Tape backups have proven to be remarkably
resilient as a storage technology. According to research from
the Enterprise Strategy Group (ESG), less than a third of the
respondents backed up to tape only, less than one quarter backed
up to disk only and about 50% used a mix of disk and tape.
"We've all known that tape backup isn't our best solution," says
Heidi Biggar, analyst at ESG in Milford, Mass. "But as users
bring disk into their environment, they're not throwing away
their tape investment." Consequently, storage professionals must
understand tape technology and its role in the enterprise. This
article highlights the basics of tape backup, outlines the main
implementation considerations, explores the experiences of
several real users and looks ahead to the future of this
long-standing storage technology.
Understanding tape backup technology
Magnetic tape remains a quintessential storage technology. Early
on, a long strip of thin plastic coated with magnetic recording
media was simply spooled onto reels. Later, the reels were packaged
into cassettes or cartridges to ease loading/unloading, simplify
storage, protect the media from accidental damage and improve
reliability by blocking dust. Once a cartridge is inserted into a
tape drive mechanism, the media is passed at a constant speed
across a stationary read/write head. These basic principles have
gone virtually unchanged in over 50 years. As time went on, tape
drives diversified into unique architectures, such as Travan, DDS,
DLT and more recently,
LTO. These tape backup architectures are not
interchangeable, so tape systems must be selected for their
technological longevity and media costs. You can learn more
details about tape technology in our
All-In-One Backup Research Guide.
Tape backup technology proved appealing for several reasons. The
cost per gigabyte is still very low compared to disk. An LTO-3
cartridge with 400 GB of native capacity costs less than $100
(about 25 cents per gigabyte) -- half that with 2 to 1 compression
enabled. Tape backup also offers "endless" storage capacity. Where
a disk storage system has a fixed maximum capacity, tape storage
can be expanded simply by using more cartridges. "You can see the
leverage you'd get with 10 cartridges [in a single tape drive]
versus 10 disk drives," says Robert Abraham, analyst at Freeman
Reports in Ojai, Calif. Further, tape offers a "write once" nature
that is ideal for long-term archival storage.
However, tape backup suffers from some disadvantages that have
been exacerbated by the ready availability of disk systems. The
principle problem is performance. Even the 160 megabytes per second
(MBps) compressed transfer rates in LTO-3 tapes pale in comparison
to disk systems. "The [tape] access times are very slow -- 1,000
times slower," Abraham says, noting that it's much more difficult
and costly to retrieve data from tape. This wasn't such an issue
years ago, but with corporate data volumes increasing from 50% to
100% each year, tape backup windows and recovery time objectives
(RTO) have simply become prohibitive for many organizations,
especially businesses that need to operate 24/7. The cost of disk
is still seen as an advantage for tape, but disk costs are falling
quickly and technologies, like data deduplication, are dramatically
improving disk efficiency, allowing data to remain on disk for
longer than ever before. See our article
Data deduplication explained for more
details and user studies on deduplication technology.
Today's tape backup technology and performance
While tape technology has changed very little, storage and
recovery needs have evolved dramatically -- yet users are reluctant
to discard their existing tape backup systems. This has changed the
role of tape from a primary backup vehicle to long-term disaster
protection. "Tape is [now] the doomsday tier," Biggar says.
There have been numerous tape backup technologies implemented in
the enterprise, but the objective has always been the same --
achieve maximum capacity and performance with minimum cost. The LTO
family has gained tremendous traction over the last few years,
bolstered by its open nature and support from industry giants like
IBM, Hewlett-Packard Co. (HP), Quantum Corp. and Tandberg Data.
Today, LTO-3 has emerged as one of the most important tape formats,
offering 800 GB of compressed storage per cartridge and compressed
data rates of 160 MBps. "Right now, it's all about LTO-3. It's very
popular, almost ubiquitous. It has displaced DLT in the enterprise
as the primary compact/small tape device," Abraham says. LTO-4
provides 1.6 TB of compressed storage per tape and should support
compressed data transfers to 240 MBps. LTO-4 is expected around
mid-2007 and will probably be available from all four vendors by
the end of 2007. Encryption will be a common option in LTO-4
drives, allowing users to protect tape data without the need of
encryption features in backup software.
Abraham segregates today's tape backup systems into three
categories. A low-end tape system typically includes one tape drive
and some form of basic autoloader that is normally limited to 10
cartridges or less. A midsized tape library is common in the
enterprise, using four tape drives with autoloading space for up to
several thousand tape cartridges. For example, a four-drive system
with an 80 cartridge library is common, with prices in the $30,000
to $80,000 range. For the large enterprise, Abraham suggests a much
larger library with up to 64 tape drives and the robotics necessary
to swap tapes from a library of 5,000 cartridges or more.
Tape reliability and life expectancy
Tape has garnered a reputation for poor reliability, but this is
a notion that experts are quick to debunk. "It's not unusual to see
data reliability specified at one million hours right now," Abraham
says, noting that some vendors call this "power-on hours" with 25%
to 50% tape motion. Some tape cartridges have been certified to
retain data for upwards of 30 years -- even up to 50 years on
premium media.
So, why do tape backups seem to suffer from such poor
recoverability? Experts say that recoverability problems are not in
writing to the tape. "The tape drive records information when it's
told to do a backup or an archive, and it verifies that the
information is recorded properly at the same time its recording --
disk drives don't do that," Abraham says. Both Abraham and Biggar
suggest that the problem isn't in the tape cartridge or drive, but
rather in other tangible errors or oversights, such as operator
errors, backup software bugs, changed/incompatible backup software
versions, lost or mislabled tapes and poor tape storage
conditions.
Implementing tape backup platforms
A tape system typically consists of four elements: the tape
drive or library, the backup software, the backup server, which
runs the backup software, and the tape cartridge(s). Although there
is considerable flexibility in the choice of these four elements,
experts note that the single greatest concern is backup server
performance. It's crucial that the server be faster then the tape
drive so that a steady stream of data is always available -- if the
data buffer empties, the tape drive must stop and reposition the
tape before more data can be written. This results in a repetitive
back and forth motion of the tape called "shoeshining," which puts
excessive wear on the tape and shortens its working life. Many
LTO-3 drives can track to slower servers, but this is a waste of
potential performance. Experts urge full-speed operation to get the
most possible value from tape backup investments.
When deploying a tape backup system, it's important to plan for
future growth. Since forklift upgrades can be extremely expensive
and disruptive to the backup process, experts suggest selecting a
tape platform that can accommodate growth with relatively
inexpensive upgrades. For example, a user needing 40 cartridges
might select a tape library with a 120 slot capacity, then
incrementally add to the remaining 80 slots over time as needs
change. Similarly, a user might select a tape library with bays for
six tape drives, but only install two tape drives to start. The
remaining four drives can be added over time to boost recording
performance. Another way to increase both capacity and performance
later on might be to upgrade the tape drive and cartridge types.
"For example if you're using LTO-3 right now, at some point in the
future you simply upgrade everything to LTO-4," Abraham says. "You
don't have to replace the robotics."
Cartridge handling and management is another area where experts
suggest added attention. With an increasing emphasis on security,
data encryption should be used wherever possible. While encryption
is primarily a software-based feature at the moment, Abraham notes
that the LTO-4 standard will support native encryption at the drive
itself, and DLT vendors will also be adding encryption on their
high-end offerings. Tape backup cartridges can become difficult to
manage, especially as the library size scales. RFID technology is
also being introduced from vendors, like Imation Corp. to track and
locate tape cartridges, or to identify tapes quickly within a
library. Secure offsite storage should not be ignored.
The impact of tape backup
Experts and users all stress that tape backup does exactly the
same job that it always did -- economically protect enterprise data
over the long term.
For home furnishing retailer R.C. Willey Home Furnishings in
Salt Lake City, the need to maintain accounting, inventory and
other business data for auditors has been a challenge not met with
disk storage. The company currently supports about 10 TB of data,
though only about 2 TB is considered "production data" located on
an EMC disk subsystem. The production data is backed up daily, with
full tape backups performed on a weekly basis. "We don't believe in
incremental backups," says Richard Sheridan, IT manager at R.C.
Willey.
Sheridan uses an aggressive array of tape storage to address
backup needs. One SpectraLogic Corp. Python T950 and two Python
T120 tape libraries handle the local tape backups. Three
SpectraLogic Python T50s support remote backups at remote
distribution centers. To make the backup process as unobtrusive as
possible, backups are staged to EMC business continuance volumes
(BCV) and staggered to four SAIT drives in the T950. "We're taking
backups throughout the day," Sheridan says. "At any given time, we
could have three or four backups happening." The daily tape backups
are also performed twice, so one copy is kept onsite, while the
duplicate tape is carried to a fire safe nearby location. With four
hours to backup 500 GB, Sheridan estimates the total daily backup
window runs between eight and 10 hours. Security concerns are met
by encrypting the backup data at the server.
This daily backup strategy yields a recovery point objective
(RPO) of 24 hours, though the recovery time objective (RTO) is
considerably shorter. "I can take my whole production environment
and put it back on a BCV in four hours." Shreidan says, noting that
tapes are available back to January 2000. Recovery testing is an
ongoing process, and Sheridan typically restores one tape per week
to locate lost files. Recovery is tested more formally on a monthly
basis by restoring to the company's disaster recovery site 45 miles
to the north.
The SAIT drives were a big improvement over older DLT
technology, but Sheridan is eagerly anticipating future tape drive
upgrades. "We're staying on tape with LTO-4," he says. "With SAIT,
it's probably a dead end." The ultimate goal is capacity, and 800
GB (compressed) per LTO-4 tape makes the economics and management
needs of tape media far more appealing.
Compliance issues
Long-term backups are also important in the gaming industry, and
American Casino and Entertainment Properties (ACEP) must be able to
furnish gaming data demanded by regulators. "All gaming-pertinent
data has to be stored for seven years," says Aaron Perez, ACEP IT
network manager. "We have huge safes full of tapes." Complicating
matters even further, recent company acquisitions have caused
ACEP's data to rocket from 8 TB to 20 TB in the course of just a
year -- nightly tape backups were just not adequate.
Today, 20 TB backup jobs are first staged to Dell Inc. PowerEdge
1950 servers and onto two Dell PowerVault MD1000 disk platforms
before being sent to an Overland Storage Inc. NEO 8000 tape library
with six LTO-3 tape drives, each running about 50% of the time.
"It's not all backed up at the same time," Perez says. "We have
three or four backup servers running concurrent backups and
dragging everything to tape." There are currently 200 tapes in the
NEO library, but Perez rotates about 10 to15 of those tapes to the
library each month.
The ability to take tape backups off site is a key element of
the ACEP disaster recovery strategy. Perez relies on a NEO 4100
tape system located in another casino for emergency recovery.
Efficiency is also a major aspect of tape economy, and Perez points
out that the move to 400 GB (uncompressed) LTO-3 from older 50 GB
AIT-2 tapes lets ACEP save far more data per tape in less time.
RTOs vary depending on the system that needs to be restored, but
Perez notes a documented RTO of eight hours for "minimum
usability." Slot systems with 80 GB to 100 GB of data may take far
less time to recover, while email systems with 400 GB to 500 GB
will take considerably longer. Recovery testing, however, is
usually an ongoing process. "We test regularly when we have to
restore a lost user file -- that happens three or four times each
week," Perez says, adding that formal biannual testing also helps
to verify the recovery process.
Looking toward the future, Perez expects to expand the NEO 8000
with additional tape slots, drives and servers, though a migration
to LTO-4 isn't in the immediate roadmap. "I'd consider it, if the
cost is right. But then you'd have to do a full overhaul [of the
tape system and process]." The big dilemma for ACEP now is tape
encryption, though Perez isn't yet satisfied with the capabilities
of encryption in backup software. "Some software manufacturers
aren't stepping up to the plate."
The future of tape backup
The demise of tape backup has been predicted for years,
especially now that disk technologies, like continuous data
protection (CDP), snapshots, replication and virtual tape libraries
(BTL) have emerged in the data center. Still, experts are quick to
emphasize that many disk technologies are basically used for
buffering tape -- removing the backup burden from primary disk.
Experts also underscore the benefits of mobility and long-term
retention, noting that tape's reputation for questionable
reliability is undeserved. "It's very easy to verify recorded data
as you write it, so the reliability is never affected as it might
be with disk," Abraham says.
The current market is stable, due in part to the current
investments made in tape technology, and experts see LTO vendors
with sales growth in the high single digits. "The total market size
on average seems to be at a steady state or a slight increase,"
Abraham says. "That overall trend will probably continue for about
five years." Experts say that the future of tape backup lies with
the open, performance-oriented technologies, such as LTO. The
eventual release of LTO-4 should support an incremental increase in
storage capacity and speed, and the inclusion of encryption in the
drive itself should attract attention from storage organizations
concerned about security and compliance issues. The entire LTO
roadmap promises growth to 6.4 GB of compressed storage with
transfer rates to 540 MBps. "LTO is not on its deathbed," Biggar
says.