the medium is considerably less expensive than disk or flash storage. Part of the reason is that, unlike disks, one tape machine can accommodate an unlimited number of tape drives or cartridges. An analysis conducted by BackupWorks.com indicates that equivalent levels of backup for tape versus disk results in about 4x cost savings for devices.

There are other cost benefits as well. These include a more than 2x savings in operational costs, and upward of 10x savings in total power and cooling costs. Today, a robotic tape library can contain upward of 278 petabytes of data; the same data stored on CDs would require almost 400 million discs.

Tape also delivers efficiency advantages. All devices, when they write bits to storage, produce an “unrecoverable bit error,” which occurs because the device writes a “ 1” instead of a “0” or vice versa. Error-correction methods do not make the problem completely go away. The result for a commonly used tape format such as an LT0-7/8 is a bit error rate of 1:1018, which is approximately one error for every 1. 25 exabytes (EB). Other enterprise-class and consumer drives perform at an error rate of about 1:1016, which translates to an error every 125 terabytes (TB). Although error-correction codes for storage technologies have improved over the years, tape is about 100 times more accurate than the best hard drives, and about 10 times better than the latest solid-state devices (SSDs).

“The challenge with any storage technology is to reduce error codes,” Heckel says. In a practical sense, this means that tape systems are more dependable than other technology solutions. The greater the unrecoverable bit error rate, the greater the risk of loss of data, along with other errors and problems, including a system seeing two bad drives simultaneously.

Yet, there’s still another consideration. Modern tape cartridges fail at a rate about five orders of magnitude less frequently than hard drives—and tapes in storage require no moving or mechanical device.

Finally, tape offers the added ap-
peal of creating an air-gapped environ-
ment when they are not in use. This
makes a tape library highly secure, as
long as it is kept physically protected.

Beyond Tape

Tape is not ideal for every situation. Recovery from a tape backup can be slow and somewhat cumbersome. Finding specific files can prove vexing. If incorrectly stored, tapes can succumb to environmental damage or become demagnetized.

There’s also a bigger problem with all current storage technologies. A typical hard drive will operate only about three to five years before failing. Portable disk storage technologies such as CDs and DVDs generally hold data for 10 to 25 years, while flash storage—which includes drives, cards, and SSDs—degrades with use, rather than with age. This means that the more a user writes and rewrites to the device, essentially using it for its intended purpose, the greater the risk of failure. Future improvements in hard drives or flash technology are likely to produce only marginal performance gains, Heckel points out. However, tape, stored under ideal conditions, can last 30 to 50 years—and perhaps even longer. Although none of these technologies can compare to the lifes-pan of paper stored under ideal conditions (about 500 years), tape emerges as a clear winner.

Yet there’s still another long-term challenge: many existing storage technologies are butting up against physical and logical limits. It is increasingly difficult to add speed and capacity through more heads, platters, or microchips. A handful of technologies may help boost the power and scale of hard drives, for example. These include heat-assisted magnetic recording (HAMR) and microwave-assisted magnetic recording (MAMR). Both of

There are a number
of technical and
practical reasons
why tape has refused
to fade into history.

ACM
Member
News

PURSUING MULTI-PARTY COMPUTATION

Tal Rabin, a
Distinguished
Research Staff
Member and
manager of
Cryptographic
Research at
Yorktown, NY, was born in
Massachusetts and grew up in
Jerusalem, Israel.

Rabin earned bachelor of science, master of science, and doctorate of science degrees from the Hebrew University of Jerusalem, all in computer science. After obtaining her Ph.D., Rabin served as a postdoctoral fellow at the Massachusetts Institute of Technology for several years, before joining IBM’s T. J. Watson cryptography group.

Her research interests are in
various areas of cryptography,
with a primary focus on multi-
party computation, a way for
two or more parties to together
compute a function over their
inputs, while keeping the
individual inputs private.
Rabin is currently engaged is
designing a distributed protocol
that would enable people to
submit a sexual harassment
complaint, which would then
identify other potential victims
of that specific harasser. Rabin
explains that victims might be
more likely to move forward
with such a complaint if they
knew a group of people would
also be involved, so they would
not have to proceed alone.
“This is the beginning
of the road for having actual
applied multiparty computation
algorithms,” Rabin says, adding
that the next few years will be
exciting as these applications
come to fruition.

Rabin is also interested in helping women advance in computer science by getting more women involved at the undergraduate and graduate levels, as well as in the workplace. She acknowledges that she does not have the answers as to how this should be done, but she hopes it will be a community effort.

—John Delaney