The IBM350 RAMACDisk File

Designated AnInternationalHistoric Landmark byThe American Societyof Mechanical Engineers

February 27, 1984

Acknowledgements

The RAMAC disk is the 16thInternational Historic MechanicalEngineering Landmark to bedesignated since the ASME programbegan in 1973. In addition 66National and 6 Regional Landmarkshave been recognized. Eachrepresents a progressive step in theevolution of mechanical engineeringand each reflects its influence onsociety, either in its immediate locale,nationwide or throughout the world.

The landmarks programilluminates out technologicalheritage and serves to encourage thepreservation of the physical remainsof historically important works. Itprovides an annotated roster forengineers, students, educators,historians and travelers, and helpsestablish persistent reminders ofwhere we have been and where weare going along the divergent pathsof discovery.

The American Society of Mechanical EngineersFrank M. Scott, PresidentJames D. Woodburn, Vice President, Region IXJoseph P. Van Overveen, History and Heritage, Region IXPaul Allmendinger, Executive Director

The ASME National History and Heritage CommitteeDr. R. Carson Dalzell, ChairmanCurator Robert M. Vogel, Secretary (Smithsonian Institution)Dr. Robert B. GaitherProfessor Richard S. HartenbergDr. J. Paul HartmanProf. Merritt Roe Smith

The ASME Santa Clara Valley SectionG. I. Skoda, ChairmanJ. M. HealzerR. B. AgarwalF. SeddiquiMichael Hunt, History and Heritage

The Santa Clara Valley Sectiongratefully acknowledges theassistance of all who cooperated onthe landmark designation of theRAMAC disk, particularly theIBM staff at San Jose.

Introduction

May 6, 1955…International Business

Machines Corporation madean announcement that went largelyunnoticed outside the computercommunity and other technicalcircles. The company reported that ateam of engineers working in a smallresearch and development laboratoryin San Jose, California, haddeveloped a new magnetic diskstorage technology.

Few could have guessed in 1955that the computer industry’s firstmagnetic disk file, the IBM 350RAMAC (Random Access Method ofAccounting and Control), would oneday prove to be of worldwidesignificance.

What the team of IBM engineershad developed was a technology thatsignifïcantly affected informationprocessing in the worlds of science,agriculture, health, education,government, finance, insurance,transportation and distribution. Thistechnology ushered in a new era ofinteractive computer applicationssuch as airline reservation systems,inventory management, automatedbanking, space flights, wordprocessing and personal computing.

It is only through hindsight that thetrue dimensions of the 350 RAMACdisk file development can beappreciated for what it was -- ahistoric technological milestone.

A technology thatspawned an industry

The first 350 RAMAC disk filebecame a commerciallyavailable product on

September 4, 1956, and was a keycomponent of the IBM 305 RAMACsystem, which also included acentral processor, card reader andprinter. (In its early development, thefile itself was called the 305. Itbecame the 350 when the 305system was announced.)

It is diffïcult to overstate the impactthe 350's disk technology has hadupon the world in the years since itsannouncement.

Making information directlyavailable for computer processing ondemand meant that no longer wouldprocessors stand idle while searcheswere made through reels ofmagnetic tape or data was punchedinto cards and sorted for processing.

Removing these obstacles helpedturn the promise of the computerinto reality and set the stage for whathas come to be called theInformation Age.

The 350 was the first step in theevolution of many direct accessstorage devices. It launched atechnology that has been improvedand refined during the past threedecades, but never superseded. Instrictly economic terms, the 350launched a direct access diskindustry whose size goes far beyonddisk drive hardware sales and intothe worlds of high-speed processors,programming and computerservices. There is no way to estimate,for example, other hardware,software, and service revenues whichwould not have been generated iflow-cost direct access disk devicesdid not exist.

Even discounting such conjecture,it is sufficient to say that fixed andflexible disk drives alone -- allderivatives of the basic 350technology--generated an estimated$12.5 billion in sales worldwide in1983 for the 72 manufacturers offixed disk drives and the 52manufacturers of flexible disk drives.

Prior to the development ofRAMAC in the early 1950's, IBMemployed a small number of peopleat its San Jose, California, punchedcard facility which had beenestablished in 1943. Today IBM'sGeneral Products Division employssome 15,000 people, with two majorsites in the San Jose area and a thirdin Tucson, Arizona.

The economic contribution of diskstorage products to California’seconomy is estimated to beapproximately as great as California'smore famous semiconductorindustry.

Development of the 350 RAMAC file atSan Jose led to the company's decisionto build its first computer manufacturingplant in California. Above engineers testunits of the 350 RAMAC System beforecustomer shipment from the new facilityin 1957.

Getting Started...

T he RAMAC file developmentcould not have had morehumble beginnings.

It all began with one man, ReynoldB. Johnson.

In mid-January 1952, Johnson, aformer high school science teacherfrom Ironwood, Michigan, who hadbeen hired to help develop the IBM805 test scoring machine, received avisit in his office at the IBM Endicott,New York, development laboratoryfrom W. Wallace McDowell, then IBMdirector of engineering. McDowellhad come with an unusual offer.

The company had decided to setup a small research laboratory onthe West Coast, and McDowellwanted Johnson to head the project.His job: to find a suitable site, do hisown recruiting and establish alaboratory of no more than 50people.

The new lab was to work ontechnologies not being pursued inthe East. Non-impact printing was

one ot the areas suggested, anotherwas data reduction. These couldmake up about 50 percent of thenew lab’s work. The rest was up toJohnson.

One of the most remarkableaspects of the 350 RAMACdevelopment effort was the rapiditywith which Johnson put his newlaboratory into full operation.

In February 1952, he:1) signed a five-year lease on avacant stuccoed cement-blockbuilding that had previously houseda printing plant at 99 Notre DameAvenue in San Jose;2) began to renovate the building;3) placed ads recruiting engineers inarea newspapers, and4) started interviewing applicantswith the help of Louis D. Stevens, amember of the Defense CalculatorDesign crew at Poughkeepsie, NewYork. Stevens, whose assignmentwas first considered temporary,returned permanently in May 1952as Johnson’s technical assistant.

The original building that housed theIBM Research and DevelopmentLaboratory. The first unit, on the corner.was 10,000 square feet in 1952, andwas enlarged to about 18,000 squarefeet in 1953.

By July 1952, IBM's new San JoseResearch and DevelopmentLaboratory was a functioningorganization of some 30 people,many hired after only one interview,working on a number of projects,with each engineer usually workingon more than one project at a time.Johnson set forth three guidingprinciples to everyone hired, whichadded to the vitality of the lab. Theywere:l It is essential that each engineer

be familiar with the purpose,function and environment of themachine or machine componenton which he is working to thedegree that his work affects theproper performance of thefunction in the ultimate

lenvironment.It is the responsibility of everyengineer to be conversant with allother projects going on in the

llaboratory.It is the most importantassignment of every engineer inthis laboratory to give assistance,in the form of consultation,experimentation or suggestions,when asked to by anotherengineer, and the second mostimportant assignment is that ofcarrying forward the project towhich he is assigned.Looking back today, it is hard to

believe that within three years one ofthe computer industry's mostimportant technologies would growfrom the efforts of this tiny fledglingoperation. Who could have thoughtso then?

A meeting of the early lab privy council.Left to right are: R. Manning Herms,William A. Goddard, Reynold B.Johnson, Louis D. Stevens, Arthur J.Critchlow, and the late John W.Haanstra.

Identifying the Goal

The project that was to leadto development of theRAMAC disk file was called

“Source Recording,” which wasdefined as encompassing “allprocesses which take alphanumericdata from any source and transcribeit in a way so that the resultingdocument may be handled bymachine methods.” The punchedcard was the most widely used suchdocument at the time, and earlywork on the project focused oneliminating the use of cardsaltogether. As the problem becamebetter understood, however, thefocus turned to minimizing oreliminating the key punching taskitself. With many shifting goals tocome, the journey of discovery hadbegun.

It is helpful to remember that in1952 there were only three ways ofstoring information for use by dataprocessing equipment: the punchedcard, magnetic tape and, to a lesserextent, magnetic drums.

Each of these methods hadsignificant limitations. Punched cardsand magnetic tape essentially limitedthe user to batch or serial

In computing's earliest days, punchedcards were the basic means of storinginformation and were often organized in“tub files”. This file was typical of the1940’s and early 1950’s. The RAMACfile development grew out of an effortaimed at automating the tub file.

processing, which meant thatexpensive central processing systemswere often idle while information wasbeing accumulated and sorted forserial processing.

Information stored on drums wasalso randomly accessible andmoderately fast, but the low volume-to-density ratio of the drumtechnology also made it costly.

From today's vantage point, onecan see that what was needed wassome form of storage device thatwas randomly accessible with agreater surface-to-area volume ratiomuch higher than drums.

In September 1952, however, thefocus was on automating the so-called punched card “tub file.” Tubfiles had come into widespread useat that time as a method for makingmaster information punched oncards more readily accessible formachine processing. Essentially, thetub files were very large rectangulartrays containing master cardsarrayed in sequence by customernumber, item number, size, color,etc. Usually, the files containedseveral copies of each master card.

In a typical operation, clerksreceiving an order would search thetubs, pick out cards containing theneeded customer and item orderinformation, and send the cards tothe machine room, where theneeded documents--shipping roominstructions, packing slips, invoices,shipping labels and bills of lading--were produced.

An early crusader for tub fileautomation and a frequent visitor tothe San Jose lab was the late EdPerkins, a special marketingrepresentative in IBM’s SanFrancisco office.

He addressed the San Joseengineers regularly on theinadequacies of the tub file methodand took many of them to visit BayArea IBM customers so that theycould see firsthand what needed tobe corrected.

Throughout this period andcontinuing through the entireRAMAC development project, IBMengineers used the tub file billingand inventory control operations of alarge paper company as a “real life”touchstone in defining and refiningrequirements and specifications forthe Source Recording study and forthe disk memory system theyeventually developed.

By November 1952, the tub fileautomation alternatives had beennarrowed to two approaches. Onewas an endless belt with mastercards attached so that operatorscould see the ones they wanted andhave them electrostatically copied.These master cards would then beelectrically sensed into a newpunched card. The other--Johnson'sfavored approach at that time--involved a matrix of parallel verticalwires of one-foot length that wouldsense a card’s information andrecord it magnetically.

It was at about this time thatJacob Rabinow, of the NationalBureau of Standards, described aNotched Disk Memory Array inwhich each disk was rotatedindependently. His paper was widelycirculated at the San Jose lab andled a shift in interest back to disks.

Disks had already been consideredearlier and discarded becausemaintaining the necessarily minutespacing between a recording headand a disk surface (about 1/1,000thof an inch) was considered aninsurmountable problem. Otheralternatives were magnetic cards,plates, strips, bands, wires and rods.All were finally discarded.

A “File-to-Card” machine first madeoperable on February 10, 1954,consisted of an 026 key punch modifiedfor input-output service, the prototypedisk file, and the control electronicsshown against the wall.

Betting on disks

In January 1953, afterreevaluating all the alterna-tives, Johnson came to a

decision. The laboratory would focusits attention on disks. A work orderfor that month reads:

“Magnetic Disks have beenselected as the best medium for theRandom Access Memory to be usedfor File Maintenance. Disks willrevolve continuously at 16 r.p.s. sothat any of 200 columns of a recordmay be read out in any sequence,from any of 20,000 records.”

Those who were there at the timerecall that Johnson’s disk decisionwas a very unpopular one. Oneengineer advised Johnson that hewas backing a mechanical folly. Thepopular name for the disk array was“the baloney slicer.”

Despite all disparagement,including cartoons on lab bulletinboards, San Jose engineers wentback to work with their focusnarrowed to disks. On February 2,

1953, the objectives for a diskstorage configuration appeared. Theconfiguration was remarkably closeto the final product. While thedecision to focus on disks clearlyreplaced the goal of tub fileautomation, the general idea at firstwas that disk memory would be partof a File-to-Card Machine andconceptually the end product wouldbe, in effect, an electronic tub file, ifnot an “automated” one.

Above

Two of the early read-write airheads(circa 1953) positioned as they wouldbe against a magnetic recording surface.Exhaust ports to remove air frombetween the surfaces are visible. Thesewere necessary to achieve the requiredspacing between head and disk, thenabout one-thousandth of an inch.

Top

Rey Johnson’s early 1953 decision toproceed with development of a disk filewas the subject of much skepticism anddisbelief. One of a number of cartoonsfound on the lab bulletin board at thetime refers to a popular song of thatyear: “How Much Is Dat Doggy In DaWindow?”

This all changed shortly to a moreambitious concept when the U.S. AirForce asked IBM to submit aproposal to provide an inventorycontrol system for its base supply.The request called for a very-largecapacity, randomly-accessiblememory rather similar to what theSan Jose researchers were trying toachieve, but with the additionalrequirement for informationprocessing capability.

The late John Haanstra and otherlaboratory systems people wereassigned to respond to the AirForce’s request and on April 1,1953, submitted a proposal to theU.S. Air Force that called for amemory complex of 10 magneticdrums that might later be replacedby disks.

On April 14, 1953, shortly aftersubmitting the Air Force proposal,Johnson made the decision that thelaboratory would abandon alltechnologies competitive to disk. Heassigned three teams to work onvarious aspects of the project.

William A. Goddard was chargedwith developing a file model, assistedby Donald D. Johnson, John J.Lynott, Geoffrey Hotham and WarrenGonder.

The late Edward Quade and hisgroup were assigned to design amagnetic read-write head whichcould be no more than 1/10th of aninch high, and Haanstra and Alton E.Ewing were to concentrate on theFile-to-Card Machine while alsostudying extended systemsapplications.

It was agreed that one of the firstproblems to be surmounted wasfinding a way to maintain constantspacing between magnetic headelements and disks withoutsubstantial runout.

Goddard recommended an air-bearing approach and asked DonJohnson to try it out. Not only didthis approach work, but by June 2,1953, a third model of an air headwith a magnetic read-write elementdesigned for a magnetic drum wasused to write 51 bits per inch on analuminum disk that had been spray-coated with magnetic iron oxidepaint.

Thus convinced that an air bearingwould “float” a read-write head abovethe surface of a disk without crashingonto its surface, the developmentteam turned to three other pressingproblems: creating disk surfaces flatenough, developing an effective diskcoating and devising disk-to-disk andtrack-to-track accessing mechanisms.

Finding the best method andmaterial for creating flat disksurfaces was a matter of old-fashioned trial and error. Among thematerials tried were aluminum, brass,glass, plastics and magnesium. Inthe sheet aluminum trial, engineerswere horrified to find their so-calledflat disks had runouts as high as1/50th of an inch and greater whenthe disks revolved at 1,200 r.p.m.

The first successful material triedwas fabricated from magnesiumlithographer’s metal, but this wasfinally replaced by aluminumlaminates clamped under pressureand heated in ovens above theannealing temperature.

Jake Hagopian, assigned to thedisk coating problem, developedspecifications for a coating madefrom a paint base. Spray coating wastried as was dipping, but neitherproduced the uniformity of thicknessrequired in the magnetic coating.Hagopian finally settled on a spinmethod in which the coating waspoured onto the inner surface of arapidly rotating disk and then spreadevenly over the surface of the disk.

Bill Crooks produced the first trulysuccessful disks by filtering thedispersion through nylon hosiery andusing paper cups to measure outcorrect amounts for the spin coating.This primitive process was used for ayear before it was automated.

Later in the development project,Marcel Vogel, Don Johnson and

Ralph Flores invented the finalcoating formulation and basically, itis still in use today.

The third major hurdle--development of disk-to-disk andtrack-to-track access mechanisms-actually posed two challenges.

The first was how to apply the loadneeded to make the air headfunction. Initially, the researcherstried to devise a method usingsprings to apply the load, but thiscaused problems in moving the armfrom disk-to-disk.

It was Norman Vogel who finallysolved this problem with a designthat retracted the air head into thearm during disk-to-disk motion of thecarriage. The design provided a self-loading force using three miniatureair pistons on the back of the airhead that were activated in the track-to-track mode.

The second challenge was how toprovide the disk-to-disk and track-to-track accessing motions. Three

Above

Norman Vogel, one of the key membersof the 350 development team, at hisworkbench testing an airhead design.

Top

First test model, used with the “File-to-Card” machine, showing heads on andinserted between disks.

teams were set up to explorealternatives: Crooks and Haanstrawere to work on an electrical-servodrive system; Don Johnson and R.Manning Hermes were to pursue anapproach based on the IBM 402 typebar mechanism, and Gonder wasassigned a system using cams andspring clutches.

The servo approach was the earlyfavorite for the track-to-track drive butwas somewhat suspect for thetougher job of moving the heavycarriage from disk-to-disk. OnOctober 28, Davis was able to locatea previously recorded track with histrack-to-track servo, and in NovemberLynott found a way to use the sameelectrical servo system in amultiplexing mode to drive thecarriage from disk-to-disk.

On November 1, 1953, ReyJohnson turned entire responsibilityfor the development of the magneticdisk assembly over to Lou Stevens,who was promoted to seniorengineer.

Things seemed to be going onnicely until one Saturday evening,during a test, the new servo went outof control and rocketed free of therail, landing in a heap on the floor.

Wesley Dickinson, associate engineer in1955 when the photo was taken, pointsto the reading and writing arm of theredesigned 305 (the disk file’s unofficialdesignation while it was underdevelopment; the 350 designation camelater).

Within 12 days, Lynott and Davis,with assistance from Trigg Noyes,had completely redesigned a newservo system.

Thus, on February 10, 1954, theSan Jose team was able to achievethe first successful transfer ofinformation from cards to disks andback.

This was a huge boost to moralefor a brief period, but soon realitybegan to set in.

Stevens remembers that the firstlaboratory model looked like a RubeGoldberg arrangement and that “notmany people believed we wouldmake such a thing practical...”

“...but it was like a religion to us.We were going to make that thingwork for sure...because if wefailed...the whole San Joseexperiment would fail. None of uswere going to allow that to happen.”

It soon became obvious, however,that neither the first model of thedisk file nor the File-to-card Machinewas performing well enough to bedemonstrated. So on March 19,1954, an in-depth reevaluation ofevery basic design was undertakenand specifications were drawn up fora revised Model II.

Trigg Noyes had been thinking ofa new design for some time, and hisfïrst decision was to make the shaftholding the disks upright rather thanhorizontal, to provide more workabledimensions for the machine, makedisk replacement easier and providemore spaces for a number ofindependent access mechanisms.

Later that year, with constructionof the revised Model II well underway,the development team got twotremendous boosts.

The first came in the form of aletter from F. J. Wesley, who hadbeen assigned to review the progressof the work at San Jose and reportback on it.

It read, in part: “...We mustimmediately...attack accountingproblems under the philosophy ofhandling each business transactionas it occurs, rather than under thepresent condition of batchingtechniques...”

“We must build storage andperipheral equipment which canspread out into individual accountsevery business fact (random accessstorage) and allow operation of anew concept for handling businessinformation concurrently with itsinception...”

“I wish to recommend for yourconsideration that we double ortreble our efforts in thisdevelopment...I am firmly convincedthat (otherwise) we cannot expect toaccomplish the real purpose and thereal use for electronics in thebusiness world.”

The second boost came inNovember 1954, in the form ofofficial Corporate sanction forRAMAC product development work.The San Jose laboratory wascharged with designing, developingand building several field-test modelsof a machine utilizing a disk randomaccess memory attached to a serial“stick” printer to provide an initialsystem utilizing random accessstorage. The machine specified wasgiven the name DRAM (for DirectReference Accounting Machine), andlater the field-test models wereknown as the 305A.

On January 10, 1955, the first offïve Model II 350 RAMAC files wassuccessfully demonstrated--slightlyless than three years from the SanJose Laboratory’s inception. A newera in information processing hadbegun.

One of the press release photos for the305 RAMAC System when it was firstannounced in 1956. Key componentwas the 350 disk file. The system alsoincluded a central processor, cardreader and printer.

Disk technology today

The rapid and widespreadacceptance of disktechnology is a direct result

of the vast improvements in accessspeed and capacity, as well asdramatic reduction in costs over theyears since the 350 was firstunveiled.

A comparison between the 350 fileand IBM’s largest current file, the3380, tells the story:

The 350’s 24-inch disks revolvedat a speed of 1,200 revolutions perminute; the resulting data rate was100,000 bits per second. Today’s3380 disks revolve at 3,600 r.p.m.and the data rate is 24 million bitsper second.

The 350’s fifty 24-inch diskscontained a total capacity of 5million binary decimal encodedcharacters (7 bits per character) ofstorage. Today’s 3380, using nine14-inch disks, offers 1.25 billion

bytes or characters (8 bits per byte)of storage. The complete 3380 hastwo spindles with 2.5 billioncharacters.

Today’s microminiature read-writeheads on the 3380 use thin-filmtechnology. They permit vastlyincreased density of storage,permitting much faster access time.For example, the original 350 diskfile stored 2,000 bits of informationper square inch and had an averageaccess time of 600 milliseconds. TheIBM 3380 packs 12 million bits ofinformation per square inch and hasan access time of 16 milliseconds.

Looked at another way, the 350technology stored one megabyte ofinformation on an area the size of apool table. By comparison, the 3380technology stores one megabyte ofinformation on an area almost thesize of a postage stamp.

Meanwhile, the price of disk datastorage has tumbled dramaticallyover the years. Users of the 350RAMAC file paid $130 a month torent a megabyte of storage; today,

the cost of leasing the same amountof storage is about a dollar. Thisdrop in cost was in real dollars,unadjusted for inflation over themore than a quarter of a century inwhich it took place.

All of these improvements andcost reductions have triggered ahuge demand for disk storagedevices.

During 1983, for example, it isestimated that the dollar volume ofrigid disk drives rose 26 percent,while the dollar volume of flexibledisk drives increased some 110percent during the same period.

Looking back today, it’s difficult tobelieve that a relatively unnoticedannouncement on May 6, 1955,about a machine called the 350RAMAC could have caused suchdramatic changes in the pace oftechnology and in our lives. But that,in fact, is exactly what happened.

$130600

Costs and access timesdown sharply.

$22 .49

75

$7.98 30$2.86 25

$0.94 16

1956RAMAC

350

1965 19702314 3330-1

19753350

19803380

5 2,00029.2 220,000

776,000100

317 3,071,000

Capacity and areal density up.

1,260.5 12,000,000