storage
by BARRY BARKER
W
hen Winchester disc drive technology was first introduced, eight years ago by IBM, it allowed greater packing densities over floppies, improved due mainly to the reliability, hermetically sealed bubble within which the disc and head were kept free of dust and contamination to allow lower head flight, reduced head size and access times. These, more or less are still the advantages offered today - the improvements have been ones of magnitude. An 8 in Winchester gives 34 Mbytes ofcapacity with a low 35 ms access time. Generally speaking, Winchester technology has also progressed down the computer range from mainframes through minis to micros. The appearance of the 16 bit micros means that Winchester technology is going to be all the more important. Networked 16 bit micros are expected to make serious dents in the minicomputer market, especially when users can deploy them in a distributed data processing setup rather than rely solely on one central processor. Winchester disc drives with Abstract: Compactness, achieved through the use of smaller DC motors, custom ISI and mi~ropro~~sor controlled circuitry, and higher precision position mechanisms are two major developments in Winchester dtic technology described in this paper. Standardization, using intelligent inte$ace architecture and future improvements in the technology are also d&cussed. Keywords: data processing equipment, data storage devices, magnetic discs. Barry Barker Peripherals.
~0124 no 7
is UK sales manager
September 1982
of CPIJ
Developments in Winchester disc technology their large storage capacities are, and will continue to be, the first choice mainstay of micros designed for that type of new market. The conventional dividers between low cost and high performance Winchesters have tended to merge with the increases in recently capacities in 8 in drives. However, it is in the area of small systems application that the main potential for Winchesters still lies. With compactness as important a consideration as peripheral cost in word processors, microcomputers, etc., the 5% in and 8 in Winchesters have been to a large extent rivals. System designers often have to choose between the extra capacity and faster access times of the 8 in and the small physical size and lower cost of the 5% in. The ‘rivalry’, however, is friendly since manufacturers can offer both sizes in various specifications. A lot has also depended on the confi~ration of the system involved. If 16 bit micros fulfil their early promise of penetrating the mini market, by being implemented as multiple machine networks, then we may well see a shift from the larger to the smaller Winchesters. There will be good reasons for equipping every workstation with its own, say, 10 Mbyte 5% in Winchester, perhaps with an archive store somewhere on the network using an 8 in or 14 in unit. That archive would be the part of the system which in the next few years may benefit from an optical storage device. The adoption of Winchesters also reflects an upward trend in demand from small business system users. The small businesses, e.g. branch offices of larger businesses and typical users of
0011-684x/82/040015-03$03.00 0 1982 Butterworth
desktop micros, are becoming better educated in what they want of a computer. They also soon find that as more cost effective applications are found the need for more processing ability and more storage soon becomes apparent. Existing 8 bit micros can boast 10 Mbytes of Winchester based storage at the top of their ranges and many users who originally thought they could make do with floppy disc versions have eventually opted for the larger models. To squeeze this capacity into the small computers, manufacturers have both reduced the physical size of the component parts and applied significant technolo~cal innovations to pro duce drives about half the size of their predecessors. Much of this has appeared only in recent months and has been made possible by the adoption of smaller, more reliable DC drive motors, more compactness on the associated connections and mountings, custom LSI, and microprocessor controlled circuitry. The trend toward higher performance has also resulted in the develop ment of higher precision positioning mechanisms. This, in fact, is one of the biggest steps forward for the Winchester since its invention. Closed loop servo mechanisms improve positioning precision over open loop designs and permit increased track densities and lower access times. In the Fastrak, a Winchester produced by Shugart, positioning information is prerecorded on the lower surface of the drive’s bottom disc, ‘Cinder the control of a microprocessor, the servo head mounted on the actuator assembly reads the servo information and positions the actuator
& Co (Publishers)
Ltd.
15
over the correct cylinder. High precision comes from the servo head’s position being connected to correctly compensate for thermal expansion etc. The result is elimination of the effects of mechanical and thermal tolerances associated with the typical open loop stepper motor. Position feedback is the key to this type of system. The closed loop servo continually provides positioning information which is used to locate the data heads. The stepper motor operates on a sequence of pulses; no feedback confirms that the data head is actually on the track to which the pulses have sent it. These improvements reduce track width by twothirds allowing 500 tracks/in on drives which results in the tripling of storage capacity - hence the 34 Mbyte 8 in drives. Because closed loop servo correlates overall seek distance with its actual position, the best accelerationideceleration rates can be generated for each length of seek. The result is a low access time, typically 35 ms. In an open loop system, with no position feedback, more cautious movement is necessary and longer access times are the penalty. Other design improvements which are appearing on the newer models come from the DC motor itself For example, DC motors can give speed control with variation of less than 1%. High reliability comes from fewer mechanical parts and the elimination of uneven bearings load from the belt normally associated with AC motors. Drive electronics are also being converted to full custom LSI, significantly reducing components and the number of circuit boards and interconnections. A further embellishment is the dedicated head landing/shipping zone designed to prevent disc damage and data loss. A drive’s heads are moved automatically to this position when the drive is powered down, when DC power is lost, or when the disc is moving at less than a safe increment away from normal operating speed.
16
A _fZxeddisc actuator pebbly
Compatible upgrading System designers’ complaints about interfacing problems whenever they wish to change one disc drive for another are common. Within one manufacturer’s range it should not be difficult to upgrade from one Winchester to another of the same physical size. But the trouble starts when one wants to upgrade f-Yomone size of floppy to another size of Winchester; or to add Winchester storage to an existing floppy based system. The problem has been one of standards and the availability of an intelligent interface capable of acting universally between peripherals of different specifications. Some months ago Shugart, along with NCR and other firms, proposed that SASI (Shugart Associates System Interface) should be adopted as an industry standard since it was well established as a de facto standard and incorporated improvements over other interfaces on offer. The formal proposal was made jointly by Shugart and NCR to the American National Standards Institute and that body is now deliberating the matter. Meanwhile SASI has been adopted officially by several leading computer manufacturers both in the USA and in Britain, perhaps an indication of the
possible outcome of the deliberations. SASI’s principle is an intelligent interface architecture using logical block addressing. This allows a host CPU simply to specify the first block address and the number of data blocks to be transferred. The physical selection of the cylinder track and sector are all performed by the controller. No alterations to the processor, interface or controlling software are required when changing peripheral storage devices except to redefine the capacity of the drive. All housekeeping operations including track and head selection and bad track handling are transparent to the host processor. Advantages to the system designer include faster debug and reconfiguration, greater flexibility to attach new types of peripheral devices and lower assembly costs. Separate hardware interfaces and X/O drives are no longer required for each type of storage device. The ml1 SASI interface is implemented in four custom LSI circuits including a pair of devices in NMOS technology for setting up addresses, queuing requests and other infrequently occuring events, and two bipolar devices responsible for high speed clock-by-clock events such as data separation and phase locked loop timing generation.
data processing
The four fir11 custom ICs, some small TTL components aud a 230 processor chip take care of data flow, error detec~i~~corre~ion and have self diagnostic capability. The CPU is available for more productive jobs like network management and local interactive processing, The net result is a major change of emphasis from standardizing the peripheral storage unit to standardizing the system itself. It means that the system has been made ‘device independent’ so to speak. This will provide the microcomputer based small systems market with a peripheral memory interface similar to the IBM I/U channel standard for mainftames, As far as Wiachesters are concerned, an interface like SASI opens up a v&St new era of possibilities. Systems designers who had been putting ofI their upgrades will find their work too easy to be put off any longer - the extra storage and faster speeds which Winchesters offer are immediately available and easily integrated into the system.
Backup Ever since Wincl~~t~~ first became adopted in various manufacturers’ systems, around 1978, the backup question has been debated. Winklester discs cannot be removed from the computer and so their data has to be duplicated to guard against data loss arising from a system failure, operator error or some other mishap. In the past this problem has been overestimated. The essential solution can only come when one has cansidered the use to which the computer will be put and the trade off between cost of backup and backing up daily procedures and the cost-risk of data toss. In a system using even only a 10 Mbyte ~Vinchester, copying the entire store ant anto another medium could be a lengthy chore. If that other medium were the existing floppy discs used for normal in/out work, it could
voi 24 no 7 september
1982
take rz large chunk out of the day’s processing capacity - not to mention being too much of a chore to be bothered with on the very day something goes wrong. Is the user, however, going to need to copy out the entire data store every day, or half day? In most business a~~licatious it is simple enough to design applications s&ware and operator routines to allow only the updated sections of the file to be copied daily and, perhaps, the entire contents at the end of the month. If a crash occurs, the feeding back of data will take longer, since the last monthly backup discs have to go in first followed by discs of updates taken since. It requires careful disc management. But the chance of this happening has to be measured against the likely failure prospects of both the system and the Winchester drive. Winchesters are IO times less likely to give problems than floppy disc drives - a figure which probably strikes home better than any mean-time-betweenfailure rates. Given that kind of reliability and the comparative ease of taking just the day% data off onto a ffappy disc, it may be that no specific backup hardware is needed at all. However, not all applications of Winchesters lend themselves to this solution. Where data just has to be protected or where an entire file could be rewritten in a smaI1 number of hours, backup is a cost worth adding. Streaming tape drives have come to be accepted as the best answer to the problem. Recording 10 Mbytes of data at 90 i/s in serpentine mode on a 34 in cartridge takes about two minutes. The simplicity of this - carried out while the operator puts on his or her coat will generally outweigh the f300 or so cost of the streaming tape drive. The problem of streaming is the complexity of the interfacing and controlling of the drives. SASI controllers capable of handling the entire copy operation provide a simple and elegant solution to a problem previously
bedevill~d by a lack of LSIC controllers. At CR-7 we often find a need for streaming tape drives, Winchesters drive, controller aad host adaptor as a total package tailored to the system designer’s needs. The complete subsystem requires only a single slot in the system backplane, with the controller mounted on the Winchester drives.
Developments 3x1hand W’inchester technology is not standing stiI1; indeed it is still probably nearer the beginning than the end of its learning curve> for example, 500 tracks/in of the closed loop servo positioning systems will soon be surpassed with positioners capable of 1 000 tracks/in. Thin film heads and media are also now under development and their main advantage will be yet higher track densities and bit densities. The product is not yet available in commercial quantities and there are still several problems to be solved. As well as thin film heads and thinner media, verticd recording and the improved encoding methods used by the new generation of intelligent interfaces will improve capacities, speeds, readability and error detection. All of these can be expected to appear in the rigid disc drive ofthe immediate future. These developments are? if anything, accelerating in the rate of appearance. The system designer has to get used ta the feeling she/he has already experienced of hearing about a new development just as she/he has finished upgrading a system to include the last one. ‘This week’s new tcchnol~’ was once a sardonic description of the rate of progress in computer systems and peripherals but somehow it is starting to lose its humorous element. u “’ CPU Peripherals Ltd, Govett Avenue, Shepperton, Middx TWI7 ~~~~t~~-~~Tharn~s 46433,
SAC& UK.
Tel:
x7