IBM 2314

1966 IBM 2314
1st drive with ferrite core heads
Arguably the first disk storage subsystem with a cost/capacity to enable all on line data processing

Why it’s important A quadruple capacity replacement for the predecessor 2311, offering 29MB per spindle. The first use of ferrite heads increased recording frequency capability and therefore higher areal density. Success of the 2314 accelerated marketing of "plug compatible" versions by a number of IBM competitors (Plug Compatible Manufacturers, PCMs).

Discussion
The 2314 is a technological extension of the 1311/2311 series of drives featuring a larger data capacity removable disk pack with 20 recording surfaces instead of 10. RPM increased from 1500 to 2400. The hydraulic actuator remained the same but improvements in its control decreased the access from 85 to 60 milliseconds even while moving twice as many heads..

The 9 drive 2314 Model 1 Direct Access Storage Facility, announced April 22, 1965, was a marketing bundle of eight drives, a spare drive and a control unit that preserved a market for the lower priced and separately offered 2311. The actual 2314 subsystem was not integrated since IBM separately shipped the Control Unit, two 4 drive modules and a one drive module; these later became the 2314 Control Unit Model B1, the 2313 four drive module and the 2312 singe drive module. The bundle lasted until 1969 when IBM unbundled in part due to pressure from the Justice department and in part due to price pressure from the PCMs all of whom did not bundle. At that time IBM announced the 2318, a two disk drive module. IBM and some PCMs are unique in offering multi-drive modules (called pizza-ovens in the trade); the rest of the world generally ships a single drives . In the case of the 2314 all PCMs shipped single drive modules.

In
a competitive response to the PCMs., IBM later announced a lower priced three drive module, the 2319, which had the effect of lowering its rental per drive price for its new customers while maintaining a higher rental price for drives installed at its existing customers.

Many believe that this was the first product with a capacity and price point to allow all data to remain on line, replacing the tape
systems in many applications. However, it should be noted that such replacement was discussed as early as 1964 in the context of the 1311 and 1301 disk files. [ref: Disk File Applications, American Data Processing, Inc., Detroit MI, (c) 1964]

Heads
The 2314 introduced both ceramic flying heads and ceramic ferrite cores, replacing stainless steel heads and permalloy metallic cores. The ceramic materials were mechanically more durable and gave better high frequency performance. Metallic heads had their metallic cores mechanically "staked" in place, since metals are ductile and could be "staked" or "swaged" in place. Ceramic components, being brittle, had to be adhesively attached. One "dark side" of the new epoxy bonded head technology included the ability of the core to move ... it doesn't take much movement to effect performance of pole tip position at a nominal flying height of 85 microinches. Alternative suppliers of similar heads soon discovered problems of pole tip "recession" and "protrusion" which drove at least one supplier, Data Industries, out of business due to pole tip protrusion, resulting in mechanical interference with the recording surface. IBM's next generation, the 3330 used glass bonding to eliminate pole tip movement.

Channel
The 2314 also inaugurated within IBM the use of self-clocking FM (2-frequency, or "double frequency") encoding, which was more reliable. In this concept a single frequency is always recorded as the reference signal for the decoder, and data bits are placed between the clock bits. Good news was greater reliability, bad news was 1/2 of the recorded data was for machine timing and not user data. Subsequent recording schemes, such as MFM and RLL, improved the ratio of user data to clocking data, but the 2314 introduced the concept within IBM

Smoke Testing
In the early 1960’s disk drive contamination testing at IBM was limited to particulates, the standard being “Arizona Road Dust”, a commercially available contaminant which was metered into a disk drive air flow to simulate airborne contamination. The time to failure was a measure of resistance to head-disk interference and effectiveness of air filters. Conventional filters consisted of an pleated paper automotive style air filter on the input side to the disk pack, plus a nylon mesh filter in the hub of the pack itself.

Engineering was called by field service to investigate an IBM 2311 installation near the San Jose airport having peculiar contamination found during an attempt to clean the recording heads. The heads were removed and brought to the Building-13 materials analysis lab, inspected by scientist Irmela Barlow. She ran infrared analysis, revealing a mixture of materials, none of which was readily identifiable. She happened to remember a technical report which had a similar spectrum … but of air pollution. The root cause was aerosol droplets small enough to pass through the automotive style air filter. At high air velocity in the vicinity of the spinning disk surface, these particles impacted and collected on leading and trailing edges of the recording head, appearing as “comet trails”.

Once the phenomenon was understood, we sought to duplicate it in the lab for a “real world” simulation. The source of aerosol smoke was tobacco, Pall Mall cigarettes to be exact. Tobacco smoke had an infrared absorption spectrum similar to what was seen on the heads. A transparent plastic chamber was constructed with heads and a spinning disk to evaluate various contamination mixtures. To our surprise, flying heads were very resistant to problems caused by either aerosols (the tobacco), or by dirt (Arizona Road Dust), but a combination of the two caused very rapid failure. Apparently the aerosol was fluid enough to cause minimal effects on the air bearing, and dust particles were swept around the head, but the combination caused a semisolid contaminant to form which caused rapid head disk interference. Initially IBM’s accounting suspected some financial misbehavior after seeing invoices for cartons of cigarettes. Lab Technician John Ramos was observed smoking a cigarette in one hand while pumping cigarette smoke into the chamber with the other hand. When asked if he had not learned something about evils of cigarettes doing this work, he replied that he would stick to the cigarette brands with filters.

The test was used to establish the need and evaluate a new form of HEPA disk drive air filter, with sufficient small pore size (approx 0.3 micron) to trap both aerosols and particulates, which was subsequently made part of the 2314 product. Mil Kahl of the Product Test Lab scaled up Engineering’s “smoke test” to fill a closet sized room with smoke from a gadget smoking about 20 cigarettes, a very smelly operation which made the drive under test reek like a cigarette butt.

A summary of 2314 features follow:

Highlights
  • First use of ferrite (non-metallic) read/write transducer in a disk drive
  • First use of non-metallic (alumina ceramic) flying head “slider”
  • First IBM use of FM encoding, a self clocking method for greater reliability (patent #3,356,934, 1964)
  • 29 MB per spindle, versus 2.0MB for 1311 or 7.25MB for 2311
  • Last IBM application:hydraulic actuator.
  • First use of HEPA "absolute" air filter to control airborne particulates

Lowlights
  • First and only bundle of 9 disk drives with a mandatory spare.
  • Increased sensitivity to airborne contamination due to lower flying height
  • Greater sensitivity to pole tip recession and protrusion
  • Sensitivity to cross-contamination from disk pack swapping,”Typhoid Mary” syndrome
  • 20 surfaces to vertically align, greater sensitivity to spindle tilt


IBM San Jose, A Quarter Century Of Innovation”, David W. Kean, 1977, CHM accession number: 102687875
IBM Archive 2314 website
[Carlson 2011] Moderator's recollection of 1960s discussions with principals of Data Industries.

Provenance note: This page was originally authored by Bill Carlson; his last approved revision was version 18

More pages