to support broader adoption in business computing environments. By the late , the lineup evolved to address growing demands for higher in open-market storage, culminating in the 0681 , introduced in 1990 with capacities reaching 857 MB formatted (using up to 12 platters in a full-height 5.25-inch form factor). Developed at 's Hursley Laboratory in the UK starting in 1983, the 0681 achieved an areal of 45.2 Mbit per square inch through innovations like partial-response maximum-likelihood (PRML) recording, the first implementation in a commercial , which employed advanced to mitigate and boost bits per inch without increasing error rates. These drives featured interfaces compatible with standards for the 0681, enabling seamless integration into servers and workstations, while earlier 068x variants supported MFM/ESDI protocols suited to PC architectures. Ruggedized construction in models like the 0681, with enhanced vibration tolerance and thermal management, made them suitable for industrial and CAD applications requiring reliable operation in non-desktop settings. The series' focus on OEM supply, including components for subsystems like the 9336 , underscored IBM's strategic shift toward competitive open-market participation beyond proprietary mainframes. This progression in the 068x line laid foundational technologies that preceded IBM's consumer-oriented Star series drives in the mid-1990s.

Floppy Disk Drives (1971–1990s)

8-Inch Floppy Drives

The development of IBM's 8-inch drive began in 1967 at the company's San Jose Research Laboratory, where a small team of engineers led by David L. Noble, under the direction of product manager and with Warren L. Dalziel as the lead inventor, worked on Project Minnow to create an inexpensive, reliable alternative to punched cards for loading into mainframe systems like the IBM System/370. The project addressed the need for a flexible, removable storage medium that could simplify and in environments. IBM first shipped the 8-inch floppy disk drive in 1971 as the model 23FD, known internally as the "Minnow," integrated into systems like the 3740 Data Entry System for mainframe data loading. This drive utilized an 8-inch flexible disk made of Mylar coated with , operating at 300 rpm with read/write heads that accessed data on a single-sided surface, providing a formatted capacity of 80 KB—equivalent to about 3,000 punched cards. Key features of the 8-inch floppy drive included its low-cost , housed in a protective with a dust-wiping mechanism for easy handling, making it suitable for program loading, updates, and backups in business and computing applications. Unlike rigid disks, the flexible design allowed for compact storage and transport, reducing reliance on cumbersome systems. The introduction of the 8-inch floppy standardized a versatile removable storage format that facilitated data interchange across systems, paving the way for an independent and profoundly influencing personal computing by enabling affordable, portable data handling. This original large-format design was later adapted into smaller sizes for greater portability in emerging markets.

Smaller Form Factor Floppies

The development of smaller form factor floppy disk drives marked a significant shift toward more compact, affordable storage solutions suitable for personal computers and portable systems, building on IBM's foundational 8-inch design. In 1981, IBM incorporated 5.25-inch floppy drives into its inaugural personal computer, the IBM PC Model 5150, which typically featured two such drives for booting and data exchange. These drives initially supported single-sided double-density (SSDD) media with a formatted capacity of 160 KB per disk, enabling efficient software distribution and in desktop environments. By the mid-1980s, advancements in recording technology allowed for double-sided double-density (DSDD) formats on 5.25-inch disks, increasing capacity to 360 KB, as seen in upgraded IBM PC configurations. The IBM PC/AT, introduced in , further elevated densities with high-density (HD) 5.25-inch drives capable of 1.2 MB per disk, utilizing enhanced error correction and track densities to support growing application sizes in business and professional use. These improvements complemented hard disk drives in systems by providing removable, low-cost media for backups and portability. The transition to 3.5-inch floppies addressed the needs of even smaller systems, with IBM adopting the format in its PS/2 line announced in 1987, marking the first widespread integration of 3.5-inch drives in its personal computers. Enclosed in a rigid plastic case for greater durability against physical damage, these drives initially offered double-density (DD) capacities of 720 KB, later advancing to high-density (HD) 1.44 MB through refinements in magnetic coating and head technology. This format gained traction in IBM's PS/2 models and portable laptops, such as the in 1986, where its compact size and shatter-resistant shell facilitated mobile computing. These smaller floppies evolved from double-density origins to high-density standards, becoming integral to IBM's for software installation and data interchange in PCs and portables throughout the and . Their legacy endures as a industry standard for removable storage until the rise of optical media like CDs in the late , with global production peaking at over 5 billion units annually in the mid-.

Star Series and Later Developments (1984–2003)

Deskstar and Ultrastar

The Deskstar series, introduced by in 1994, marked the company's entry into 3.5-inch hard disk drives targeted at desktop personal computers and consumer systems. The initial model, the DALA-3540, offered a capacity of approximately 541 MB, setting the foundation for a line that evolved rapidly through the to meet growing storage demands in PCs. Subsequent models expanded capacities progressively, reaching up to 20 GB by the late in drives like the 16GP and 75GXP, utilizing ATA (IDE) interfaces for broad compatibility with PC OEMs such as and . These drives featured rotational speeds of 5400 to 7200 RPM, enabling faster access times suitable for and general applications. A pivotal technological advancement in the Deskstar line came in 1997 with the introduction of Giant Magnetoresistive (GMR) heads in the 16GP model, which achieved areal densities up to 3 Gbits per square inch and capacities of 8.4 GB to 16.8 GB. This innovation, building on 's earlier magnetoresistive head research from 1990, allowed for significantly higher storage densities without increasing physical size, sustaining the industry's growth trajectory during the PC expansion of the late 1990s. However, the series faced severe reliability challenges with the 2000-launched 75GXP models (15 GB to 75 GB capacities), notorious as the "Deathstar" due to failure rates exceeding 30% in some deployments, attributed to design flaws in the disk suspension and GMR head integration. This led to widespread customer complaints and a class-action filed in 2001, which settled in 2005 by offering compensation up to $100 per affected drive without admitting liability. Complementing the consumer-focused , the Ultrastar series debuted in 1994 as IBM's enterprise-grade offering for servers and , with the first model, the DFHS, providing 4 GB capacity via interfaces optimized for configurations. Designed for reliability in data centers, Ultrastar drives supported levels through compatible controllers, featuring error-correcting codes and hot-swappable designs that enhanced uptime for mission-critical applications. By 1999, the Ultrastar 72ZX achieved a landmark 73 GB capacity at 10,000 RPM, incorporating GMR heads and glass platters for improved shock resistance and performance in arrays. These drives powered the server infrastructure during the late-1990s surge, where held approximately 35% of the overall hard drive market share, supplying OEMs like for web hosting and . Throughout the 1990s, both and Ultrastar lines paralleled IBM's parallel development of mobile storage in the Travelstar series, but focused on stationary desktop and server environments with higher capacities and enterprise features. Their adoption by major PC and server OEMs underscored IBM's dominance in the storage market until reliability issues and competitive pressures contributed to the sale of the HDD business to in 2003.

Travelstar

The IBM Travelstar series marked a significant advancement in mobile storage, debuting in as a line of 2.5-inch hard disk drives designed specifically for portable computing devices. The initial Travelstar LP models, shipped in October 1994, featured a low-profile height of 12.5 mm with two platters, offering capacities of 360 MB, 540 MB, and 720 MB, and data transfer rates up to 11.1 MB/sec, making them the fastest 2.5-inch drives available at the time. These drives incorporated innovations from IBM's Viper prototyping project, including a deep dish flat cover design for enhanced aerodynamics and high track densities to support compact form factors. Key innovations in the Travelstar series emphasized suitability for battery-powered systems, with low power consumption profiles such as 0.85 W in low power idle mode for later models, enabling extended runtime. Shock resistance was a core feature, with operating tolerances up to 10 G and non-operating up to 175 G, crucial for rugged mobile use. Early variants supported the PCMCIA interface, allowing integration as removable drives, such as the Travelstar 8E model, which facilitated easy expansion in notebooks without internal mounting. The Travelstar drives found primary application in IBM's laptops, powering models like the 600 and 700 series, and were adopted in other portable systems to enable on-the-go data access, including mobile databases and office productivity suites. By 2000, capacities had scaled to 32 GB in models like the Travelstar 32GH, supporting richer and larger datasets in mobile environments while maintaining the series' focus on reliability for nomadic users. This progression contributed to the broader series' evolution before its eventual decline amid market shifts. Early Travelstar models faced design challenges inherent to pioneering 2.5-inch technology, including elevated heat generation during sustained operation and occasional reliability concerns in high-vibration scenarios, though overall rates remained competitive for the era.

Transition and Sale to

During the 1990s, IBM's (HDD) division encountered intensifying competition from established players such as and , which captured larger market shares through aggressive pricing and innovation in consumer and enterprise storage. Concurrently, IBM reoriented its corporate strategy under CEO Louis Gerstner toward high-margin software, services, and consulting, viewing commodity hardware like HDDs as increasingly unprofitable amid and price erosion. By the early 2000s, the HDD unit reported substantial losses, including $423 million in 2001 and $92 million in the first quarter of 2002, with further losses in the second quarter, prompting IBM to explore divestiture options. In April 2002, IBM and Hitachi announced a strategic alliance to combine their HDD operations into a joint venture, with Hitachi acquiring the majority of IBM's assets for $2.05 billion; this culminated in the full transfer on January 3, 2003, forming Hitachi Global Storage Technologies (HGST). The deal included IBM's manufacturing facilities, intellectual property, and approximately 3,900 employees, positioning HGST as the second-largest HDD maker behind Seagate with projected 2003 sales of $5 billion. IBM's final HDD releases in 2002 included the 180GXP, a 3.5-inch desktop model offering up to 180 GB capacity at 7,200 RPM, and the Travelstar 40GN, a 2.5-inch mobile drive with 40 GB at 5,400 RPM, both leveraging giant magnetoresistive (GMR) head technology from prior Star series advancements. Following the sale, carried forward IBM's core technologies, including thin-film heads and advanced recording methods, to produce subsequent generations of drives until its acquisition by in 2012; , in turn, fully withdrew from HDD design and manufacturing, concentrating on enterprise systems integration and services.

Non-Disk Magnetic DASD (1950s–1980s)

Magnetic Drum Storage

IBM's magnetic drum storage devices served as an essential precursor to random-access , providing intermediate-speed, that bridged the performance gap between high-speed but limited-capacity or core memory and slower bulk storage like . These units featured a rotating metal coated with ferromagnetic material, accessed via fixed read/write heads positioned over each track, enabling direct access without mechanical seeking. Typical access times ranged from 10 to 50 milliseconds, dominated by rotational latency, while data remained volatile and required constant power for retention. They played a critical role in high-speed buffering and temporary storage for early computing systems, facilitating real-time processing before the advent of fixed-head disks and full hard disk drives. An early example was the magnetic drum storage integrated with the computer, introduced in 1952, which provided secondary storage divided into two blocks of 1,024 words each for a total of 2,048 36-bit words (approximately 9 KB). This configuration supported buffering for scientific calculations on the 701, a vacuum tube-based system, marking one of IBM's initial forays into drum technology for random-access applications. By the mid-1950s, drums had evolved into main memory for mass-produced machines like the , where average access times reached 2.4 milliseconds, underscoring their utility in commercial . In 1962, IBM introduced the 7320 Drum Storage unit for the 7090 and 7094 systems, offering 830,000 bytes of online random-access capacity on a single magnetic module. Designed to complement core storage, the 7320 operated through the 7631 File Control, delivering data rates up to 202,800 characters per second in six-bit mode and serving as a high-speed intermediary for applications requiring faster access than disks or tapes of the era. With the launch of the System/360 architecture, advanced drum technology for paging and buffering in multiprogramming environments. The , introduced in the late for models like the 65, 67, 75, and 85, featured 200 logical tracks (each comprising four physical tracks) with 20,483 bytes per track, yielding a total capacity of approximately 4.1 MB per unit. Its 17.5-millisecond rotation time enabled average access latencies around 8.75 milliseconds, making it ideal for paging in high-end configurations, where up to two units could provide 8 MB combined for rapid data swapping. Configurations often paired the 2301 with the 2820 Storage Control for attachment to System/360 channels, enhancing throughput to 1.2 MB per second. Complementing the 2301, the IBM 2303 Drum Storage, announced in 1964 for lower-end System/360 models such as the 40, 50, 65, and 75, utilized 800 addressable physical tracks with 4,892 bytes each, for a capacity of about 3.9 MB. Unlike the 2301's multi-track transfers, the 2303 handled one physical track at a time at 312,500 characters per second, positioning it as a cost-effective cache-like buffer for mid-range systems. It attached via the 2841 Storage Control, supporting applications needing moderate-speed random access without the full performance of the 2301. Drum storage units like the 2301 and 2303 were eventually supplanted by semiconductor memory technologies and higher-performance disks in the 1970s and 1980s.

IBM 2321 Data Cell

The IBM 2321 Data Cell Drive, announced in April 1964 alongside the System/360 mainframe and first shipped in 1965, represented an innovative approach to high-capacity, removable bulk storage for large-scale data processing. Designed to extend online random access capabilities beyond traditional disk systems, it targeted applications requiring substantial archival volumes, such as data libraries or transaction logs too voluminous for fixed media. The core design centered on removable data cells housed in plastic cartridges, each containing 200 short magnetic strips made of Mylar coated with iron oxide. These strips, measuring 2.25 inches wide by 13 inches long, were organized into 20 subcells of 10 strips each for efficient selection. A single drive accommodated up to 10 interchangeable cells, providing a total capacity of 400 million bytes—or equivalently, 800 million packed decimal digits—per unit. Up to eight drives could connect to an IBM 2841 Storage Control Unit, scaling total online storage to over 3 billion bytes. Operation involved a hydraulic picker mechanism that selected a subcell, positioned it at an access station, and wrapped individual strips around a rotating read/write drum for data transfer at up to 1,000 characters per second. Access times varied by strip position: average selection latency ranged from 175 to 600 milliseconds, plus a 25-millisecond rotational delay and up to 95 milliseconds for cylinder seeking. This rotary transport system, originally developed under the codename MARS in the late 1950s, emphasized interchangeability to support off-line libraries of cells, enabling users to swap media for expanded archival capacity without system downtime. Despite its ambitious scale and role in supporting System/360's direct-access storage hierarchy—following earlier magnetic drum technologies—the 2321 faced significant reliability challenges from the outset. Particulate contaminants in the strip transport path degraded recording integrity, while mechanical issues like inconsistent clutches, noisy lamps, and difficulties in precisely reinserting strips into their clips after use led to frequent errors and strip damage. Early field deployments required extensive retrofits, including fiber optic sensors and viscous dampers, but stabilization took nearly a year, and overall dependability remained problematic compared to emerging disk drives. IBM withdrew the 2321 from marketing in January 1975, as it was obsoleted by more reliable and higher-density alternatives like the IBM 3330 disk drive. The device's automated strip-handling innovations, however, foreshadowed advancements in robotic tape libraries for mass storage systems.

IBM 3850 Mass Storage System

The 3850 Mass Storage System (MSS), announced in October 1974, was developed for use with mainframes to provide large-scale, automated storage for infrequently accessed data. It addressed the limitations of manual tape libraries and direct access storage devices (DASD) by introducing a hierarchical approach that extended virtual storage concepts to mass data management. The system evolved conceptually from earlier bulk storage ideas like the IBM 2321 Data Cell, but utilized tape cartridges instead of strips for higher density and automation. The architecture centered on the 3851 Mass Storage Facility, a robotic resembling a vault that housed up to 4,720 removable tape cartridges, each holding approximately 50 million bytes, for a maximum total capacity of 236 GB in initial configurations; a 1980 update doubled this to 472 GB. Cartridges were automatically retrieved by dual accessors (robotic arms) and staged to high-speed helical-scan tape drives in the 3850 Storage Control, emulating standard 3330 disk packs with up to 100 million bytes per emulated pack (two cartridges). This setup allowed data sets to be accessed via familiar OS/VS methods like BSAM (Basic Sequential Access Method), with the system handling staging transparently to minimize user intervention. Average access from the vault to staging took around 6-10 seconds, enabling near-online performance for archival volumes without manual handling. Primarily targeted at offline archival in sectors like banking, the 3850 supported applications such as storing historical transaction records or MICR adjustment data, where volumes could reach tens of billions of bytes without requiring constant online DASD space. Integration with BSAM and other access methods allowed seamless use in , treating mass storage volumes like removable DASD for and retrieval. For instance, banks automated tape handling to reduce operational overhead, supporting up to 1,728 emulated disk packs in larger installations for total capacities approaching 472 GB. Despite its innovation, the 3850 proved complex due to intricate microcode, mechanical reliability issues in the robotic accessors, and high costs, leading to limited adoption. By the mid-1980s, falling prices for DASD and advancements in cheaper technologies rendered it obsolete, with withdrawing the product in August 1986 and ending maintenance support in 1991.

References

Add your contribution
Related Hubs
User Avatar
No comments yet.