A Storage Primer

Here are the major players among the plethora of electronic data storage options.

by / June 30, 1998 0
The amount of electronic data piling up in agencies around the country is staggering. Document management, data warehousing, multimedia and the World Wide Web, combined with traditional transaction-driven applications, are causing storage to become one of the fastest-growing segments of the industry.

To meet the demand, solutions are coming out of the woodwork. There are more storage technologies than ever before for laptops, desktops, networks and the enterprise, and new ones are on the horizon. If you're wondering which ones will take us into the 21st century -- disk, tape, magnetic, optical -- actually, it will be all of them. Even tape, the oldest of them all, because tape libraries can store more bytes for the buck.

You'd have your head in the sand not to notice how cheap disk drives have become over the past couple of years, cramming more bits and tracks into the same space for even lower prices. There seems to be life left in those hard disks, at least for a few more years.

But combination magnetic/optical technologies from Terastor and Quinta show tremendous promise. There's also DVD-RAM -- a pure optical recording technology. Seemingly aimed at the desktop market with 2.6GB disks and access times like a CD, DVD-RAM could be a universal storage technology within the next five years. DVD-RAMs are just coming to market.

The next five years are going to be very interesting. There will no doubt be failures, but things have a tendency to stick in this biz. After all, they're still making open reel tapes! Here are the various storage trends and solutions.

Enterprise Storage Management

Enterprise storage is the collection of all online, nearline and offline disks and tapes within the organization. While online implies instantaneous access, nearline devices -- primarily tape and disk libraries -- take several seconds to a minute. Offline means on the shelf --either locally or offsite.

Enterprise storage management includes routine backup, archiving, disaster recovery and hierarchical storage management (HSM), which moves online data to nearline to offline. The year-2000 problem only adds to the burden, requiring copies of production databases in order to test real data.

Storage management is becoming critical, and one would wish for a virtual storage solution that could simply take all the data you hand it and automatically take care of all backup, mirroring and related tasks. We're not quite there yet, but we're heading in that direction.

Software solutions are becoming more integrated. For example, IBM's ADSTAR Distributed Storage Manager (ADSM) is a comprehensive storage management system that supports more than 30 platforms and offers integrated backup, disaster recovery and HSM.

The Storage Network Industry Association (SNIA), formed in 1997 and spearheaded by Strategic Research Corporation's Michael Peterson, is looking into the "missing glue." SNIA is researching the categories that should become standard protocols for performing fundamental storage tasks, so that hosts can talk to storage systems using a common language.

Hardware solutions are becoming increasingly intelligent, with more combinations of memory, disk and tape combined in one box. For example, IBM
introduced the virtual tape server, which caches tape files on disk, and CD-ROM servers are about to follow suit. By the end of the year, NSM Jukebox will release a CD-ROM file server with hard disks that cache frequently used data.

RAID

A redundant array of independent disks (RAID) is now commonplace for mission-critical data. There are various fault-tolerant configurations of RAID, but what it boils down to is that, if a drive fails, RAID rebuilds the lost data automatically. While big RAID boxes abound, Microarrays from Aiwa come in 5GB to 16GB capacities and plug into a single full-height drive bay, and that includes two or three drives. Appearing as a single SCSI drive to the server or workstation, you can't beat that for simplicity.

Network Storage and Storage Networks

The common network storage is a file server or database server that contains its own SCSI (sometimes IDE) hard drives directly connected via cable, either within the cabinet or in an external enclosure. The file server stores programs and data, and shares files with users on the network. The database server stores the data as well, but it contains and runs the database management system. It's also called an application server, because it does data processing -- direct retrieval, search, etc. -- and sends the results back to the clients.

As network storage proliferates, a type of file server known as a network attached storage (NAS) system is becoming very popular. Companies such as Network Appliance, Auspex and Artecon are promoting the NAS, a specialized file server known for its ease of installation. Instead of a full-blown operating system inside, the NAS typically uses a microkernel -- a small, custom-control program that processes only file I/O requests, such as NFS (UNIX), SMB/CIFS (Windows) and NCP (NetWare).

Meridian Data has brought the NAS down to the workgroup level, where you can get a quick storage boost by simply plugging a Snap!Server into your network hub. Your Windows, OS/2, DOS and UNIX clients have an extra 4GB to 12GB of storage almost instantaneously. The Snap!Server is indeed a snap, with nothing more than an on/off switch and an Ethernet port. One would wish all computer components were designed this way.

The NAS can be added and removed from the network like any other node, and thus wins for ease of use, whether it's a large unit or a small one. But the NAS suffers because it is a network-attached device using communications protocols, such as TCP/IP, to transport the data to the requesting users. Network traffic is often erratic, and that has led to the concept of the storage area network (SAN).

The Storage Area Network (SAN)

The SAN is a back-end network connecting storage devices via peripheral channels such as SCSI, SSA, ESCON and Fibre Channel. Unlike the NAS, the SAN is channel attached. The NAS is connected to the front-end communications network (Ethernet, TCP/IP, etc.), while the SAN connects hosts to the storage system via a high-speed, dedicated storage network. What gets confusing is that NASs are often contrasted with SANs, but they are included under the "storage network" umbrella. The easiest way to remember the difference is that the NAS is network attached, and the SAN is channel attached.

There are two ways of implementing SANs: centralized and decentralized. A centralized SAN ties multiple hosts into a single storage system -- a big RAID box with huge amounts of cache and redundant power supplies. The cabling distances allow for local as well as campus-wide and metropolitan-wide hookups over peripheral channels rather than an overburdened network. Even SCSI distances have been extended. GigaLabs' SCSI switches can talk to each other over 20 kilometers of fiber. This centralized storage topology is also not entirely new, being one commonly employed to tie a server cluster together for failover.

EMC pioneered the centralized SAN. Its Symmetrix line has made the company very successful. Also, the CLARiiON line from Data General is widely respected. Storage Computer offers an interesting twist: Its StorageSuite systems employ a virtual memory architecture that separates the hosts from physical storage boundaries and enables each one to see just the amount of storage assigned to it.

Having data in one place is obviously easier to manage. In addition, some storage systems can copy data for testing, routine backup and transfer between databases without burdening the hosts they serve. The glass house may be coming back; not because we want to centralize processing necessarily, but because we want to keep our data manageable and safe.

Fibre Channel (FC) is a driving force in the SAN arena, because it supports existing peripheral interfaces as well as network interfaces. FC can be configured point-to-point, in an arbitrated loop (FC-AL) or via a switch. FC-AL runs at 100MB per second. That's bytes, not bits, which means FC is equivalent to Gigabit Ethernet. In fact, the FC-0 physical level of FC is where Gigabit Ethernet came from. An FC switch runs at 100MB per second in both directions from each port. What's interesting about FC is that your hosts can talk to the storage system via SCSI, and to each other via IP, all over the same topology.

If a centralized storage system is not feasible, a SAN can connect multiple hosts with multiple storage systems. FC easily supports a distributed SAN environment, while GigaLabs is breathing new life into SCSI topologies. Considering the proliferation of file servers in an enterprise, distributed SANs will be widely used over the next few years.

Removable Hard Disks

SyQuest disks were the industry standard for desktop cartridges in the early 1990s, but they took a back seat to Iomega's Zip and Jaz disks the past couple of years. The 1GB Jaz disks are quite popular, running less than $300 for the drive and $90 for the disk. The new 2GB version bumps the drive up to $550 and $150 for the disk.

However, SyQuest is coming back strong with a very affordable line that includes the SyJet, SparQ and Quest. The $400 SyJet uses 1.5GB cartridges that cost about $120, while the $200 SparQ drive offers extraordinary value with 1GB cartridges costing only $35. SyQuest might have a real winner with its $600 Quest drive, which uses 4.7GB cartridges that cost about $200. There's nothing faster with more capacity in the removable market. If it proves reliable, the Quest could become a standard second drive shipped on all PCs.

Magneto-Optic Disks

For robust storage, there's nothing quite as good as magneto-optic (MO) disks, which are almost as fast as hard disks and far more durable. MOs have a 30-year shelf life and can withstand a million rewrites. Robotic MO libraries provide fast, high-capacity nearline storage. For example, Plasmon IDE provides a line of 5.25-inch MO jukeboxes from 52 cartridges with 135GB to 500 cartridges with 2.6TB.

The 3.5-inch drives cost about $500, and the cartridges, which come in 230MB and 640MB capacities, cost from $10 to $20. The 5.25-inch drives range from $1,900 to $3,000, use $25-$75 cartridges and come in 1.3GB, 2.6GB and 5.2GB capacities. The only downside to MO has been its higher cost.

With MO, the laser reads by sensing the reflection through the magnetic bit, which differs depending on its polarity. For recording, the laser heats up the bit to the Curie point, and the magnet changes the polarity. Since the laser and magnet are located on opposite sides of the platter, double-sided 5.25-inch media must be flipped. For example, on a 2.6GB disk, only 1.3GB is online. Cartridges in single drives are flipped manually, while the robotics in the jukebox does it automatically.

CDs and DVDs (Phase Change)

Phase change is a pure optical recording technology developed by Panasonic. The laser alters the material in the recording layer to make it less reflective (amorphous) and, using a different-strength pulse, restores it back to its crystalline, reflective state. Phase change was first used in the 1980s for write-once (WORM) drives, then in a line of 5.25-inch drives, and, more recently, in Panasonic's $500 Phase Change Dual (PD) drive, which supports CD-ROMs as well as a 650MB rewritable, phase change cartridge that costs about $30.

Phase change drives have not been popular in the United States, but that is about to change, because phase change happens to be the technology in rewritable CD-RWs and DVD-RAMs. In fact, PD and DVD-RAM cartridges are almost identical, except that single DVD-RAMs hold 2.6GB (single sided) and 5.2GB (double sided).

It's going to be a toss-up between CD-RW and DVD-RAM drives. Although CD-RW drives are in the $400 range, and the media is below $20, DVD-RAM drives are finally coming to market. DVD-RAM drives cost more than CD-RWs (about $800 instead of $500) right now, but the media is less when you compare a $19 CD-RW with 650MB to a $24 DVD-RAM that holds 2.6GB. An advantage to CD-RWs is that CD-ROM drives can read them, but only the newer MultiRead drives are shipping now, not the millions already out there.

Access times on a DVD-RAM disk are slightly slower than a CD-ROM, so they will not compete with hard disks or MO disks. But DVD-RAM could be a no-brainer for cloning an entire disk for desktop backup. For high-capacity nearline storage, consider that a 150-disk DVD-RAM jukebox will hold nearly 800GB and should cost about $15,000. DVD-RAM drives will support all CD formats, and Panasonic's drive is backward compatible with its PD cartridges.

You may have heard of DVD+R/W. Storing 3GB instead of 2.6GB per side, this technology from Sony and Philips competes with DVD-RAM. If DVD+R/W muddies the waters, DVD-RAM will be slower to take off. You may have also heard of NEC's MMVF, a 5.2GB rewritable DVD. NEC has no current plans to bring this to the United States. Let's hope it doesn't change its mind.

Write-Once Media

With CD-R drives in the $300 range, and media less than $2 a pop, you can archive 650MB at a pretty penny. CD-R discs can be read in most all CD-ROM drives, so CD-ROM towers and jukeboxes can serve double duty.

DVD-R is another animal. Currently, Pioneer makes the only drive on the market, which costs $17,000. Intended for DVD mastering only, the second generation is expected next year at $5,000, which will include DVD-R/W capability. DVD-R/W is a rewritable version of DVD-R, not intended to compete with DVD-RAM, thank goodness.

The only vendor left in the large-format WORM market is Philips LMS. Although Kodak is coming out with newer 14-inch media this year, it is exiting the business in 1999. Philips 12-inch 12GB media are used in large libraries that provide up to 1.2TB with very fast nearline speeds of only two to three seconds to load a cartridge. A six-drive, 107-cartridge unit costs $295,000. Philips is switching from ablative WORM to 30GB phase change disks next year.

Enterprise Magnetic Tape

In 1997, IBM introduced its Virtual Tape Server (VTS) for big iron MVS systems. The VTS combines disks and a tape library to improve performance (the disk is a cache) and to eliminate tape waste. A legacy of older systems makes it more convenient to put one file on a single tape, but since files are often considerably smaller than the cartridge, there is tremendous waste. The VTS keeps track of all the files and stores multiple files on one volume. While IBM's VTS uses its own Magstar library, the Scimitar system from Sutmyn uses whatever library is attached to the mainframe.

IBM's Magstar tapes are an enhanced 128-track version of its half-inch 3480 and 3490 single-hub tape cartridges, the successors to the venerable open reel tapes. Magstar (3590) cartridges hold 10GB (raw) with 20GB expected in the fall.

Don't confuse Magstar MP with Magstar. Magstar MP uses an 8mm, 5GB dual-reel, cassette-style cartridge, not the single-spindle type. It's SCSI based, not ESCON, and it starts in the middle of the tape. Although MP (multipurpose) media is more expensive ($50 for 5GB compared to $40 for 10GB), an MP drive is $8,000 compared to more than $30,000 for the Magstar.

DDS - 4mm

Users really like the compactness of the small 2-inch-by-3-inch, 4mm DDS tapes, first introduced in 1989. This helical scan technology is popular in the low-end server market, and DDS-3 brings the ante up from the original DDS-1 spec of 2GB to 12GB. Drives run $600 to $1,100, with cartridges $15 to $45.

Although not mainstream, there is actually something smaller than 4mm tape. If you're looking for a $299 parallel port tape backup with 1GB cartridges smaller than a matchbook, check out the Pereos drive, available from J&J Peripherals. The only downside is that there is hardly any room to write the contents on the label.

QIC Lives On

With more than 14 million quarter-inch-tape (QIC) drives on the market, QIC is the leading tape-backup technology. Although primarily a desktop product, QIC is poised to become a leader in the server market. QIC comprises two form factors: the 3.5-inch minicartridge (DC-2000) and the 5.25-inch data cartridge (DC-6000). Imation has breathed new life into the minicartridge with its Travan format, which uses 8mm tape that ranges from 400MB to 10GB -- a far cry from the first 40MB DC-2000 cartridge.

Verbatim provides a line of QIC-EX tapes, which extend the housing for more tape length and capacity. Travan drives support QIC and QIC-EX cartridges and run about $250 to $350 for the 4GB (TR-4) models.

Tandberg Data is doing for data cartridges what Imation did for minicartridges. Once thought to be on the verge of extinction, Tandberg has taken the DC-6000 into the midrange market with its MLR-3 drives, which use 25GB cartridges that sell for about $70. The $2,000 drive ought to become very popular, because it adds up-to-date components to traditional, reliable, linear tape technology at a competitive price. A 40GB version is expected in early 1999.

Floppy Replacements

The Zip drive has become the most-widely used rewritable, removable medium since the floppy disk. The drives are getting close to the sub-$100 level, and the media is about $13. But Zip is finally going to get some competition from Exabyte's LS-120 and Sony's HiFD, which offer something Zip does not. They read and write regular floppies, so the same drive serves double duty. Plus, the LS-120's 120MB and HiFD's 200MB provide more storage. LS-120 drives sell for about $120 and media is $15.

HiFD drives should be shipping by now. Sony could be the winner here, if it moves quickly and makes the media affordable. After all, Sony invented the 3.5-inch floppy!

Laptops and Handhelds

Avatar offers laptop users its 250MB Shark hard disk for $250 (less $50 with the current rebate). Connecting via PC card or parallel port, Shark uses a very thin, lightweight 2.5-inch by 3-inch cartridge that costs about $35. It ought to become very popular for road warriors. The Pro version attaches via PC card and parallel port.

Iomega, makers of the popular Zip and Jaz disks, has developed Click! -- a tiny hard disk for PDAs, digital cameras and other handhelds. The 40MB cartridges are expected to sell for about $10.

A really neat Optical ROM (OROM) data card is coming out next year from Ioptics. OROM provides 10ms access to 128MB in a thin, plastic 2.25-inch-by-1.75-inch data card that connects via PC card or USB. It uses stationary organic LEDs (OLEDs) that shine through 5,000 "data patches" on the card, which are two-dimensional images that each hold 32KB of data. OROM cards are pressed in a manner similar to CDs. Write-once OROM writers are also expected for desktop PCs.

To the Future

When one considers the amount of paper that still sits in file cabinets, if the paperless office is ever to truly exist, we need all the technological breakthroughs we can get.

Happy Storing!

Alan Freedman's Computer Desktop Encyclopedia on CD-ROM is "the" award-winning reference on the computer industry. It contains more than 9,000 entries from micro to mainframe and includes technical drawings, photographs and charts. It runs on Windows 3.1, 95, NT or Mac SoftWindows with a cost of $44.95 for the CD-ROM and $34.95 for the floppies. Contact The Computer Language Company at 215/297-8082. Fax: 215/297-8424.
July Table of Contents