Network Storage Buyers Guide

That sentence is wrong grammatically, but information systems pros who must store and
manage the burgeoning bits and bytes their agencies produce will tell you it’s right
on the money.

And an exponential growth of networks ensures that the flow of data won’t stop.
Network managers are finding that using sneakernet to back up hard drives and
floppies—even for small networks—no longer works.

On larger networks the small single-drive tape and disk backup systems on network
servers cannot begin to back up and archive the hundreds of gigabytes of information
generated daily. The situation calls for automated storage.

Setting up an automated storage system can be daunting, but failing to do so can be
disastrous.

According to most estimates, the data storage requirements in typical government
departments or agencies will double annually for the foreseeable future.

Meanwhile, networks everywhere suffer from shrinking backup windows—the amount of
network-generated data escalates but the time allotted to back it up doesn’t. Many
organizations are working in vain to back up and store hundreds of gigabytes per hour
during a single overnight shift, according to an Exabyte Corp. white paper on tape
storage.

Struggling along with an antiquated manual storage system saves, in the long run,
neither money nor labor. Network management expenses, including backup and storage, can
exceed $200,000 per year for even a small PC LAN, according to a report from Strategic
Research Corp., storage market researchers in Santa Barbara, Calif.

The same report notes that PC LAN managers spend an average of 954 hours per site per
year administering data backup tasks, and an average of 154 hours per year of user
productivity is lost because of backup errors.

This Buyers Guide focuses on three types of automated backup and storage systems: tape
libraries, tape autoloaders and optical jukeboxes, any of which could become the core of a
coordinated network storage strategy for your organization.

Automation pays off. Automated tape or optical disk storage subsystems result in
reductions in data storage costs, greater systems reliability and increased data security.

Because they use multiple drives and hundreds of cartridges across several servers on
the network, tape libraries eliminate the need for an expensive disk or tape backup system
on every server. According to Exabyte, even a low-end $10,000 dual-drive tape library plus
software, which costs about $100 per server, can be shared across 10 servers. Such sharing
yields dual-drive performance for about $1,700 per server—much less than the cost of
a slower single drive system installed in each server.

Management costs drop when time spent on backup sessions, manual media management file
and disk management, and data recovery tasks is reduced.

Security improves because automated storage technology is inherently reliable and, in
many cases, fault-tolerant, eliminating data loss from a power outage. If the IS staff
fails to do a backup session on time, or insert the wrong cartridge in a drive, the
results can be costly. Automated systems don’t make human errors.

I’ll get into details of the technologies that support tape libraries, tape
autoloaders and optical jukeboxes, as well as some of the tradeoffs of using tape vs. hard
disk media. But first, let’s look at optimum situations for using each type of
automated storage:

But because most data on a hard disk isn’t accessed regularly, IS managers are
finding it cost-effective to move it to an automated tape system for nearline storage
until it is required again. HSM software packaged with many tape libraries or bought from
third-party suppliers automates data storage and retrieval.

All automated storage systems use basic sets of hardware storage media such as tape
drives or optical or CD-ROM drives, but they use robotics and management software to
automate storage.

Loading or unloading tape cartridges or optical disks in and out of drives is handled
automatically by the system’s robot instead of manually by IS personnel. The
automation brings with it improved performance and reliability, higher levels of data
security, lower management costs and unattended backup. It also usually results in lower
per-megabyte storage costs than those achieved using standalone tape or disk drives in
network servers.

DLT libraries are high-capacity systems for storing, retrieving, reading and writing
multiple magnetic tape cartridges. They contain multiple storage racks for holding the
cartridges and a robotic mechanism for moving the cartridges to the drives.

A small tape library usually contains several drives and may hold between six or eight
to several hundred cartridges. Large libraries may hold 100 or more drives and thousands
of cartridges.

Different tape libraries are designed to hold different types of tape cartridges,
including quarter-inch cartridge (QIC), 4-mm digital audio tape (DAT), 8-mm helical scan
and newer tape technologies, including DLT and advanced intelligent tape (AIT) (see story,
Page 51).

For several reasons, this guide focuses on DLT libraries—DLT media is widely
available, the system offers high capacity, solid reliability and high data throughput
rates and data-access times.

Most DLT libraries are also highly modular, which lets you easily scale them up to much
larger systems simply by adding more drives and cartridges. However, other tape
technologies such as AIT are also catching on. Many listed DLT library vendors make
several kinds of drives.

The high costs of typical tape libraries are offset in part by their amazing
functionality. Tape libraries can handle the entire data management process without human
intervention. Data cartridges are loaded and unloaded automatically. Online storage
problems are automatically identified and recover steps are initiated immediately in case
of data loss.

Cartridge inventories are maintained along with routine drive maintenance and cleaning.
Most tape libraries support both proprietary and third-party application and data
management software for handling data backup and restoration as well as hierarchical
storage management tasks.

Look for most or all of the following design features in a DLT tape library:

Multiple drives can double, triple or quadruple the data transfer rates (DTR) offered
by the unit. Theoretically, if a DLT4000 system with one drive, such as Andataco-IPL
Systems Inc.’s EBS4DSA, offers a 1.5-Mbps uncompressed DTR, the same unit with four
drives can provide a 6-Mbps throughput rate.

Generally, use of multiple drives permits concurrent operations. For example, the
library can back up data with one drive while restoring data to a server with another.
This feature is especially important for HSM, which requires continuous migration of data
back and forth between tape and disk.

In a fault-tolerant library, if one drive fails, others will ensure that data backup
and recovery procedures continue without a hitch.

Consistent, modular design features also allow flexibility in packaging of the
units—desktop, tower or rack-mount.

Your system should also come with a variety of device drivers that allow its use with
any operating system.

Transitional Technology Inc.’s Q Series comes with a laser calibration system that
provides self-correction and realignment of the unit should it receive a knock or be
subjected to sharp movement.

Tape autoloaders, also called autostackers, are a step removed from tape libraries in
both price and performance. Autoloaders automatically load tapes sequentially, usually
into a single drive, to provide unattended backup. Their hardware robotics and management
software components are generally less sophisticated than those for tape libraries. This
relative lack of complexity makes tape autoloaders a good choice for network managers
exploring automated network storage for the first time.

Like libraries, autoloaders depend heavily on the SCSI-2 interface, and the range of
sustained data transfer rates is about the same.

The units offer support for Open Systems platforms and commonly used software from
Novastor Corp., Cheyenne, Legato Systems Inc. and others.

A jukebox is a storage device for multiple sets of CD-ROMs, tape cartridges or optical
disks. It could even be said that a tape autoloader is a type of tape jukebox. In this
guide, however, we list only optical jukeboxes—those that hold and manage optical
disks.

Optical jukeboxes hold as few as five and up to 100 or more disks. Like the
platter-spinning Wurlitzer record players of the 1950s, the optical media is moved by a
robotlike arm from a storage slot in a carousel to the drive.

Optical disks are more durable and stable than tape. They are less sensitive to
temperature and humidity extremes than tape and have a shelf life of 30 years or more.
However, the per-megabyte cost of optical storage is generally higher than that for tape
storage.

If you can settle for less overall capacity than with most tape libraries and
autochangers but still require quick access times with durable, though expensive media,
optical jukeboxes are a good bet. You can easily store and manage CAD/CAM drawings,
invoices, images and other documents, and access them in seconds.

Several technologies shape the optical jukebox marketplace. Magneto-optical disk
drives, which use 5.25-inch cartridges, can read and write more than one type of storage
medium.

Use MO jukeboxes for operations such as large document management or computer output to
laser disk applications, or enterprise environments in which large, memory-intensive
document and image files are stored.

MO systems offer the high capacities, reliability and media durability required by
mission-critical applications, according to Peter Way, product marketing manager for
Hewlett-Packard’s Optical Jukebox Business Unit.

Of special note to MO users is the new family of 5.25-inch MO drives and cartridges
introduced by Hewlett-Packard in May 1998. This 5.2G family doubles the capacity of
previous 2.6G optical products and offers significant speed and transfer rate advantages
over older MO product lines. HP is using the systems in its new SureStore line of
jukeboxes. Maxoptic Corp. and Plasmon Data Inc. have followed up with 5.2G cartridges
packed in their jukeboxes.

At the low end of storage requirements, inexpensive CD-ROM jukeboxes that use the
standard CD-ROM read-only drives used in most desktop PCs provide access to relatively
static information.

Jukeboxes using CD-recordable, write-once (CD-R) or CD-rewritable (CD-RW) media are
well suited for recording and archiving data that needn’t be accessed daily.
Rewritable digital video disk (DVD-RW) technology, which has greater storage capacity and
faster speeds than CD-R, is coming on strong.

A relatively new optical technology, near-field recording, is also showing up. By
bringing the recording head very close—4 micro inches—to the recording service,
near hard disk speeds can be attained, but the susceptibility of this technology to dust
and vibration may be a problem.  

Most organizations store information four ways: online on hard drives, nearline in
libraries or jukeboxes, and offline either on removable disks and tapes, or on paper or
microfiche.

Each offers advantages and drawbacks, depending on user requirements. For example, only
about 20 percent of the information on hard drives (online storage) is actively used, and
80 percent is accessed infrequently if at all. And even the largest hard drive can’t
begin to hold the hundreds of gigabytes required by today’s storage environments.

Nearline data includes records that are used only occasionally but must be readily
accessible, for example, in 30 seconds or less. Nearline data storage relies on tape and
optical systems using removable media.

Because access to tape data is sequential, many users associate tape with slow recall
performance. Techniques for advancing the tape to the exact block of information required
make advanced digital linear tape, advanced intelligent tape and 8-mm tape storage systems
superior to optical media when it comes to data throughput speeds and access times.

Although it is more expensive and slower than tape, optical storage with write-once,
read-many capability is a good option for nearline file storage on media that will last
for years.

Offline storage depends on removable media that may be stored in remote locations until
used. Both tape and optical media may be used for this function, though tape is cheaper.

Paper and microfiche continue to be the world’s most common storage media, but
they present storage and access problems for all but the smallest organizations.

Look for an arsenal of innovative tape drives and media with a 100G capacity and
10-Mbps data transfer rates to arrive by the end of this year.

Meanwhile, new formats for older tape technologies are also in development. Take a
quick look at today’s most common tape formats and some new ones that will arrive in
the marketplace by year’s end.

Quarter-inch—named for the width of the tape—is a misnomer. Although some
quarter-inch systems are still around, most of today’s QIC tapes are 0.31 inches
wide. A standard 0.31-inch QIC tape is 400 feet long and holds 2G of uncompressed data.

NS8 is a network-optimized version of the TR4 format. It adds internal hardware
compression and a read-while-write verify technique for 8G of storage per tape. Neither
Travan nor QIC drives are heavy iron—they have relatively small capacities and slow
600-Kbps data transfer rates.

Three digital data storage (DDS) DAT specifications were developed by Hewlett-Packard
and are used today by DAT drive makers.

DDS-1, DDS-2 and DDS-3 all use 4-mm cartridges, but DDS-3 tapes are 125 millimeters
long and offer up to 12G uncompressed data and 24G of compressed data storage using the
2-to-1 ratio DCLZ compression algorithm.

DDS-4, an iteration of DAT just coming out, provides 20G of uncompressed and 40G of
compressed storage.

DLT’s basic design features, including linear recording, stationary read/write
heads and calibrated tape tensioning, provide more robust systems with longer tape and
head lives than competing 8-mm helical scan technologies.

Although three types of DLT drives—DLT2000XT, DLT4000 and DLT7000—are
currently in use, the older and slower DLT2000XT is rapidly being phased out. DLT4000
drives can hold 20G of uncompressed and 40G of compressed data with a sustained native, or
uncompressed, transfer rate of 1.5 Mbps.

Using a modified linear serpentine method of recording called symmetrical phase
recording, DLT7000 drives can hold 35G of uncompressed and 70G of compressed data with a
sustained native transfer rate of 5 Mbps.

They provide 54G of storage capacity using LZ data compression and transfer data at 10
Mbps.

The newest AIT drives by Seagate Technology Inc. offer 70G of storage. It will store up
to 40G of data—compressed using the IDRC algorithm—per cartridge and has a
transfer rate of 3 Mbps native and 6 Mbps compressed.

There are many other new technologies on the horizon. They include linear tape-open
developed by Hewlett-Packard Co., IBM and Seagate in both the Accelis and Ultrium formats.

LTO drives will provide throughput speeds of 10 Mbps to 20 Mbps and capacities of 25G
and 100G capacities.

Automated storage systems attached to network servers are far from the last word in
storage technology.

Next-generation storage area networks will let tape and optical subsystems attach
directly to the network without linking to a server. The SAN acts as an intelligent
switched interconnect for any of these subsystems, becoming, in effect, a giant data
repository. Users can access information anywhere on the SAN, regardless of data origins
or subsystems used.

But backing up mission-critical information without going through storage servers
brings a new set of design challenges to information systems managers.

Because tape systems operate far differently than disk-based systems, it is difficult
to make both work together seamlessly in a heterogeneous environment. Members of the
Storage Networking Industry Association are developing standards that let different
storage technologies operate as peers.

Both Gigabit Ethernet and Fibre Channel standards offer the ability to create
high-speed links between tape and disk subsystems, but industry experts and advocates of
both camps disagree as to which should prevail. Some observers opt for Gigabit Ethernet
because it leverages the huge installed base of Ethernet networks. Others prefer Fibre
Channel because it was designed from the outset for storage purposes. Its proponents call
it the natural successor to SCSI, which is widely used among tape and optical system
builders.

As a first step toward SANs, some independent vendors are working on ways to provide
uniform management of tape drives, tape libraries, optical jukeboxes and media pools in
existing networks.

Microsoft Corp. and IBM Corp. have bought software from HighGround Systems of
Marlborough, Mass., that creates a consistent interface for removable storage management
across Microsoft Windows NT and Unix systems. The agreement is based on the Removable
Storage Management service that HighGround developed with Microsoft for Windows 2000.

IBM plans to use RSM as the focal point for applications that share media, devices and
libraries in a distributed, heterogeneous environment with an eye toward exploiting the
potential of Fibre Channel SANs, according to Bud Broomhead, vice president of business
development for HighGround.

Check out the software at http://www.highground.com.
 

J.B. Miles writes about communications and computers from Carlsbad, Calif.