Storage technology roundup
The scoop on the newest developments in magnetic, optical, and even holographic media
Storage technology has come a long way, and manufacturers continue to improve its speed, reliability, and throughput. This article examines six types of storage and takes a peek ahead at emerging technologies like holographic media and laser optical tape systems. (4,500 words)
Since the days of punch cards, computer manufacturers have strived to squeeze more data into smaller spaces. That mission has produced competing and complementary data storage technology including electronic circuits, magnetic media like hard disks and tape, and optical media such as compact disks.
Today, companies constantly push the limits of these technologies to improve their speed, reliability, and throughput -- all while reducing cost. The fastest and most expensive storage technology today is based on electronic storage in a circuit such as a solid state "disk drive" or flash RAM. This technology is getting faster and is able to store more information thanks to improved circuit manufacturing techniques that shrink the size of the chip features. Plans are underway for putting up to a gigabit of data onto a single chip.
Magnetic storage technologies used for most computer hard disks are the most common and provide the best value for fast access to a large storage space. At the low end, disk drives cost as little as 25 cents a megabyte and provide access time to data in ten milliseconds. Drives can be ganged to improve reliability or throughput in a Redundant Array of Inexpensive Disks (RAID). Magnetic tape is somewhat slower than disk, but it is significantly cheaper per megabyte. At the high end, manufacturers are starting to ship tapes that hold 40 gigabytes of data. These can be arrayed together into a Redundant Array of Inexpensive Tapes (RAIT) if the throughput needs to be increased beyond the capability of one drive.
For randomly accessible removable storage, manufacturers are beginning to ship low-cost cartridges that combine the speed and random access of a hard drive with the low cost of tape. These drives can store from 100 megabytes to more than one gigabyte per cartridge.
Standard compact disks are also gaining a reputation as an incredibly cheap way of delivering data to desktops. They are the cheapest distribution medium around when purchased in large quantities ($1 per 650 megabyte disk) which explains why so much software is sold on CD-ROM today. With desktop CD-ROM recorders, individuals are able to publish their own CD-ROMs.
A number of other emerging technologies promise to make significant advances over today's existing storage technology. Researchers are looking at holograms, electron trapping, and laser optical tape to store even more data in less space.
Hard disk drives
Hard disks are the most common type of mass storage device today thanks to their low cost, high speed, and relatively high storage capacity. The main differences between them are cost, capacity, transfer rate, and access time. Disk drive prices range from 25 cents to a $1 per megabyte, depending on the capacity, size, and manufacturer. They generally have a seek time of less than 20 milliseconds and are able to transfer data at more than 4 megabytes per second.
Drive capacity has been growing thanks in part to improvements in drive platters and read/write heads which access them. The earliest types of read/write heads were made from a single block of magnetic ceramic material (ferrite). The first major improvement resulted in a composite material made from a mixture of iron and other substances. Today, most drives use thin-film heads. Thin-film technology allows head vendors to achieve much smaller physical dimensions and greater control over the production process. This results in better performing products.
To improve the capacity and reliability of their mass storage systems, many companies are using RAID to connect multiple disk drives and have them appear as one device. There are several different RAID configurations which offer varying degrees of reliability and performance.
When trying to maximize performance, RAID can be configured so that a part of each piece of data is written to each drive in the array. When trying to provide maximum reliability, RAID can write each piece of data to two drives for backup.
Most people choose to implement a combination of these approaches. Each piece of data is striped across multiple drives and a parity drive to guarantee reliability. If a drive goes down, data can be reconstructed from the other drives automatically. The best RAID systems even enable the system administrator to hot swap one of the hard drives while the rest of the array is running.
Fully configured RAID systems cost from $1 to $2 per megabyte depending on the system. The throughput of a RAID system is only limited by the channel used to connect it to the workstation. Many vendors such as Andataco, Digital Equipment Corp., and Storage Dimensions have started to ship storage SCSI-2 based storage systems which can move as much as 20 megabyte per second of data. Falcon Systems has raised the bar even higher with its Fast File Pro, an NFS server that uses Fiber Channel to move data at up to 1,066 megabytes per second. (Falcon also offers a line of conventional RAID subsystems.)
RAID is not just for high-end servers. Many manufacturers are targeting desktop RAID systems as well. For example, Integrix recently unveiled the RD10 desktop storage device with a standard Sun pizza box enclosure. The current system can be configured for 6 or 12-gigabytes, and a 24-gigabyte version is due out by the end of the year. The base price of the 12-gigabyte system is $12,000, which includes 32 megabytes of cache memory.
"Our RAID box is the first desktop system specifically designed for the SPARC platform," said Jason Lo, president of Integrix. "A lot of people purchase RAID controllers from one manufacturer and the chassis from another manufacture and put them together under their own label. We developed the design ourselves," he said.
One of the key components for any RAID system is storage enclosure. Andataco has taken this to the next level with its Enterprise Storage Packaging (ESP), a modular architecture of different component building blocks that can be connected for creating storage systems including solid state, tape, optical, and hard disk drives.
The system is designed to enable hot swapping of disks for RAID applications. Inside each chassis is an intelligent controller that combined with PRISM, Andataco's Internet storage management software, can monitor the environment and the disk array as needed. Network managers with proper security clearance can access a disk drive from anywhere on the Internet using Netscape Navigator to communicate with PRISM.
Frank Berry, director of marketing for Andataco noted, "We can manage multiple clusters of drives using our Netscape software." He also pointed out that the same Netscape browser used for interfacing with PRISM can also be used to search through the entire users manual located on Andataco's Web site. (For more about RAID, see SunWorld Online's "RAID-5 explained" September, 1995, feature story.)
Floppy drives get fat
Magnetic disks have become a de facto standard for distributing software in the computer world. However, their key limitation has been their small storage capacity of 1.44 Mb or less. In 1993, the first high-capacity magnetic disk drives, called Flextra, were introduced by Insite Peripherals with a capacity of 21 megabytes per disk. (Silicon Graphics uses these drives in some of its workstations.) By November 1993, the company reports it sold more than 300,000 disks and 60,000 drives.
More recently, the Flextra has been eclipsed by new drives from Iomega and Syquest. Although these drives are designed for the PC environment, the manufacturers have drivers that enable them to work on other platforms as well.
Iomega started shipping the Zip drive last March. The $200 Zip drive is capable of storing up to 100 megabytes of data on a $15 disk. The transfer rate is 10 megabytes per second with a seek time of 29 ms. Iomega has also developed the 1-gigabyte 3½-inch Jaz drive. This drive sells for less than $600, and disks cost under $100 each. The transfer rate is 44 megabytes per second, and the access time is 72 milliseconds.
Syquest was quick to counter with its own 130-megabyte disk drive called the EZ135. It has better performance than the Zip drive with an advertised transfer rate of 19.2 megabits per second and a seek time of 13.5 ms.
For the high end, Syquest unveiled the SyJet 1.3 at Macworld last month. It is capable of storing up to 1.3 gigabytes on each cartridge. The SyJet 1.3 drive will be priced at under $500 each, and disks will sell for $95.
3M (St. Paul, MN), Compaq (Houston, TX), and Matsushita announced the joint development of a 120 megabyte floppy disk drive system that is backward compatible with existing 1.44 megabyte floppy drives. Compaq plans to install the drives as an option on its new desktop computers. This drive is able to achieve such high densities with a laser aligning the disk head with the disk. Consequently, data tracks can be packed tightly.
Larry Teien, a spokesman for 3M said that the drives will be priced in line with the Zip Drive and the EZ-135. The main attraction for users is that a single slot on their computers will be able to handle both old and new disks.
Solid state speed
Although magnetic disks are getting faster, they are still a far cry away from seek times of solid state storage devices. Solid state disks allow seek times on the order of 100 microseconds -- at least 20 times faster than most magnetic disks making them ideal for high speed storage which requires a lot of read and write operations.
However, with prices ranging from $20 to $100 per megabyte, solid state devices cost much more than magnetic media. At the low end, solid state PC Cards (a.k.a. "PCMCIA") are starting to gain ground for mobile computing and other similar applications. At Comdex last November, Apple, Canon, Eastman Kodak, Hewlett-Packard, Lucky Goldstar, Matsushita, Motorola, NEC, Seagate, and Epson announced the CompactFlash standard for storing from 2 to 15 megabytes of data in a single PC Card for computers, cameras, phones, and answering machines.
At the high end, solid state devices are used as caching devices for database applications to speed the effective throughput. Zitel (Fremont, CA) has developed the Storage Co-Processor-II (SCP-II), a caching system designed to enable magnetic disk drive systems to deliver the throughput and response time of solid state systems. A complete SCP-II system costs about $3 per megabyte.
Randy Serafini, director of marketing at Zitel said, "Zitel's original SCP system proved extremely popular with customers that needed high-end database performance, regardless of cost. That's why a SCP system is being used by Oracle Government Group to showcase Oracle database performance and why SAS Institute uses SCP to show performance increases for its application programs. SCP-II is an entirely new platform design for our next generation of intelligence-based technology. It incorporates the latest design philosophies for performance and data security at a significantly lower cost."
Each SCP controller can cache data from up to four 4.3 gigabyte drives into 64 to 256 megabytes of solid state memory. It stands by and watches the read and write patterns generated on the disk and caches the most commonly used data into solid state storage.
For database applications, where the same data is frequently accessed, the system is able to retrieve 85 to 90 percent of all requests directly from cache. According to Serafini the SCP-II is able to accelerate such applications by 600 percent. Applications accessing data in a more random fashion will see less acceleration.
Optical technology moves ahead
Compact disks have emerged as a low cost way of distributing data. Each disk can store 650 megabytes of data and only cost $1 each when purchased in large quantities. For other applications in which fewer copies of the disk are required, a number of desktop CD-ROM recorders are on the market in the $1,000 price range from Hewlett-Packard, Yamaha, Philips, Kodak, and Sony.
One of the drawbacks of CD-ROM is that certain computers work best with certain formats. For example the standard PC format is ISO-9660, does not have any facility for storing long file names or icons. Rock Ridge extensions are added to Unix software to provide file permissions, long file names, and other capabilities that can be used on that platform. Macintosh computers have their own set of extensions for adding file names and icons.
CDR Publisher, from Creative Digital Research, lets developers write all three formats to a CD-ROM simultaneously, allowing them to access their data from any platform. Although developers will still need to create the applications for different platforms, CDR publisher provides a convenient way of carrying text, data, pictures, and video to multiple platforms. The software is available for trial from CDR's Web site.
Writable CDs can also be used for network backup or archiving. Once data has been saved to CD, a company can use a product such as OptiDriver Net from OptiSys to permanently mount these disks on the network. OptiDriver allows up to 254 users to access a bank of CD-ROM drives. Frequently accessed data can be cached to a hard disk to provide better overall throughput. Directory entries in the CD-ROM can be buffered into a local workstation's memory for even faster access.
A new generation of optical disk called the digital video disk (DVD) is scheduled to hit the streets sometime this year. Storing from 4.7 to 17 gigabytes of data per disk, the DVD will primarily be used for distributing data to massive audiences due to the high anticipated costs of the recording process.
Rewritable optical drives can also be used for transaction intensive operations, if they are configured correctly. For example, Frontier Telecommunications (Rochester, NY) recently upgraded to an automated database for capturing call detail records using an array of rewritable optical disk drives. More than 500 transactions per second are recorded into the database, and it is designed to grow to 400 gigabytes of active storage. They are, however, slower than other media such as magnetic or solid state disks.
The rewritable disk drive jukebox for the system was supplied by ATG Cygnet (San Jose, CA). The real key to the system is the Open Archive storage software from InformationView (Rochester, NY) that enables an array of optical disks to be used directly as a database management system.
There are two primary applications of optical disk technology. The first archives documents to a disk for more compact storage. The second way (for hierarchical storage systems) backs up data automatically to different storage systems based on its frequency of use. Providing a third option, Open Archive, lets companies use the large capacity of optical storage directly with a database.
Jim Echter, a spokesman for InformationView, said the company began looking for a database that could scale to extremely large sizes for some telecom applications they were working on in 1991. But they could not find any one that could store a terabyte of data in a database. "We took a look at the terabyte database problem and decided that optical was the only thing that would cost effectively hold it. Then we found out that the database vendors could not understand optical directly."
Using Open Archive, it is possible to store up to 1.6 terabytes of data in a single database. The system costs about 20 to 90 cents per megabyte, depending on the size of the array. Although magnetic media is starting to approach this range in terms of cost, Echter pointed out that managing a large magnetic disk array could be a nightmare. A manager has to deal with the overhead of maintaining all the disks; and if just one goes out, data can be lost. If an optical drive crashes, the disk can be moved to another drive.
Tapes take off
Tapes have traditionally been the cheapest media for storing data, but because they are not as randomly accessible as other media such as disk drives, it often takes longer to find and begin retrieving data. Optimally, tape should be used for archiving data that can all be dumped at once. Storing numerous files that tend to be modified and saved often would yield less than satisfactory results.
More often tape is used in hierarchical storage systems. In this case, users can archive data from hard disks and other sources. When data is needed, the hierarchical storage management (HSM) software can be used to automatically cache data onto the faster hard disk system.
There are three main tape formats primarily used in the open systems arena: 8mm, 4mm (also known at digital audio tape or DAT), and digital linear tape (DLT). 8mm tapes have been around the longest and were the original alternative to reel-to-reel tape and low-capacity QIC tapes.
The current state of the art in 8mm technology is a 7 gigabyte drive from Exabyte The drives cost $2,500 to $4,000, and cassettes run about $15 or $.003 per megabyte. The average time to find a file is about 85 seconds, and it can support a sustained transfer rate of 4 megabytes per second.
Exabyte is actively working on Mammoth, the next generation of this technology which should begin shipping in mid 1996. Capable of storing up to 20 gigabytes of uncompressed data on a single drive, Mammoth uses a new magnetic media called advanced metal evaporated, which is more sensitive than the magnetic particle tape used in most tape systems.
In order to squeeze data into the same-sized cartridge, the tape used by Mammoth is much thinner than conventional tape. Strain on the tape is reduced through the use of equal force applied to both reels inside the tape to turn it instead of with a capstan.
Mammoth also incorporates a new more sensitive scanner design for reading the tape. It is larger than the scanners for most 8mm tapes and can read and write over more area. A preamplifier is built into the scanner making it more sensitive to the tape. In fact, Exabyte purchased the portion of the Grundig Corp. in Germany that designed the scanning system so it could control the technology.
Mammoth is only the first of a line of high-capacity drives from Exabyte. Tim Gerchar, senior product marketing manager at Exabyte noted, "We did not build Mammoth to drive one product out of it. There will be multiple generations of Mammoth products."
After the consumer giants failed in bringing digital audio tape into the home several years ago, the tape lived on and assumed a new life as the 4mm cartridge used for computer applications. Although the recorders tend to be cheaper than 8mm tape, the cost of the tape is higher.
The latest in 4mm technology is Sony's SDT 7000. Each tape can store up to 4 gigabytes of data with a transfer rate of 6.2 megabytes per second. The average file seek time is 30 to 35 seconds which is slightly better than with 8mm tapes today. Drives cost less than $1,200, and tapes are from $10 to $15 each.
One of the problems with DAT is that the drives are not sealed. Consequently, DAT drives must be cleaned every 20 to 50 hours to ensure reliable operation.
Sometime this year, Hewlett-Packard plans to release the C1537A 4mm tape drive. Capable of storing up to 12 gigabytes on a cartridge, it will have a transfer rate of 4 megabytes per second and drives should cost about $1,200 each.
DDS-4, the next generation of 4mm technology capable of storing 24 gigabytes on a single tape should be available in 1997.
Digital linear tape, first introduced by Digital Equipment Corp. in 1991, has quickly been adopted for high-end storage applications due to its large capacity and high transfer rate. Today the drives sell for $6,500 each, and the retail price for a 6 gigabyte cartridge is about $35. At the high end, DLT systems from Conner Peripherals can store up to 20 gigabytes on a single cartridge and support a transfer rate of 20 megabytes per second.
By the end of 1996, the capacity of DLT could go to 30 gigabytes and support a transfer rate of 40 megabytes per second, according to Linda Kempster, president of Strategic Management Resources (Bowie, MD).
For many storage applications, a single tape is either too slow or too small. In order to improve the speed of tape systems, a number of companies are installing a Redundant Array of Inexpensive Tapes (RAIT), similar to technology used in disk drives. By ganging the disk drives together, it is possible to write to all four of them simultaneously. RAIT controllers are available from a number of vendors such as Andataco and Qualstar (Canoga Park, CA).
When a company needs to archive large quantities of data, tape libraries are used to automatically load and play tapes as they are needed. Most of these systems consist of a box that is designed to work with one or a few types of media. One of the most versatile robotic loading systems around is ARCHIE (ARCHival Information Exchanger) from Borret Automation Technologies (Westlake Village, CA). ARCHIE is a robotic system that can be built into the room. The robotic arm moves around the tracks and finds and loads media as needed. Because the system is so flexible, the media can be upgraded without affecting ARCHIE's operation. In addition, it can scale to support enormous sizes. At the high end, a 39-tower ARCHIE system could store over 23 petabytes of data.
The future of storage
There are a number of technologies that promise to increase storage capacity and performance when they arrive. One of the most exciting lines of research is holographic storage technology. These units will eventually store terabytes of data in a sugar-cube size crystal. The Advanced Research Project Agency (ARPA) recently kicked in another $22 million dollars to improve the technology, which builds on other holographic storage research ARPA has been funding over the past decade.
With this technology, a laser is used to etch a hologram into a crystal or other holographic media. That pattern can then be read back from the crystal by beaming a laser at the appropriate angle. Different patterns can be read by changing the angle. Consequently, thousands of holograms can be stored in a single crystal simultaneously.
It is a fascinating concept, but so far no one has found a medium that will work for the device. Initially, the focus was on using some kind of crystal. After many disappointments, however, one of the leaders in this research, Tamarack Storage Devices (Austin, TX) has turned its attention to using polymer disks, which are similar to compact disks.
T. C. Lee, director of optical engineering at Tamarack Research said, "We have developed all the bells and whistles to make a drive, but the media is far away."
The ideal material will have a high signal to noise ratio so that data can be found and read quickly without having to flood the holographic media with so many photons that it gets hot or consumes a lot of laser power. In order to confirm data, the medium also needs to be very sensitive to light and be stable enough so that holographic fringes do not move around inside.
Lee estimated it will take three to five years before holographic storage makes it to the market. He pointed out that no one has found a material that will work for this yet. Consequently holographic storage could either take off tomorrow if the technology emerges, or never if no one can find an appropriate material for making it.
Another exciting technology for high capacity storage is the Laser Optical Tape System being developed by LOTS Technology. Using a 3480 cartridge form factor with internal optical tape, the drive will support SCSI interfaces and will fit into the standard 3480 cartridge bays of automated tape libraries. Each cartridge will be able to store a terabyte of data and support a transfer rate of 120 megabytes per second. The first generation of LOTS systems are scheduled to ship at the end of 1996.
LOTS uses a patented head to keep the tape in place without touching the sides, minimizing wear. Thanks to these heads, the tape has a rated life of 25 years. The head will also improve the read and write speed by splitting the laser output into 45 separate beams.
Optex (Rockville, MD) is working on a new kind of optical disk that uses the principal of electron trapping to store data. Data is stored by changing the state of electrons in the media. The first generation of ETOM devices was unveiled in the fall of 1992. They can store 14 gigabytes of uncompressed data on a single double-sided disk with a transfer rate of 120 megabytes per second. For now, the technology remains in the lab as Optex works on developing a cheaper laser and improving disk manufacturing techniques. They expect to start shipping prototypes within the next year.
There is a wide variety storage technologies available. At the low end, it is possible to merely pick the cheapest technology that will store the amount of data you need. But as your applications scale up, the complexity of storage options increases.
As Richard Lee, president of Data Storage Technologies acknowledged, "At the lower end of the spectrum, people shop based on the best advertisement. Over time as you need to get more capacity or bandwidth, you can get stuck into a corner. If you build up all these different storage formats, you have to figure out how to migrate it from one to the next in the future. It is not hard to imagine 100 terabytes of data that has to be migrated and if you have to do that at some SCSI rate, you will be there till the end of time."
If you have technical problems with this magazine, contact email@example.com
Linda Kempster first became interested in storage technology while working at Northrop Aircraft in 1983. That interest got here involved in evaluating the storage requirements for a number of very large scale projects. For example the Government-sponsored Earth Orbiting System (EOS) will require 125 gigabytes of storage a day. Every day.
Since then, Kempster has been evaluating all removable mass storage technologies. She has compiled a book about the industry called Media Mania. The book starts by detailing the needs of future storage systems such as EOS, the IRS, and the CIA.
Media Mania then digs into the meat of different kinds of technology. Thirty-two chapters cover all of the different tape and optical formats and provide hard data on price and performance for each of them. There are also chapters on tape libraries. One of the dangers of this kind of book is that the data is quickly outdated. But Kempster has been quick to keep up with the times and has just recently finished the third update in only two years.
The $57 book is available from the publisher, Avedon Associates based in Potomac, Md., or directly from the author at firstname.lastname@example.org.
About the author
George Lawton (email@example.com) is a computer and telecommunications consultant based in Brisbane, CA. You can visit his home page at http://www.best.com/~glawton/glawton/. Reach George at firstname.lastname@example.org.