The disk drive: winner and still storage champion
As the disk-drive industry moves to GMR and faster interfaces to meet user demand, the capacity ramp still continues. The vendors that make the smoothest transitions in this technology-centric business should emerge as winners.
By Maury Wright, Technical Editor -- EDN, January 21, 1999
I’ll bet you think of the disk-drive industry as a commodity business in which cost is king. Cost certainly matters, but you can hardly find a computer-system component that comprises more technology than a disk drive. The modern electromechanical hard drive relies on a microminiature air frame (the flying head); a 10,000-rpm motor stuffed inside the spindle; perhaps the most advanced electromagnetic element in all of electronics; and state-of-the-art DSP, interface, and mixed-signal ICs. Somehow, disk vendors have been marrying new revisions of these diverse technologies each year to deliver 60% more capacity every 12 months or so. Now, the industry has entered perhaps the most difficult transition ever with a move to giant magnetoresistive (GMR) heads to continue the ramp in capacity and data rate. To go along with the new heads, drives need faster motors, read channels, and interfaces. More than ever, you need to evaluate the technologies inside the black box to pick winners in the disk industry. Like a mysterious mix of Superman and Houdini, disk-drive designers appear capable of overcoming any obstacle to producing higher capacity and faster products. Despite past miraculous growth, just two or three years ago a 17-Gbyte desktop drive seemed like a mirage. Now, you can buy one for less than $400, and, perhaps even more impressive, you can get 6-Gbyte drives for less than $150. The rampant capacity increase isn’t limited to fixed drives either. Removable hard drives have hit 2 Gbytes, and drives with flexible media seeking to replace the floppy now store 250 Mbytes (see sidebar " Floppy-replacement battle continues "). Rotating-magnetic-storage devices are impressive examples of what our industry can accomplish.
Moreover, the disk drive is unique in longevity in the electronics industry. Few technologies other than disk drives and CRTs have lasted half a century in the face of constant evolution of replacement technologies. In the case of data storage, vendors have presented bubble to flash memory, optical storage, and holographic techniques as replacements for disk drives. Nevertheless, magnetic-drive vendors believe they have at least another decade of prosperity ahead, despite constant technical hurdles and potential challenges from alternative technologies (see sidebars " Optical technologies emerge—but for auxiliary or replacement roles? " and " COL and microactuation push bit-per-inch and track-per-inch specs "). System designers working on near-term designs must both examine disk technology and—because of rapid change—read between the lines to see what the vendors might have in store over the next six months. Moreover, you almost have to examine the industry based on technology characteristics, such as heads, drive capacity, performance, and interfaces relative to market segments. Today, you can break the fixed-disk-drive business into the desktop, mobile-market, and enterprise-market segments. The desktop segment largely comprises 3.5-in., 1-in.-high drives that use the AT attachment (ATA) interface. Quantum and Seagate also offer some low-end SCSI drives that target high-end desktops. Moreover, Quantum’s 5.25-in., ATA-based Bigfoot family also targets the desktop and presumably costs less because it uses more spacious media ( Reference 1 ). Quantum has been the desktop-market leader for several years, and IBM and Seagate are not far behind. Western Digital has lost market share of late, and Maxtor has been the biggest gainer. Fujitsu and Samsung also offer desktop drives. In the desktop segment, the vendors emphasize cost and capacity along with reliability, which is key across all segments. IBM leads the mobile-market segment for 2.5-in. drives (see picture). In this segment, "thin is in," along with low power and rugged design. For example, all mobile drives include head-loading technology to ensure that the heads don’t come into contact with the media—a luxury the desktop segment simply can’t afford. Toshiba runs a strong second in mobile drives and has led the way with 8.45-mm-high, slim-line drives that store more than 4 Gbytes targeting markets such as ultraportable and handheld computers. Hitachi and Fujitsu also have solid families of mobile products. Long-time leader Seagate remains atop the enterprise-market segment, with IBM a close second. Other participants include Quantum, Fujitsu, Hitachi, and Western Digital. This segment includes workstations and very-high-end PCs as well as servers, external drive arrays, and other enterprise-network-storage applications. Enterprise-segment drives rely on the SCSI or Fibre Channel Arbitrated Loop (FC-AL) interface. As the companies release products at new and higher capacities, those new 3.5-in. drives typically come in 1.6-in.-high packages. As the companies ship each succeeding generation of drive, previous-generation capacities move to 1-in.-high packages. The highest volume shipments typically come in 1-in.-high, second-generation drives. Typically, enterprise vendors now offer 36-Gbyte drives in 1.6-in. packages and 9- or 18-Gbyte drives in 1-in. packages. Heads transition to GMR The transition to GMR heads will likely be the No. 1 technology trend across all market segments for near-term drive evaluations. ( References 1 and 2 include information on the transition from thin-film inductive heads to MR heads.) MR and GMR heads work similarly. Both require a separate inductive write element to store data by magnetizing tiny regions on the disk. As the MR or GMR read element crosses flux transitions on the media, the magnetic orientation in the sense element changes, thereby causing a change in resistance. The difference between MR and GMR comes down to construction of the head and the physical properties that cause the change in resistance. (IBM invented both of these technologies. To get more details on the physics, go to www.ibm.com/harddrive.) The important point for disk-drive customers comes down to areal density. As bit cells get smaller, MR heads produce insufficient resistance change for the electronics to accurately read the data patterns. GMR heads exhibit resistance changes of two to almost four times the magnitude of those that MR heads exhibit. The transition to GMR began in the second quarter of last year in the mobile market in which capacity poses the biggest challenge. Now, you will find GMR heads appearing across all market segments, although some vendors may not make the transition until midyear. Vendors do have somewhat of a choice in the near term. For example, Maxtor recently announced the DiamondMax 4320 series of drives using MR heads from Read-Rite Corp ( www.readrite.com). The family tops out at 17.2 Gbytes on four platters, or 4.3 Gbytes per platter. Realistically, the 17.2-Gbyte capacity is beyond what mainstream customers are buying for their desktops. Moreover, the areal density in the 3 to 3.5 Mbits/in.2 is equal to what early GMR drives deliver. Fujitsu, for example, has moved to GMR heads (see picture), yet its newest products feature similar areal density. You can find similar comparisons in the enterprise segment. Quantum has stayed with MR heads thus far but has matched the 36-Gbyte capacity of Fujitsu’s GMR drives. Even IBM is using a mix of MR and GMR across its enterprise line. That situation will change, however. IBM has just announced the Deskstar 25GP family using GMR heads. The devices include 25-Gbyte, five-platter and 20.3-Gbyte, four-platter drives. The 5.1-Gbyte/platter capacity ups areal density to nearly 4 Gbits/in.2—a level MR heads may be unable to match. IBM has also upped mobile-drive capacity to beyond 14 Gbytes using GMR heads. In further evidence of just how profound the GMR transition can be, Western Digital has moved from shipping drives storing 3.4 Gbytes/platter to ones storing 5.1 Gbytes/platter in about five months. The company has just introduced a 20.4-Gbyte, four-platter drive in the Caviar family. Ironically, Western Digital was a late convert to MR heads and a good example of the rewards and perils implicit in such transitions. The company stayed with thin-film inductive heads after most of the competition moved to MR. The decision initially paid off because the company had experience with the older, less-expensive heads. The delay in transition, however, ultimately cost the company significant market share as its drives lost ground in areal density. To catch up, Western Digital signed a license with IBM to acquire GMR expertise and to buy heads from IBM and now finds itself again on the leading edge. When you judge a vendor’s plans for heads, you must estimate just how difficult a transition might be. You must also consider whether the vendor owns head technology or will buy heads on the open market. As you might guess, the vendors and independent head suppliers have different opinions on the GMR transition. Ted Deffenbaugh, director of strategic marketing at Quantum, believes the transition will be a relatively simpler challenge than the thin-film-inductive-to-MR transition, which entailed changing the way a head worked. Deffenbaugh claims that Quantum has been able to test and plan deployment of GMR heads using the drive platforms and media that it’s currently shipping with MR heads. He believes that Quantum needs to be ready with GMR heads by midyear. Quantum feels so comfortable with the state of the merchant market for such heads, it recently sold off its head-research effort to manufacturing partner MKE (Panasonic, www.panasonic.com). The move continues a trend by Quantum and some other companies to minimize vertical integration. Vertical integration/merchant market Vendors on the GMR forefront, however, see vertical integration differently. IBM, Fujitsu, and Seagate clearly have moved the fastest with GMR and own the most experience and head technology. The companies believe that vertical integration—at least when it comes to owning the head technology—will be a requisite for vendors to succeed down the road. The companies don’t so much disagree with Quantum’s Deffenbaugh about the difficulty of evaluating GMR but believe that moving the technology into volume manufacturing will be a roadblock for many vendors. GMR heads are notoriously sensitive to ESD; IBM and Fujitsu have spent years solving the static problems but don’t say how they did so. Over the last few years, companies have discussed techniques from careful handling of the heads to active protection in preamps as possible static-protection techniques. Mike Chenery, Fujitsu’s vice president of architecture development, claims that disk drives have become more than ever before a technology-centric product and that the companies with the greatest investments in the head/media technology will get the biggest pay-back. Chenery points out that areal densities are approaching the point at which bit-cell sizes approach the size of grains in the media coating. He thinks companies must develop heads and media together to continue the disk-capacity ramp. Officials at Read-Rite, the leading independent head vendor, obviously disagree. Vice President of Advanced Technologies Subrata Dey points out that Read-Rite has demonstrated higher GMR areal densities than any other company. Read-Rite has revealed tests of a GMR head working at 13.5 Gbits/in.2 But Read-Rite has not acknowledged volume shipment of GMR heads. It will ship heads this quarter that support areal densities of 4.5 Gbit/in.2 and claims that it can move to 7.5 Gbit/in.2 by the end of 1999. Read-Rite’s challenge is different from that of the head developers at IBM, Seagate, and Fujitsu. For Read-Rite to be successful, it must test and validate its heads with media from multiple merchant suppliers and captive sources. Its opportunity is also large because it can sell heads to many vendors and has built a company of nearly 20,000 employees on that premise. Some drive vendors are shipping drives with third-party GMR heads. Western Digital, for one, is shipping GMR because of the company’s IBM partnership, which may be an exclusive relationship. But Toshiba is also shipping GMR drives in its notebook products from one or more undisclosed vendors. Toshiba does maintain a head-research organization and partners with third-party head vendors in bringing new technologies into volume production. Hitachi takes a slightly different slant and develops head technology for in-house production yet privately shares that technology with head vendors to ensure multiple sources. It would be tough to judge the correct strategy in the GMR transition. IBM has demonstrated the advantage of methodically deploying the technology when ready to maintain an areal-density lead in all market segments. But the highest volume purchases in any of the market segments rarely fall at the leading-edge capacity points. Spin rates drive performance After head technology, performance comes next on the list of disk-selection criteria, and spin rate appears to be the hot story when it comes to performance. Spinning a drive faster boosts data-transfer rate and reduces rotational latency, which is one component of seek time. Mainstream desktop drives currently spin at 5400 rpm; mainstream enterprise drives, at 7200 rpm. The movement is well under way to boost those spin rates to 7200 and 10,000 rpm, respectively. In the desktop segment, users are demanding faster performance and therefore higher spin rates to interactively manipulate audio, graphics, and video files. Again, IBM leads the way in capacity and performance with its Deskstar 22 GXP that comes in 22- and 18-Gbyte models. Fujitsu and Western Digital also have 7200-rpm drives that top out at the 18-Gbyte capacity level. Both IBM’s and Western Digital’s Web sites have white papers that detail the performance advantages of 7200-rpm drives. In the enterprise market, most vendors and analysts expect servers to drive the demand for 10,000-rpm drives. Although server designers are buying the drives, workstation manufacturers are also using the products, especially for applications such as video editing. Seagate, Fujitsu, IBM, and Quantum all offer such drives in 9-, 18-, and 36-Gbyte models; the Fujitsu units have the fastest data rate. Because Fujitsu has gone to GMR across its product line while others have stayed with MR in the 10,000-rpm products, Fujitsu’s drives top out at 45-Mbyte/sec data rates. The GMR heads deliver the higher data rate because of a higher bit density. You can buy a 9-Gbyte, 7200-rpm drive for $500 to $600 in single units, and the 10,000-rpm drives carry a $100 to $200 premium. Hitachi has even developed a 12,000-rpm drive, although it technically doesn’t fit within the 3.5-in. form factor. Most of the vendors believe the next major change will be to a 15,000-rpm spin rate, although such drives won’t emerge until 2001. The newest 10,000-rpm drives solve the major problems—heat and noise—associated with earlier 10,000-rpm products. The vendors have all gone to smaller 2.5- to 3-in. platters in these drives. The smaller platters weigh less, helping the move to the higher spin rates. The smaller media also reduces wind friction, which, in turn, reduces noise and heat. A year ago, workstations or servers that used 10,000-rpm drives needed active cooling schemes dedicated to the drives. You should always consider and prepare for heat problems with any disk drive, but you could probably slap the newest 10,000-rpm drives into any cheap PC case and successfully use them. A number of other technologies, including read channels and interfaces, must also fall into line to support the higher data rates. A few years ago, it appeared that read-channel technology might become a roadblock at rates below 200 Mbps. Now, however, the read-channel vendors, such as Texas Instruments Storage Products, Cirrus Logic ( www.cirrus.com), and Marvell ( www.marvell.com), are looking at rates higher than 500 Mbps. Marvell has already introduced a digital CMOS read channel (88C4200) that supports 500-Mbps rates. New interfaces support higher rates You’ll also find new interface issues in both the desktop and enterprise segments. For desktops, Quantum led the development of Ultra ATA/66, yet another enhancement of ATA. The 66-Mbyte/sec interface yet again doubles the data rate available to the desktop. According to Quantum, however, the benefits of the interface start with reliability rather than performance. The new standard builds on the foundation laid in Ultra ATA/33 to add reliability via error checking and parity. Moreover, the Ultra ATA/66 standard mandates an 80-conductor ribbon cable with added grounds to improve the reliability of data transfers. Ironically, Western Digital in October 1998 was first to market with an Ultra ATA/66 drive. Western Digital also has a thorough Ultra ATA/66 white paper on its Web site. To date, Quantum, IBM, Fujitsu, and Maxtor have all announced support. Unfortunately, Ultra ATA/66 drives won’t provide an advantage until support is available on the host side. You may be able to buy add-in cards that support the faster rate in the first quarter, but don’t expect motherboards with support until after midyear. In the enterprise end, SCSI and FC-AL will continue to coexist for some time. Some observers believe that FC-AL drives will start to supplant SCSI drives in arrays, whereas others believe that SCSI drives will work on the inside of a box with Fibre Channel connections between a host system and an array or other storage subsystem. List prices from market leader Seagate would indicate near price parity for SCSI and FC-AL drives, but the list prices obviously aren’t holding up in the distribution channel. Quantum’s Deffenbaugh points out one problem for all FC-AL drives. The drives require powerful transceivers that add 2W of power dissipation, and that power draw isn’t welcome in drives or arrays. Except for Quantum, all of the enterprise vendors offer Ultra2 SCSI or FC-AL versions of their products. IBM also still offers Serial Storage Architecture (SSA) drives that find use mainly in IBM systems. SSA drives established an early beachhead in some video-server applications as well ( Reference 3 ). Quantum pledges support for FC-AL on the 10,000-rpm Atlas family, which will debut in the first half of this year (see picture). Quantum, however, concentrated on being first to market with the latest flavor of SCSI. The SCSI Trade Association ratified the SCSI 3 standard in July ’98, and Quantum and several partners shortly thereafter described a flavor called "Ultra 160/m SCSI" that adopts several key features—chiefly double-transition clocking—from SCSI 3 to deliver 160-Mbyte/sec data rates. Quantum has announced Atlas drives that support the new interface that is arguably faster than FC-AL. Looking ahead The rest of 1999 should prove interesting for designers who follow the drive industry. By year’s end, it should be clear whether vendors that don’t own head technology can compete in the GMR era. Maxtor, for example, has been a recent leader in market growth using outside suppliers for virtually every component. But Maxtor, like Quantum, must now move to a new technology without the aid of training wheels. IBM, conversely, has always possessed the leading storage technologies, but the huge company has at times not been quick enough to devote resources and capture available market share. The company seems intent on leading in all market segments, and perhaps only Fujitsu has a competitive research base. Seagate also seems intent to move back to vertical integration with a renewed emphasis on R&D. As for new products or concepts that might emerge, you can expect a few goodies. IBM should during the first half of this year ship the 1-in. Microdrive that stores 170 or 340 Mbytes. Several companies, including Hewlett-Packard ( www.hp.com), demonstrated handheld computers using the Microdrive at Fall Comdex (Las Vegas, NV, www.comdex.com). Sony ( www.sony.com) is also cranking up the pressure on drive vendors to build a product more suitable for consumer audio-visual devices. It appears that efforts are almost dead to build an IEEE-1394-based drive with a SCSI-like command set. But Sony claims that the consumer-electronics companies need a drive with inherent support for 1394 isochronous data transfers that enable digital video applications. Both Quantum and Western Digital have signed agreements to work with Sony on such products, and you could see prototypes by midyear. References
|
||||||||
 Maury Wright, Technical EditorYou can reach Technical Editor Maury Wright at 1-619-748-6785, fax 1-619-679-1861, maury-wright@home.com. |
||||||||
GMR heads drive areal densities past 5 Gbits/in.2For more information:
For more information on products such as those discussed in this article, use EDN's InfoAccess service . When you contact any of the following manufacturers directly, please let them know you read about their products in EDN.
Fujitsu Computer Products of America Inc
San Jose, CA
1-408-432-6333
www.fcpa.com
Hitachi America Ltd
Brisbane, CA
1-415-589-8300
www.hitachi.com
IBM Storage Systems Division
San Jose, CA
1-888-426-5214
www.ibm.com/harddrive
Imation Corp
Oakdale, MN
1-651-704-4000
www.imation.com
Iomega Corp
Roy, UT
1-801-778-1000
www.iomega.com
Maxtor Corp
Milpitas, CA
1-408-432-1700
www.maxtor.com
Mitsubishi Electronics
Sunnyvale, CA
1-408-730-5900
www.superdiskdrive.com
Quantum Corp
Milpitas, CA
1-408-894-4000
www.quantum.com
Samsung America
San Jose, CA
1-408-544-5200
www.sem.samsung.co.kr
Seagate Technology Inc
Scotts Valley, CA
1-408-438-6550
www.seagate.com
Sony Electronics
San Jose, CA
1-800-352-7669
www.sony-hifd.com
Teac America
Montebello, CA
1-213-726-0303
www.teac.com
Terastor Corp
San Jose, CA
1-408-914-4000
www.terastor.com
Texas Instruments
Storage Products
Tustin, CA
1-714-573-6000
www.ti.com
Toshiba America
Irvine, CA
1-949-457-0777
www.toshiba.com/taecdpd
Western Digital Corp
Irvine, CA
1-949-932-5000
www.westerndigital.com
Talkback
