• Keine Ergebnisse gefunden

All about Plug-Compatible Disk Drives

Im Dokument Contents of Volume 2 datopro (Seite 79-101)

Independent peripheral manufacturers introduce disk drives compatible with IBM System/360 mainframes at lower cost and with higher performance than IBM's own.

1MB introduces improved models. Independents introduce new models with similar capabilities. IBM lowers its prices through extended leasing plans. Independents cut their prices. IBM repackages its disks to give lower prices.

Independents shave their prices again. IBM acts. Inde-pendents react. IBM reacts in turn. And so the story goes on and on. Pity the data processing manager who must keep up with all these events if he is to remain aware of the equipment that will most economically serve his constantly changing needs.

Have we overstated the situation? Consider these events that have occurred during the last few years and have significantly affected the selection of disk equipment:

• IBM 2314-disk-based processing "arrives."

• Independents provide many alternatives to the 2314.

• IBM 3330-lots more bang for the buck.

• Double density-independents offer an alternative that requires more than just price-tag comparisons.

• IBM 2319-a thinly disguised price cut for the 2314.

• Independents announce "3330" drives.

• Integrated attachments-IBM moves part of the disk controller inside the mainframe.

• Growing evidence that independents' drives work well.

• IBM adds virtual storage to the System/370-and more disk storage is called for.

• Expanded attachments-the controller portion of many IBM disk subsystem configurations is reduced in cost.

• IBM 3340-new concept in disk pack drives.

• Technological announcements by AT&T, IBM, and others hint that disks may be obsolete in 3, 5, or 10 years.

Add to these items the constant price adjustments by the independents, elimination of overtime charges and long-term leases by IBM (read price cuts), complex and changing configuration possibilities, etc., and you begin to draw a true picture of the position an EDP manager has been forced into.

Compatible replacements for IBM's 2311, 2314, and 3330 families of disk subsystems abound. They promise worthwhile savings in cost and/or improvements in performance. This report analyzes the pros and cons of acquiring disk equipment from a source other than IBM, presents the specifications of 31 subsystems from 11 vendors, and summarizes the results of an extensive user survey.

The development and growth of the plug-compatible peripheral market has removed one of the security blan-kets formerly available to those responsible for selecting a company's EDP gear. To many, selection of an IBM computer is comforting because it works and because so many others have done it that it must be right. However, even with this point of view, a user today cannot ignore the presence of plug-compatible peripherals. He must again make a decision to stay with IBM or try someone else's devices. Just as there are viable alternatives to IBM computer systems, there are viable alternatives to IBM peripherals. They are widely available, their prices are generally lower - and they work.

This report sorts out the alternatives that independent manufacturers offer for IBM disk products, summarizes users' experiences with more than 1700 non-IBM drives, and provides some sound buying advice. t>

The l"rfemorex 3670, a direct replacement for IBM's high·

performance 3330 subsystem, has been well received by users.

MAY 1973

©1973 DATAPRO RESEARCH CORPORATION, DELRAN, N,J, 08075

700-010-4Ob Peripherals

All About Plug-Compatible Disk Drives 1:> The Game Begins

The concept of independent peripherals-those units de-signed to replace or complement the peripheral devices offered by the mainframe manufacturers-is far from new.

Companies such as Ampex, CalComp, Memorex, Potter, and Telex, to mention a few, have supplied computer peripheral equipment for years. Until 1968, however, most of their products were sold to mainframe manu-facturers rather than directly to end users of computers.

Typically, only large users ever acquired peripherals other than terminals from a source other than the mainframe manufacturer-and even then, these "foreign" peripherals were generally confined to specialized systems. A few hardy individuals with specialized needs played around with special peripherals such as digital plotters, and the scientific community experimented with unusual types of input devices. But by and large, to the great majority of users, a computer system was an entity put together exclusively from the building blocks furnished by the mainframe manufacturer. Use of peripherals from sources other than the mainframe manufacturer normally involved special "black box" interfaces and costly adaptations of the standard software.

Though IBM's share of the computer market has changed little since 1960, a variety of systems, each with its own special input/output logic and structure, characterized IBM's product line prior to 1964, the year of the 360.

Since then, with about three-quarters of the computer users drawing from a common pool of peripheral devices and with IBM's reliance on its own resources, at least for all conventional peripherals, the attractiveness of selling only to the mainframe manufacturers has greatly de-creased for the independent peripheral makers. The OEM (original equipment manufacturer) market, though iarge, was only a drop in the bucket compared to the potential of serving the whole data processing community. In addition, IBM has no monopoly on technical talent. Many peripheral companies have proven that they can design and build first-rate equipment-and that they can deliver and maintain the equipment at lower prices than IBM's.

There is every reason to believe that they can continue to do so in the face of stiffer competition from IBM.

The attraction to a company to produce products for a known market is strong. And this points up one edge that the peripheral companies had on IBM: they were not developing a new market. They did not have to convince anybody that magnetic tape units or disk drives are a good idea. This work had already been done for them by IBM and the other mainframe manufacturers. So with this proven market waiting, and with the example of what aggressive marketing had done for the key-to-tape and minicomputer concepts, the door was open to big money.

The subject of peripheral devices is a broad one that encompasses many concepts, the old and proven as well as the new and exciting. The independent peripheral manufacturers-those who do not also produce main-frame-have concentrated on two of the relatively old and proven devices: magnetic tape drives and disk pack drives.

The reason is simple. The other two staple types of on-site peripheral devices are punched card units and line printers-and in these categories IBM's products are ex-tremely well engineered and hard to surpass on a price/

performance basis. By contrast, when the independents began their assault on magnetic tape and disk drives in System/360 computer installations, IBM was vulnerable on both a technical and a price basis. (Developments since then, however, have greatly reduced IBM's vulnerability on both these grounds.)

The decision of several peripheral manufacturers to con-centrate on the IBM replacement market bore fruit in 1970, the first big year for the independents.

In late 1970, when the independents had usurped an estimated 8 to 12 percent of its tape and disk drive market, IBM began reacting. Following the announcement of new IBM peripherals that effectively beat the prices of the independent manufacturers' products, the inde-pendents quickly reacted in turn. Some lowered prices on existing units. Some announced new product lines. And some bemoaned at great length and in public the audacity of IBM to prick their great balloon. As outlined previous-ly, this was just the beginning of many rounds of product and pricing announcements by both IBM and the inde-pendents throughout 1971 and 1972. And the game is still going on.

The net result for the user is more processing power per dollar. However, the fattest days of the independent peripherai market have passed, just as they did for OCR, CRT display terminals, key-to-tape data recorders, and time-sharing, to mention some of the computer product concepts that have attracted large amounts of venture capital and trade press headlines during the past few years.

From now on it's going to be a tough, competitive market-and this is to the advantage of the computer user, if he stays aware of who, what, and how much.

Just Step Up and Plug It In

Plug-compatible disk drives are widely available, have been installed and put into successful operation, and are saving their users money.

A plug-compatible peripheral device is one that can replace an existing unit made by the mainframe manu-facturer (or, for that matter, by another independent maker) by the simple process of unplugging the old unit, wheeling it out, wheeling in the new unit, plugging it in, 1:>

MAY 1973

©1973 DATAPRO RESEARCH CORPORATION, DELRAN, N.J. 08075 REPRODUCTION PROHIBITED

700-010-4Oc Peripherals

All About Plug-Compatible Disk Drives

This Memorex box represents the only attempt by the indepen-dents to come directly to grips with the IBM internal storage control features in the System/370 computer family. It can be used to connect Memorex 3330-compatible drives via the IBM Model 125 DDA, Model 135 IFA, and Model 145, 158, and 168 ISC's. Single-density and/or double-der"sity drives can be attached, although IBM software does not support double-density drives on the Model 125.

> Just Step Up and Plug It In

Plug-compatible disk drives are widely available, have been installed and put into successful operation, and are saving their users money.

A plug-compatible peripheral device is one that can replace an existing unit made by the mainframe manufac-turer (or, for that matter, by another independent maker) by the simple process of unplugging the old unit, wheeling it out, wheeling in the new unit, plugging it in, testing it, and resuming operations. No change to existing software, either systems or application programs, is required. (In actual practice, some "plug-compatibles" are not this easy to install.)

Actually, there are two types of plug compatibility. One we will refer to as subsystem compatibility, and the other as component compatibility. A subsystem is compatible if it will interface directly with the computer's I/O channel and function in the same manner as the subsystem it replaces. A component is compatible if it will interface directly with the appropriate controller. Other manufac-turers cannot obtain detailed IBM device-to-controller interface specifications (ethically , anyway) until three

months following the frrst delivery. Consequently, a manu-facturer cannot be absolutely sure of achieving direct com-ponent compatibility until then.

Although the interface specifications for the input/output channels of IBM's System/360 and System/370 have long been known, subsystem compatibility is not necessarily an automatic procedure. The exact command structure must be known in order for the plug-compatible devices to function within the framework of IBM-supplied software without changes.

Therefore, plug-compatible devices will normally lag behind IBM's products. Thus far, the independent disk drive makers have elected to imitate the hardware/

software interfaces of IBM's products, with two note-worthy exceptions. One of these occurred between the announcement and delivery of IBM's 3330 drives (about 14 months). The independents tried to steal a march with double capacity 2314-style drives that were not directly software-compatible; this approach met with some success and a couple of notable failures. The other software incompatibility is current. Two vendors, Itel and Memorex, are marketing 3330-style drives for System/360 Model 65's. This requires provision of the equivalent of a block multiplexer channel, which IBM does not make available for the 65. Itel uses a software approach, and Memorex uses a combination of software and control unit microcode; all require some modification of the standard operating system. In a recent development, Itel announced the availability of full-speed 3330-type drives for the IBM 360/50. Previously, the company offered a reduced-speed version. Itel states that engineering changes over the life of the 360/50 increased the I/O rate capability to the point of its being able to accept the full-speed 3330 data stream. According to reports, the company is now looking at the 360/40.

At present, the independents are taking dead aim on IBM's 3330 product line. Plug-compatible 2311 's and 2314's are still available, including double-density 2314's, but new production and development of these devices have essentially stopped.

There Goes the Old Ball Game

IBM revolutionized auxiliary storage concepts when it introduced the first disk pack drive, the 1311, in 1962.

IBM revolutionized the disk pack drive concept when it introduced the frrst high-capacity subsystem, the 2314, in the mid-60's. IBM revolutionized the cost/performance characteristics of disk pack drives when it introduced the 3330 in 1970, and again in July 1973 with the announcement of the double density 3330-11. In between, IBM announced the curious 3340 system. Other computer vendors during this period generally accepted the inevitable and offered 2314 and 3330-style drives in place of or in addition to their other products.

The most outstanding feature of the IBM 3340 is the inclusion of the moving arm, the read/write heads, and at

t:>

700-010-4Od Peripherals

All About Plug-Compatible Disk Drives P least some of the electronics directly in the 3348 Data

Module (disk pack). This feature has two notable effects.

One is the increased reliability of a sealed container, which eliminates any danger of damage due to airborne particles and enhances read/write reliability because the same head always serves the same track. The second is the increased cost of the pack. An IBM 3330 pack costs $775 (100 megabyte capacity) or $1,150 (200 megabyte capacity). The 3348 Data Module costs $1,600 (35 megabyte capacity) or $2,200 (70 megabyte capacity).

With the ftxed-head option, the 70-megabyte module costs $4,400.

True, the drive portion of the 3340 subsystem is reduced in complexity and cost. But if your application needs require many packs (a service bureau is an extreme example), the overall costs are going to be high. As of this writing, 3M and BASF have announced compatible Data Modules for use with the 3340, and Memorex is also expected to enter this market.

If you couple IBM's 3340 announcement with the expansion of integrated attachments for the 3330, the double-density 3330-11, and IBM's statement that it has become practical to hold an installation's entire data base on-line at the same time, you come up with a curious anomaly. Is it possible that we are working ourselves full circle, back to ftxed-disk units and away from the pack concept? This question is intensilled by the 1974 IBM announcement of an additional storage segment on the 3348 Data Module that is served by ftxed heads; i.e., no arm pOSitioning is required to retrieve data from the 0.5-million-byte portion consisting of 5 cylinders of 12 tracks each. The only delay is rotational. This feature, although expensive, improves access to frequently referenced data such as indexes.

By removing the head mechanisms from the drive and moving the controller inside the mainframe, IBM has certainly left little for the independents to play with in terms of profitable items, except for those who also build packs. Is the improved environment of the sealed 3348 Data Module required for reliable recording at the higher density? (The 3330-11 's operate at 3340 densities without sealed packs.)

The Game's Not Over Vet

How important is it to you to save money on 2314 and 3330 drives? When, if at all, do you plan to convert to 3340's, or to 370's for that matter? IBM's planned migration strategy to move its users from the System/360 to the 370 has resulted in something less than an overwhelming tide. The 3340's and the 3330's are not available for the System/360 (except for the large and rare Model 85's and 195's). And there are restrictions on the number of 3340's in System/370 models and on software support. (Details on conftguration and support are contained in the Computer section; see Report 70C491-03 for the System/360 and Report 70C-491-04 for the System/370.) Such restrictions can, of course, be modilled at any time, as they were for the 2319 modules.

Is the 3340 the modern equivalent of the 2314 and a lower member than the 3330 in a hierarchical group of devices? Or is it the precursor of a new family of devices that will obsolete the 3330 and all others? Will such a family include models with access times sufftciently fast to support large-scale virtual memory processing in a really effective manner? And how does IBM's newest mass storage device, the 3850, ftt in? The 3850 uses magnetic tape technology to provide low-cost bulk storage of data.

It is coupled with 3330 drives to form a two-tier hierarchy of storage. Data mes are called to the 3330's as needed and returned when processing is completed, when they are no longer needed, or when superseded by a higher-priority task. (You can now visualize a computer system with both virtual main memory and virtual disk memory, but visualizing a virtual magnetic tape facility takes some doing.)

One last question: are we raising more questions than we can answer? Only the answer to the last question is definite : Yes. Only time will bring full and accurate answers to the other questions raised.

For many computer users, buying from an equipment vendor other than IBM is a new experience that has some disquieting ramillcations. Datapro recognizes the need for a practical guide through the conflicting claims, technical jargon, and marketing ballyhoo to the goal of best use of your data processing dollars. Accompanying this report is a comprehensive set of comparison charts listing both IBM's products and the alternatives from the indepen-dents for IBM, Honeywell, Univac, and Xerox products.

The convenient side-by-side format allows quick, meaning-ful comparisons. Also in this report are the results of an extensive survey of our subscribers' experience; this will give you a quick insight into the luck that others have had with particular products. In addition, we will continue to publish individual reports on important devices that need closer scrutiny; at present tMs group consists mainly of IBM 3330 replacements. If 3340-type devices are announced by the independents, they will also get individual coverage.

The Almighty Dollar

The principal reason for considering the acquisition of peripheral devices from a source other than the manufac-turer of the mainframe is cost. Secondary reasons include improved performance and shorter delivery time.

There are two ways of looking at savings: the actual number of dollars saved and the percentage of the total system price saved. Both ways are relevant. The ultimate cost figure that an installation must be concerned with is, of course, the cost per item processed. The percentage figure becomes important when evaluating several alterna-tives for spending a fixed number of dollars.

For the majority of computer users, disk subsystems represent an appreciable part of their total equipment costs. Any significant reduction in the cost of these subsystems is clearly a worthwhile goal.

t:::::

©1975 DATAPRO RESEARCH CORPORATION, DELRAN, N.J. 08075 REPRODUCTION PROHIBITED

MAY 1975

70D-010-408 Peripherals

All About Plug-Compatible Disk Drives

USE_RS' RATINGS OF DISK SUBSYSTEMS

Number Number of Weighted Averages and Response Counts* Problem Counts**

Disk Subsystem of User Spindles Overall Ease of Hardware Maintenance

Installation Operation

Installation Operation

Im Dokument Contents of Volume 2 datopro (Seite 79-101)