Introduction

aY 6,1955... Few could have guessed in 1955 It is only through hindsight that the International Business that the computer industry's first true dimensions of the 350 RAMAC MMachines Corporation made magnetic disk file, the IBM 350 disk file development can be an announcement that went largely RAMAC (Random Access Method of appreciated for what it was .. a unnoticed outside the computer Accounting and ControI), would one historic technological milestone. community and other technical day prove to be of worldwide circles. The company reported that a significance. team of engineers working in a small What the team of IBM engineers research and development laboratory had developed was a technology that in San Jose, California, had significantly affected information ~eloped a new magnetic disk processing in the worlds of science, storage technology. agriculture, health, education, government, finance, insurance, transportation and distribution. This technology ushered in a new era of interactive computer applications such as airline reservation systems, inventory management, automated banking, space flights, word processing and personal computing. A technology that spawned an industry

) he first 350 RAMAC disk file Removing these obstacles helped Prior to the development of became a commercially turn the promise of the computer RAMAC in the early 1950's, IBM Tavailable product on into reality and set the stage for what employed a small number of people September 4, 1956, and was a key has come to be called the at its San Jose, California, punched component of the IBM 305 RAMAC Information Age. card facility which had been system, which also included a The 350 was the first step in the established in 1943. Today IBM's central processor, card reader and evolution of many direct access General Products Division employs printer. (In its early development, the storage devices. It launched a some 15,000 people, with two major file itself was called the 305. It technology that has been improved sites in the San Jose area and a third became the 350 when the 305 and refined during the past three in Tucson, Arizona. system was announced.) decades, but never superseded. In The economic contribution of disk It is difficult to overstate the impact strictly economic terms, the 350 storage products to California's the 350' s disk technology has had launched a direct access disk economy is estimated to be upon the world in the years since its industry whose size goes far beyond approximately as great as California's announcement. disk drive hardware sales and into more famous semiconductor Making information directly the worlds of high-speed processors, industry. available for computer processing on programming and computer demand meant that no longer would services. There is no way to estimate, processors stand idle while searches for example, other hardware, were made through reels of software, and service revenues which magnetic tape or data was punched would not have been generated if into cards and sorted for processing. low-cost direct access disk devices did not exist. Even discounting such conjecture, it is sufficient to say that fIXed and flexible disk drives alone - all derivatives of the basic 350 technology -- generated an estimated $12.5 billion in sales worldwide in 1983 for the 72 manufacturers of fIXed disk drives and the 52 manufacturers of flexible disk drives.

Getting Started ...

he RAMAC file development one ot the areas suggested, another By July 1952, IBM's new San Jose could not have had more was data reduction. These could Research and Development Thumble beginnings. make up about 50 percent of the Laboratory was a functioning It all began with one man, Reynold new lab's work The rest was up to organization of some 30 people, B. Johnson. Johnson. many hired after only one interview, In mid-January 1952, Johnson, a One of the most remarkable working on a number of projects, former high school science teacher aspects of the 350 RAMAC with each engineer usually working from Ironwood, Michigan, who had development effort was the rapidity on more than one project at a time. been hired to help develop the IBM with which Johnson put his new Johnson set forth three guiding 805 test scoring machine, received a laboratory into full operation. principles to everyone hired, which visit in his office at the IBM Endicott, In February 1952, he: added to the vitality of the lab. They New York, development laboratory 1) signed a five-year lease on a were: from W. Wallace McDowell, then IBM vacant stuccoed cement-block • It is essential that each engineer director of engineering. McDowell building that had previously housed be familiar with the purpose, had come with an unusual offer. a printing plant at 99 Notre Dame function and environment of the The company had decided to set Avenue in San Jose; machine or machine component up a small research laboratory on 2) began to renovate the building; on which he is working to the the West Coast, and McDowell 3) placed ads recruiting engineers in degree that his work affects the wanted Johnson to head the project. area newspapers, and proper performance of the His job: to find a suitable site, do his 4) started interviewing applicants function in the ultimate own recruiting and establish a with the help of Louis D. Stevens, a environment laboratory of no more than 50 member of the Defense Calculator • It is the responsibility of every people. Design crew at Poughkeepsie, New engineer to be conversant with all The new lab was to work on York Stevens, whose assignment other projects going on in the technologies not being pursued in was first considered temporary, laboratory. the East. Non-impact printing was returned permanently in May 1952 • It is the most important as Johnson's technical assistant. assignment of every engineer in this laboratory to give assistance, in the form of consultation, experimentation or suggestions, when asked to by another engineer; and the second most important assignment is that of carrying forward the project to which he is assigned. Looking back today, it is hard to believe that within three years one of the computer industry's most important technologies would grow from the efforts of this tiny fledgling operation. Who could have thought so then?

The original building that housed the IBM Research and Development Laboratory. The first unit, on the comer, was 10,000 square feet in 1952, and was enlarged to about 18,000 square feet in 1953.

Identifying the Goal

he project that was to lead processing, which meant that Throughout this period and to development of the expensive central processing systems continuing through the entire TRAMAC disk file was called were often idle while information was RAMAC development project, IBM "Source Recording," which was being accumulated and sorted for engineers used the tub file billing defined as encompassing "all serial processing. and inventory control operations of a processes which take alphanumeric Information stored on drums was large paper company as a " real life" data from any source and transcribe also randomly accessible and touchstone in defining and refining it in a way so that the resulting moderately fast, but the low volume­ requirements and specifications for document may be handled by to-density ratio of the drum the Source Recording study and for machine methods." The punched technology also made it costly. the disk memory system they card was the most widely used such From today's vantage point, one eventually developed. document at the time, and early can see that what was needed was By November 1952, the tub file work on the project focused on some form of storage device that automation altematives had been eliminating the use of cards was randomly accessible with a narrowed to two approaches. One ahogether. As the problem became greater surface-to-area volume ratio was an endless belt with master better understood, however, the much higher than drums. cards attached so that operators focus turned to minimizing or In September 1952, however, the could see the ones they wanted and eliminating the key punching task focus was on automating the so­ have them electrostatically copied. itself. With many shifting goals to called punched card "tub file." Tub These master cards would then be come, the journey of discovery had files had come into widespread use electrically sensed into a new begun. at that time as a method for making punched card. The other··Johnson's It is helpful to remember that in master information punched on favored approach at that time-­ 1952 there were only three ways of cards more readily accessible for involved a matrix of parallel vertical storing information for use by data machine processing. Essentially, the wires of one-foot length that would processing equipment: the punched tub files were very large rectangular sense a card's information and card, magnetic tape and, to a lesser trays containing master cards record it magnetically. extent, magnetic drums. arrayed in sequence by customer It was at about this time that Each of these methods had number, item number, size, color, Jacob Rabinow, of the National significant limitations. Punched cards etc. Usually, the files contained Bureau of Standards, described a and magnetic tape essentially limited several copies of each master card. Notched Disk Memory Array in the user to batch or serial In a typical operation, clerks which each disk was rotated receiving an order would search the independently. His paper was widely tubs, pick out cards containing the circulated at the San Jose lab and needed customer and item order led a shift in interest back to disks. information, and send the cards to Disks had already been considered the machine room, where the earlier and discarded because needed documents--shipping room maintaining the necessarily minute instructions, packing slips, invoices, spacing between a recording head shipping labels and bills of lading.­ and a disk surface (about 1/ 1,OOOth were produced. of an inch) was considered an An early crusader for tub file insurmountable problem. Other automation and a frequent visitor to alternatives were magnetic cards, the San Jose lab was the late Ed plates, strips, bands, wires and rods. Perkins, a special marketing All were finally discarded. representative in IBM's San In computing's earliest days, punched Francisco office. cards were the basic means of storing He addressed the San Jose information and were often organized in engineers regularly on the "tub files". This file was typical of the inadequacies of the tub file method 1940's and early 1950's. The RAMAC and took many of them to visit Bay file development grew out of an effort Area IBM customers so that they aimed at automating the tub file. could see firsthand what needed to be corrected.

Betting on disks

n January 1953, after This all changed shortly to a more reevaluating all the alterna­ \ ~'( =E ___O _L_O_E-___ ?__ E-_ I ___ ~__ ~_ OP_P_t _~ -=~ I ambitious concept when the U.S. Air I tives, Johnson came to a Force asked IBM to submit a decision. The laboratory would focus proposal to provide an inventory its attention on disks. A work order control system for its base supply. for that month reads: The request called for a very-large­ "Magnetic Disks have been capacity, randomly-accessible selected as the best medium for the memory rather similar to what the Random Access Memory to be used San Jose researchers were trying to for File Maintenance. Disks will achieve, but with the additional revolve continuously at 16 r.p.s. so requirement for information that any of 200 columns of a record J processing capability. OollT OOfoG.', I~ . ... """",, 0 • ..,. may be read out in any sequence, The late John Haanstra and other from any of 20,000 records." laboratory systems people were Those who were there at the time 1953, the objectives for a disk assigned to respond to the Air recall that Johnson's disk decision storage configuration appeared. The Force's request and on April 1, was a very unpopular one. One configuration was remarkably close 1953, submitted a proposal to the engineer advised JohnsOlI Ulat he to the final product. While the U.S. Air Force that called for a was backing a mechanical folly. The decision to focus on disks clearly memory complex of 10 magnetic popular name for the disk array was replaced the goal of tub file drums that might later be replaced "the baloney slicer." automation, the general idea at first by disks. Despite all disparagement, was that disk memory would be part On April 14, 1953, shortly after including cartoons on lab bulletin of a File-to-Card Machine and submitting the Air Force proposal, boards, San Jose engineers went conceptually the end product would Johnson made the decision that the back to work with their focus be, in effect, an electronic tub file, if laboratory would abandon all narrowed to disks. On February 2, not an "automated" one. technologies competitive to disk He assigned three teams to work on various aspects of the project William A. Goddard was charged with developing a file model, assisted by Donald D. Johnson, John J. Lynott, Geoffrey Hotham and Warren Gonder. The late Edward Quade and his group were assigned to design a magnetic read-write head which could be no more than 1/1Oth of an inch high, and Haanstra and Alton E. Ewing were to concentrate on the File-to-Card Machine while also studying extended systems applications. It was agreed that one of the first problems to be surmounted was finding a way to maintain constant spacing between magnetic head elements and disks without Above Top substantial runout Two of the early read-write airheads Rey Johnson's early t 953 decision to (circa t 953) positioned as they would proceed with development of a disk file be against a magnetic recording surface. was the subject of much skepticism and Exhaust ports to remove air from disbelief. One of a number of cartoons between the surfaces are visible. These found on the lab buUetin board at the were necesscuy to achieve the required time refers to a popular song of that spacing between head and disk, then year: "How Much Is Dat Doggy In Da about one-thousandth of an inch. Wmdow'?" Goddard recommended an air­ Ralph Flores invented the final bearing approach and asked Don coating formulation and basically, it Johnson to try it out. Not only did is still in use today. this approach work, but by June 2, The third major hurdle-­ 1953, a third model of an air head development of disk-to-disk and with a magnetic read-write element track-to-track access mechanisms-­ designed for a magnetic drum was actually posed two challenges. used to write 51 bits per inch on an The first was how to apply the load aluminum disk that had been spray­ needed to make the air head coated with magnetic iron oxide function. Initially, the researchers paint tried to devise a method using Thus convinced that an air bearing springs to apply the load, but this would "float" a read-write head above caused problems in moving the arm the surface of a disk without crashing from disk-to-disk onto its surface, the development It was Norman Vogel who finally team tumed to three other pressing solved this problem with a design problems: creating disk surfaces flat that retracted the air head into the enough, developing an effective disk Bill Crooks produced the first truly arm during disk-to-disk motion of the coating and devising disk-to-disk and successful disks by filtering the carriage. The design provided a self­ track-to-track accessing mechanisms. dispersion through nylon hosiery and loading force using three miniature Finding the best method and using paper cups to measure out air pistons on the back of the air material for creating flat disk correct amounts for the spin coating. head that were activated in the track­ surfaces was a matter of old­ This primitive process was used for a to-track mode. fashioned trial and error. Among the year before it was automated. The second challenge was how to materials tried were aluminum, brass, Later in the development project, provide the disk-to-disk and track-to­ glass, plastics and magnesium. In Marcel Vogel, Don Johnson and track accessing motions. Three the sheet aluminum trial, engineers were horrified to find their so-called flat disks had runouts as high as 1/ 5Oth of an inch and greater when the disks revolved at 1,200 r.p.m. The first successful material tried was fabricated from magnesium lithographer's metal, but this was finally replaced by aluminum laminates clamped under pressure and heated in ovens above the annealing temperature. Jake Hagopian, assigned to the disk coating problem, developed specifications for a coating made from a paint base. Spray coating was tried as was dipping, but neither produced the uniformity of thickness required in the magnetiC coating. Hagopian finally settled on a spin method in which the coating was poured onto the inner surface of a Top Above rapidly rotating disk and then spread First test model, used with the "File-to­ Norman Vogel, one of the key members evenly over the surface of the disk Card" machine, showing heads on and of the 350 development team, at his inserted between disks. workbench testing an airhead design. teams were set up to explore Within 12 days, Lynott and Davis, It read, in part: ..... We must altematives: Crooks and Haanstra with assistance from Trigg Noyes, immediately... attack accounting were to work on an electrical-servo had completely redesigned a new problems under the philosophy of drive system; Don Johnson and R. servo system. handling each business transaction Manning Hermes were to pursue an Thus, on February 10, 1954, the as it occurs, rather than under the approach based on the IBM 402 type San Jose team was able to achieve present condition of batching bar mechanism, and Gonder was the first successful transfer of techniques ... " assigned a system using cams and information from cards to disks and "We must build storage and spring clutches. back. peripheral equipment which can The servo approach was the early This was a huge boost to morale spread out into individual accounts favorite for the track-to-track drive but for a brief period, but soon reality every business fact (random access was somewhat suspect for the began to set in. storage) and allow operation of a tougher job of moving the heavy Stevens remembers that the first new concept for handling business carriage from disk-to-disk. On laboratory model looked like a Rube information concurrently with its October 28, Davis was able to locate Goldberg arrangement and that" not inception... " a previously recorded track with his many people believed we would "I wish to recommend for your track-to-track servo, and in November make such a thing practicaL" consideration that we double or Lynott found a way to use the same ..... but it was like a religion to us. treble our efforts in this electrical servo system in a We were going to make that thing development...1 am firmly convinced multiplexing mode to drive the work for sure ... because if we that (otherwise) we cannot expect to carriage from disk-to-disk. failed ... the whole San Jose accomplish the real purpose and the On November 1, 1953, Rey experiment would fail. None of us real use for electronics in the Johnson turned entire responsibility were going to allow that to happen." business world." for the development of the magnetic It soon became obvious, however, The second boost came in disk assembly over to Lou Stevens, that neither the first model of the November 1954, in the form of who was promoted to senior disk file nor the File-to-Card Machine official Corporate sanction for engineer. was performing well enough to be RAMAC product development work. Things seemed to be going on demonstrated. So on March 19, The San Jose laboratory was nicely until one Saturday evening, 1954, an in-depth reevaluation of charged with designing, developing during a test, the new servo went out every basic design was undertaken and building several field-test models of control and rocketed free of the and specifications were drawn up for of a machine utilizing a disk random rail, landing in a heap on the floor. a revised Model II. access memory attached to a serial Trigg Noyes had been thinking of "stick" printer to provide an initial a new design for some time, and his system utilizing random access first decision was to make the shaft storage. The machine specified was holding the disks upright rather than given the name DRAM (for Direct horizontal, to provide more workable Reference Accounting Machine), and dimensions for the machine, make later the field-test models were disk replacement easier and provide known as the 305A. more spaces for a number of On January 10, 1955, the first of independent access mechanisms. five Model II 350 RAMAC files was Later that year, with construction successfully demonstrated--slightIy of the revised Model II well underway, less than three years from the San the development team got two Jose Laboratory's inception. A new tremendous boosts. era in information processing had The first came in the form of a begun. letter from F. J. Wesley, who had been assigned to review the progress of the work at San Jose and report Wesley Dickinson, associate engineer in t 955 when the photo was taken, points back on it. to the reading and writing ann of the redesigned 305 (the disk tile's unofficial designation while it was under development; the 350 designation came later).

Disk technology today

he rapid and widespread bytes or characters (8 bits per byte) the cost of leasing the same amount acceptance of disk of storage. The complete 3380 has of storage is about a dollar. This Ttechnology is a direct result two spindles with 2.5 billion drop in cost was in real dollars, of the vast improvements in access characters. unadjusted for inflation over the speed and capacity, as well as Today's microminiature read-write more than a quarter of a century in dramatic reduction in costs over the heads on the 3380 use thin-film which it took place. years since the 350 was first technology. They permit vastly All of these improvements and unveiled. increased density of storage, cost reductions have triggered a A comparison between the 350 file permitting much faster access time. huge demand for disk storage and IBM's largest current file, the For example, the original 350 disk devices. 3380, tells the story: file stored 2,000 bits of information During 1983, for example, it is The 350's 24-inch disks revolved per square inch and had an average estimated that the dollar volume of at a speed of 1, 200 revolutions per access time of 600 milliseconds. The rigid disk drives rose 26 percent, minute; the resulting data rate was IBM 3380 packs 12 million bits of while the dollar volume of flexible 100,000 bits per second. Today' s information per square inch and has disk drives increased some 110 3380 disks revolve at 3,600 r.p.m. an access time of 16 milliseconds. percent during the same period. and the data rate is 24 million bits Looked at another way, the 350 Looking back today, it's difficult to per second. technology stored one megabyte of believe that a relatively unnoticed The 350's fifty 24-inch disks information on an area the size of a announcement on May 6, 1955, contained a total capacity of 5 pool table. By comparison, the 3380 about a machine called the 350 million binary decimal encoded technology stores one megabyte of RAMAC could have caused such characters (7 bits per character) of information on an area almost the dramatic changes in the pace of storage. Today' s 3380, using nine size of a postage stamp. technology and in our lives. But that, 14-inch disks, offers 1.25 billion Meanwhile, the price of djsk data in fact, is exactly what happened. storage has tumbled dramatically over the years. Users of the 350 RAMAC file paid $130 a month to rent a megabyte of storage; today, Costs and access times down sharply.

25 $2.86 $0.94 16 1956 1965 1970 1975 1980 RAMAC 350 2314 3330-1 3350 3380

5 2,000 29.2 220,000

317

Capacity and areal density up.