Introduction by Jim Koch -
I have actually spent a lot of time over the last few weeks
with the person I have the honor to introduce next. That
is Al Hoagland. Today’s program would not have been
possible without Al's vision and perspective and ability to convene
old friends and individuals that can offer such a rich perspective
on this magnificent industry. Rey Johnson's name has been
mentioned. Al actually joined Rey Johnson's new San Jose
lab in 1956 to head research on magnetic disk recording.
Al received his doctorate in Electrical Engineering at Berkeley
where he taught in the faculty there and pursued research in digital
magnetic recording prior to joining IBM. In 1957 Al started
two research projects in the IBM San Jose laboratory. The
first research project focused on single disk drives while the
second focused on small scale magnetic strip files or more commonly
referred to as replaceable cartridge technology. He developed
for track following servo techniques high track density, back
then that were used later throughout the disk drive industry.
He also investigated longitudinal and perpendicular digital magnetic
recording techniques and their basic head designs. In 1959
became head of engineering science for advanced magnetic storage
technology. His work then expanded to include signal processing
for magnetic recording channels and air bearing design for controlling
the spacing between magnetic heads and disk surfaces. Al
has been instrumental, as others have mentioned, in the formation
of university centers in storage technology, including our center
here at Santa Clara University. He is a fellow IEEE, past
president of IEEE Computer Society and a trustee of the Charles
Babbage Foundation for the History of Computing. He has
written numerous articles on the scientific and technological
underpinnings of this industry, including a book in its second
edition, Digital Magnetic Recording.
Al it is a pleasure to welcome you here today
AL HOAGLAND
 |
I call the arrival of digital magnetic recording a paradigm shift because having heard so far today of the invention of magnetic recording and the early decades of activity you recognize that they were dedicated to the analog recording of sound and music. The history I am going to cover is that of computer data storage, to serve radically different applications and requiring entirely new technological implementations. (By the way, this change to digital storage is now also being exploited in sound, music and video recording). The companies and individuals leading this shift were driven by the needs of computing and information processing. Now the introduction of this new era is identified with the development during the 1940’s of the ENIAC. This computer was the first based on a processor using electronics, vacuum tubes at that time. The electronics provided unparalleled calculation speed then. The computer had 20,000 vacuum tubes and for input and output used paper tape and punched cards. That’s where it all began. But the arrival of the electronic processor generated a demand for much faster data input and output rates with the memory to take advantage of the processor computational speed. The memory basically provided the source of data that the processor could immediately access. The early computer block diagrams in those days did not include a block identified as storage that would fit between I/O and memory to buffer and better match the relative speeds of the processor and input/output. |
 |
Now the early developments were first driven by scientific computation, associated with the major increase in computer capabilities available and consequent new problems you could effectively undertake. Government funding was key in those early days. Commercial applications that then existed made use of punched card equipment and relay based computers and there was little anticipation of a major change. Universities played a leading role as a consequence at this stage. The most well known super computer that followed ENIAC was the Whirlwind at MIT. High speed memory. implementation went through the sequence of mercury delay lines, cathode ray tubes, magnetic core, and eventually ended up at semi-conductors. But, in addition to the "super" computer, there was great interest in computers that, while more modest in capability, could be low enough in cost to be widely available. UC Berkeley received funding from the Office of Naval Research to develop such an "intermediate" type computer. I was fortunate as a graduate student to work under Paul Morton as a member of this program from its formation. This computer (named the CALDIC), had to be low cost and a magnetic drum was chosen at that time as the optimum memory technology. |
 |
The fact that the magnetic drum
was a nonvolatile storage device was a plus. The required
features for memory, and this is where a real paradigm shift
in magnetic recording devices became evident; was that first
of all you had to be capable of all modifying a single "word"
or very short block of data, located in the midst of many other
such words. Also, these blocks necessarily were binary
encoded digital data. The memory unit needed to provide continuous availability as well as very high reliability to the stored information. Further, the drum had to have a very short direct access time to all the data stored and be able to either read, write or update in place the stored information. That required non-contact spacing between the head and recording medium to avoid wear (the relative surface speed between the two was in the range of 1600 inches a second.) The desire for a high RPM arose both from the short access time sought and the desire for a high data rate. A unique advantage of the drum over other memory approaches at the time was that the storage device was nonvolatile. |
 |
Now, as I mentioned, this program at Berkeley was run under professor Paul Morton from 1948 to 1952. We had heard of a some work at a company called ERA on a magnetic drum device, but our project was basically starting from scratch and, in particular, was one of the first major efforts to design a computer system based on a magnetic drum. I put the drum specs on the slide, and obviously you can see that the magnetic heads were not based on an air bearing for spacing. While the drum was stationary, you would move the head to just touch the drum surface and then back off slightly until you felt no contact would occur when the drum was turning. That led to approximately two thousandth's of an inch spacing between the head and medium. On the other hand, a drum could rotate relatively fast and the rpm was three times as high as later chosen for the RAMAC. Recording density was 800 bit per square inch. Design really focused on achieving the functional requirements for a memory. The capacity was the ten thousand words. Access time 8.3 milli-seconds. Every track had a magnetic head so you could get very quick access to any of the data. If you put two heads on the same track you could get much shorter access by just recirculating your data between them. Now, not only did this program lead in advancing magnetic drum technology and low end computer design but was a key source of trained students for the nascent computer industry as commercial efforts expanded. Student colleagues of mine who also went on to IBM and worked on the RAMAC included Lou Stevens and John Haanstra. This Berkeley computer project proceeded the formation of the San Jose Laboratory under Rey Johnson). This early work at Berkeley focused on a magnetic drum but the same digital magnetic recording techniques have much in common with other implementations of what I will call direct access data storage such as the magnetic disk. |
 |
This is a picture of the drum. I think it is quite impressive, given Paul Morton had to run this program with the usual turn over in students which occurs in an academic environment. The facilities that the university had were also limited. The operational drum had 200 heads, a head per track, all supported by four head bars. After what we saw in the earlier presentation, you must admit this looks sort of neat. I was worried at first about showing this picture because of the way people might react to such an ancient piece of hardware but after seeing even earlier recording equipment in the first talk I now feel very comfortable. |
 |
Now, how were things going to change? I will give you a little historical perspective here. Business data processing was a growing area. Magnetic tape made inroads into paper tape and punch cards. However, it also was a sequential medium and therefore the mode was still batch data processing. However, there was a growing interest in being able to do transaction processing. For example, in inventory control you could update all the records affected by a sale immediately (invoice, stock status, shipping order, etc) rather than sorting and running against a master tape. What was needed were the functional features associated with a magnetic drum but with a very high capacity at a reasonable cost. It is clear why such developments would be driven by computer systems companies, such as IBM, Univac, NCR, etc |
 |
Now how do you get high capacity with a low cost per megabyte? You could not put a head on every track, so you had to go to head positioning, and you needed a awful lot of recording area in a reasonable volume to get that capacity. This slide summarizes what is required. And we still need to update individual blocks of data. |
 |
Now head-medium registration tolerances in tracking and clock timing are critical since the recording medium does not have predefined bit cells and we need to rely on self clocking to read, write and update in place This picture illustrates the entirely new challenges imposed to provide the capabilities of direct access magnetic recording storage. |
 |
This chart is probably not readable
as projected but was done by Bill Turner of IBM about thirty
some years ago. I will use words to give you its message,
Rey Johnson came to San Jose in 1952. I was invited to
consult at his new Laboratory while a graduate student because
of my on going magnetic recording work at Berkeley. In
1956 I went to work full time for Rey. He was a tremendously
creative guy who loved to explore new ideas. What I am
trying to say is Rey was great at exploratory research and he
was a wonderful guy to work for, particularly if you had any
ideas you wanted to work on. But the down side was if you
wanted to get a product out. He protected his advance
technology projects and resources. One of the things that
struck me when I came into the Lab was that I saw this tremendous
challenge and opportunity of the RAMAC project and only a small
group of people assigned to make it happen. Then I walked
to the other rooms and saw a whole bunch of people doing a lot
of weird exotic things. And Rey was already starting to
put in place a storage device to be much more advanced than the
RAMAC. Rey got a lot of things started, and then other
people would take over to provide the follow through. This
turned out to be a very advantageous situation because Rey was
able to create a lot of new project activity that really set
the long range directions for the next generations of disk drives
In keeping with this spirit, the ADF program about which I will comment later was initiated and led by Rey well before the RAMAC was even announced. In fact this chart shows the number of design evolutions started before the first product had succeeded in the marketplace. Again, you didn't design disk drives as consumer products. They needed to be integrated into systems and IBM was the primary mover in the computer systems market place. There were other company efforts but IBM disk drive products were the mainstream standard and the ones that carried the industry for many years. For that reason I am focusing on their activities in the period this talk is covering. |
 |
This early RAMAC prototype model was intended to demonstrate that you could get a lot of magnetic surface area if you pack disks close together, showing the advantage of a disk stack implementation. |
 |
This shows the air pressurized head used on the RAMAC. This is the only device I brought with me to this Conference, actually the only item I kept over the years since I was intimately involved in this magnetic head design. In those days it was hard to find a component you could carry in your pocket and show. Actually it is close to the size of the new IBM microdrive. This head has a little nozzle on it and a plastic tube carrying pressurized air to keep the head off the disk. The force of the pressurized air counterbalanced a force loading the head towards the disk and the spacing was set at the point these two forces were in equilibrium The cost and complexity of an air compressor and head assembly to operate heads this way led to the use of a single head pair for the disk stack. |
 |
The RAMAC actuator arm assembly. The design obviously cost you a great deal in access time to any block of data since the head pair required positioning up and down the stack and then in and out to get on the desired track. |
 |
Here is a picture of the RAMAC system. When you look at a picture of the disk drive alone it certainly looks very large (which indeed it was). Within the context of the actual computer system it does not appear nearly so imposing. For a disk drive it was huge but not out of proportion with the other system components of that era. |
 |
The RAMAC disk drive had fifty disks twenty four inches in diameter. Access time was 800 milliseconds. Areal density of 2000 bits/inch. (100bpi/inch and 20 tpi/inch). The main initial application for the device was to replace punch card tub files that had to be manually accessed in trying to do transaction oriented data processing. |
 |
Before the RAMAC program was even announced,
Rey was heading up an advanced file program called the ADF.
This program was exceedingly important for a number of reasons.
The basic design, with a head per surface was really the optimum
way to design a disk drive in terms of access time and modularity.
The capacity objective for the ADF was ten times that of RAMAC.
The much larger capacity with an access time almost one tenth
that of the RAMAC (due to reducing head positioning to only one
dimension) opened up dramatically the applications that could
be considered. The first major commercial test was the
American Airlines reservation system which would for the first
time upgrade a data processing system to a geographically dispersed
real time transaction processing capability. The first model of the ADF was to be shipped with the Stretch computer, a major step forward in scientific computers committed to the government by IBM. And the ADF was a critical component to achieve the performance specifications. IBM bet its credibility on meeting the schedule for this super computer. Every lab in IBM had their particular component responsibilities for this system and San Jose, of course, had the disk drive. San Jose being the newest Lab and the ADF disk drive being a major challenge placed the Lab under an unusual degree of pressure. The greatest exposure was that this drive involved three entirely new technologies for a disk drive. First, the use of “flying heads” (heads that did not require an air supply) . Second, a hydraulic actuator was selected to deal with the heavy head-arm assembly required by a head/surface design. Third, the choice of vertical (or perpendicular) recording based on the desire to have a much harder magnetic surface than that of the coated iron oxide then being used on disks. There were concerns about head/disk contact and the damage it would do and the idea of using an oxidized steel disk appeared to be a solution. Since under the oxidized layer the steel was soft magnetically, vertical recording was then possible and vertical head structures were a obvious choice to consider as well as appearing potentially cheaper than the longitudinal type of head. The ADF was like starting anew, there being almost no relation to the RAMAC technologies. The only thing that was really common was the disk diameter. |
 |
Shows is the flying head, which generates a self acting bearing from the pressure generated from the boundary layer of air on the disk through the contour design and keeping the head off the surface. |
 |
This shows the vertical recording head that was in pilot production for the ADF. What you see is a simple vertical probe, a coil with a lot of turns, and a magnetic capsule, if you will, into which the coil and probe were inserted to provide shielding from adjacent tracks. |
|
The bottom line, it is a challenge to change one technology
but changing three is extremely difficult to say the least. The ADF
was way behind schedule and the testing was only uncovering more difficulties.
(Rey had already moved on to start some even more advanced projects).
So there was a major audit of the program and corporate review. It
became clear that the surface quality of the steel disks made them crash
prone. (I had the dubious privilege to see 40 heads crash at one time
on a test module). Al Shugart was selected take over management of
the program. I was on temporary assignment to oversee the recording
magnetics. Among major changes was the decision to abandon steel disks and
move ahead through advancing the recording technologies developed for the
RAMAC and already in production. Al Shugart, at this point,
really didn't have much prior experience in the technology, His leadership
became evident in two things he did in spite of the crisis nature of the
atmosphere He was able to keep corporate executives from headquarters
at arms length and trusted the engineers enough to let them do what they
felt necessary. In turn the “troops” gave him their full
support and he successfully turned around a program that was close to being
written off.. I view this period as the real start of the career of
Al in the disk drive industry. While the Stretch model had to be shipped
with pressurized air bearing heads due to the schedule, the 1301, which
was the commercial version of the ADF, shipped in 1961 with flying heads.
This accomplishment secured without question the mission in the disk drive
area for IBM San Jose. The head arm module of the ADF. This drive supported two such modules. A picture of the 1301. This drive is the true precursor of all the succeeding generations of disk drives, being the first to use flying heads and a head/arm assembly providing a head per surface . In this context the RAMAC was a one of a kind product . The 1301 deserves more recognition for its place in the history of disk drives. |
 |
Shows ADF chief engineer, Al Shugart, who obviously is adjusting the positioning of the module in the disk drive to personally insure that there will never be any failures. |
 |
First of all if you look at all successive generations of disk drives, the flying head per surface combined with a comb actuator type positioning system is pervasive. Also, by providing a new level of capacity/access time performance the drive paved the way for on-line real time transaction processing. The American Airlines Saber System was the pioneering application (providing a 3 second response time to a very large database) and set in motion a major new direction in information processing. |
 |
The single disk file (SDF), which I started under Rey Johnson even before the ADF crisis erupted, again reflected the nature of Rey. The objective was to develop a track following servoing system based on a pattern on the disk itself in order to achieve a major increase in track density capabilities well beyond the open loop systems then being used. A single replaceable disk drive was the implementation and provided a unit that at that time could match the capacity of a tape reel but with direct access. This approach to tracking was not actually incorporated in a product until the 3330 which was announced in 1971. Since then all drives use data on the disk for track positioning. |
 |
PICTURED is structure of the SDF. Reasons there are several heads per surface include: reduced head positioning travel; reduced access time; reduction in space taken by servo pattern and therefore better utilization of disk real estate; and the ability to read byte wide data as is done on tape. |
 |
The replaceable disk pack was an approach to bringing transaction processing capabilities to small businesses through lower cost systems. The drive used a two megabyte disk pack, (with 14 inch diameter disks). The concept was that with a number of packs per drive various such applications could be run when desired. However, on-line storage was really expected to be on-line so the customers would buy additional drives as they could afford rather than procure many additional packs. This low cost drive (for that time) greatly expanded the market for disk-oriented processing. |
 |
Pictured is the IBM 3330 which also shows a disk pack as well as the multiple drive configurations that were popular. The 3330 is the first disk drive to implement head positioning by servoing on a disk pattern. (Given the customer preferences and the advantages in favor of achieving higher and higher densities on a fixed disk rather than a replaceable pack with the announcement of the 3350 in 1976 IBM returned to the fixed disk drives only). |
 |
This graph illustrates the scaling laws that allow simple extrapolations of performance based on key geometrical parameters. It clearly demonstrates the density gains from scaling down dimensions. From the introduction of the disk drive in 1956 to the 1990's storage density increased at a 32% CAGR while starting in the nineties it has been advancing at better than a 60% CAGR due to the introduction of MR read head technology. John Best will cover this latter decade and beyond in his talk coming up. |
 |
There was a time when the leading technology and products in the San Jose area were those of the magnetic disk drive. The disk drive was invented here and is a home grown industry while semi-conductors activity was imported. When this area first began to be called Silicon Valley there was a strong feeling by many that it really should be called Iron Oxide Valley. In that period the disk drive industry was very bullish and felt eventually the earth would be “covered” with iron oxide. This slide reflects that period but we all know that times have changed and reality has set in. |
 |
I am showing this picture primarily because Al Shugart is the luncheon speaker. It shows an alternative “direct access” mass storage device based on using cartridges loaded with strips of magnetic “tape”, another attempt to get a lot of capacity with a relatively short access time. All these alternative designs to a disk drive never really succeeded in the market place over any long period of time. Here, the tradeoff is capacity at the expense of access time. |
 |
The memory/storage hierarchy is shown You note it can be viewed in terms of high speed/high cost on line memory at the top low cost/high capacity off line flexible media at the bottom and direct access storage in the middle. The disk drive provides the optimum compromise between high capacity and short access time and has been the only product that really has succeeded in surviving in the so called “access gap” between memory and removable storage. I am sure all these three storage implementations will survive. |
 |
This picture, symbolizing the three dominant memory storage technologies, goes back many years and the message it conveys is still true today, even though it was made more than two decades ago. |
 |
In summary, Rey Johnson created an environment
and atmosphere that really accelerated the research and development of storage
technology and in particular the magnetic disk drive. Al Shugart contributed
immensely in product management of these innovative devices and creating
a business out of the opportunities they posed. The disk drive industry
started in San Jose and San Jose is still its center. I feel very
comfortable believing that will remain so. The disk drive is the major
factor I identify as responsible for the incredible continuing growth in
computer applications. I am pleased to wrap up by saying when I started
on the CALDIC, the state of the art of areal density was 800 bits per square
inch. Now recent disk drives operate in the 100 gigabits per square
inch range.. That means in my personal time frame I have witnessed
improvement in areal density by a factor of 100 million. I can think
of no other technology where such dramatic progress could occur over the
span of your career. Thank you. |
