Stuck on tape
For large enterprises, tape libraries are still a cost-effective storage option
As disk prices fall, many organizations
are switching from tape
storage to disk-based virtual
tape libraries (VTL) as the preferred
means of long-term data storage. The
Agriculture Department's Food and Nutrition
Services (FNS) in Alexandria, Va., for
example, started using NetApp Virtual
File Manager to back up its servers onto a
set of NetApp FAS200 series devices.
'Since doing disk-to-disk backup,
tapes are hardly being used,'
said Harold Russell, a Wyandotte
NetTel project manager working
at FNS.
Despite predictions of their imminent
demise, however, tape libraries
' which have been around
since Remington Rand built the
Uniservo in 1951 ' still have a place
in the data center.
FNS has two Spectra Logic Spectra
20K tape libraries, each with four drives
and 24 tape slots, which it still uses for
backing up its test environment. And next
summer, FNS will start doing monthly
archives of its e-mail messages to meet the
seven-year retention requirements.
'Knowing the federal government, we will
probably never completely dump tapes, unless
we switch to another durable type of
media like optical disks,' Russell said.
Disk vs. tape'Tape is the most cost-effective storage out
there, and people will continue to leverage
tape products for many years to come,'
said Tom Coughlin of consulting firm
Coughlin and Associates.
Robert Stevenson, managing director of
storage research at TheInfoPro, said tape
usage is particularly strong among large
organizations.
'They are starting to see more regulatory
compliance requirements, so we are not
seeing a reduction in tape growth,' he said.
Smaller organizations, however, are
adopting VTLs as an alternative to tape.
Those that have 15,000 to 20,000 tapes
are weaning themselves
off tape, and many could be completely
off tape in about five years. Those
with less than 15,000 tapes are using VTLs
as a storage consolidation method.
'They like VTLs, not so much because
they have a bottleneck in terms of their
tape drives being busy, but they have different
tape drive technologies,' Stevenson
said. 'They are using VTL to consolidate
onto one single tape format while being
more aggressive on trying to minimize tape
creation.'
Coughlin said one main driver in keeping
tape is not only the equipment and supplies
cost but also the power consumption
differences between disk and tape storage.
'As fuel prices go up and energy costs increase,
tape automation systems can be a
big part of using your resources more effectively
and lowering your total operating
cost,' he said.
Analysts David Reine and Mike Kahn of
the Clipper Group compared costs and energy
consumption for five-year data storage
on Serial Advanced Technology Attachment
disks compared with Linear
Tape-Open Generation 4 (LTO-4) tape
(GCN.com/1191.) The study assumed
an initial storage base of 50T that
grows 50 percent annually and runs daily
differential backups containing 5 percent
of the total data, weekly full backups stored
for 13 weeks and quarterly archives of the
data. The archives were either stored on
the disk array or moved to near-line storage
in a tape library. Reine and
Kahn found that storing the quarterly
archives on the disk array cost
23 times as much as moving them
to tape and that energy costs were
290 times as high.
Disk-to-disk (D2D) backup, they
concluded, 'is not a replacement for
tape; it should be a complement to tape.
The costs associated with a pure D2D scenario
are simply too great for any midsized
or larger business to consider using
this technology to establish a comprehensive
policy to save their entire data store
for any extended length of time.'
New capacityAlthough tape automation systems are
mature product lines, there are still incremental
improvements being made to their
functionality.
The biggest change recently was the introduction
of LTO-4 in 2007. The new format
doubled the capacity of the third generation
of LTO (LTO-3) tapes to 800G
native storage and boosted the uncompressed
transfer rate to 120 megabits/sec.
Sun Microsystems and IBM have already
gone beyond the LTO-4 capacity with new
drives they introduced in July. Sun's new
StorageTek T10000B tape drive has a 1T
native capacity, double that of its predecessor,
and costs $37,000. IBM's TS1130 drive is also 1T and costs $39,500. Both use 4
gigabits/sec Fibre Channel or Fiber Connectivity
connections.
LTO-4 retained the write-once, readmany
functions that came out of LTO-3
tapes to comply with certain data retention
regulations. But for the first time, it
also includes built-in encryption ' Advanced
Encryption Standard 256, approved
by the National Institute of Standards
and Technology and which meets
the requirements of Federal Information
Processing Standard 140-2.
Stevenson said the release of LTO-4 has
been a driver in federal agencies' increased
tape library purchases.
'We are seeing a heavy amount of refresh
rates and moving to LTO-4 drives to minimize
the number of tape drives they have to
license,' he said. 'The encryption feature is
also a big component in the federal sector.'
The other key factor in making tape library
decisions is data deduplication. With
rapidly expanding storage capacity, deduplication
is increasingly important,
whether it is part of the tape library or a
stand-alone product.
Tiered storageIf storage volumes stayed constant, falling
VTL prices would make them a viable
option to replace tape. But with storage
requirements growing at a 50 percent
annual rate, tape still has its place. But
where tape fits into the overall solution is
changing.
'With the push toward data center consolidations,
the influx of servers within
the primary hosting centers with their
own separate backup solutions overwhelmed
the initial infrastructure,' said
John Kozitzki, enterprise backup and recovery
manager for Michigan. 'Backup
windows were continually being missed,
success rates were in the 70 [percent] to
80 percent range, and backup administration
staff was having a hard time
keeping up with configuration and
maintenance.'
Kozitzki set up an enterprise backup
system using both disk and tape. The
state uses Symantec Netbackup running
on 22 assorted Sun servers. Netbackup
writes to EMC Symmetrix disk
storage volumes for disk-to-disk backups,
EMC EDL4400 disk libraries for
virtual tape use ' disk-to-disk or diskto-
disk-to-tape ' and to a Sun StorageTek
SL8500 tape library with 60
drives and 5,000 slots, using 9940B or
T10000 tape.
'We currently configure servers that require
enterprise backup and recovery
services based on the amount of storage
allocated to the server, the retention policy
required, and the recovery point/recovery
time objectives,' Kozitzki said. 'This
information is taken into consideration
when crafting a solution for our agency
partners.'
The tape library is used for direct nearline
storage and off-site archiving on
servers that have retention policies greater
than 30 days. With the new setup, he runs
about 20,000 jobs each week to back up
200T of data and is now achieving a 96
percent to 98 percent success rate.
'Moving to a standardized, consolidated
enterprise solution made sense,' Kozitzki
said. 'It allowed for easier administration,
increase in success rates and made it more
scalable.'