How VTL and de-dupe are driving tape’s slow death

Written By: Peter Williams
Published:
Content Copyright © 2008 Bloor. All Rights Reserved.

Magnetic
tape storage will not disappear any time soon but, increasingly, it is being
driven to the edge of system storage, into being used primarily for deep
archive. There is good reason—and it involves a combination of virtual tape
library (VTL) systems that utilise low cost SATA disks and de-duplication.

It is
many decades since disks (even with very small capacities) opened up an on-line
market by providing random access; this displaced tape in the primary storage
tier. However, since then, even though disk drives have massively increased in
capacity and speed, tape technology has largely kept pace and maintained a
considerable price differential to deter detractors from replacing it for sequential
backup and recovery operations.

However,
the massive and inexorable increase in storage and shrinking backup windows
have together created a bigger management headache when tape is used, for
instance in locating and loading the correct tapes when recovery is needed. When
tapes are physically transported off site the risks and problems are
multiplied; there is the danger of damage or theft in transit, plus potentially
huge delays in retrieving and returning the correct tapes through busy traffic even
before recovery can begin. Not very 21st century, that.

De-duplication
appliances for backup started really gaining popularity in the mid-market about
three years back. Inserting a NAS de-dupe appliance, transparently (and this is
important), into the backup process means an average 20:1 (95%) saving on the
output data storage need with little performance overhead even over a straight
disk-to-disk backup. This alone dramatically altered the economics and ROI
versus tape; a disk backup (de-duped) suddenly cost no more than a tape backup
(not de-dupable)—as well as being far faster (especially for recovery), more
reliable and simpler to manage.

A
further bonus is that off-site replication in de-duped form can be achieved
very fast and reliably over WAN, at very low line cost, so can remove the risks
and time normally incurred if transporting backup tapes between locations; this
has also opened up a low-cost storage disaster recovery (DR) capability. So the
ROI case has become very straightforward for most mid-market companies. We have
therefore seen de-dupe appliance vendors Data Domain, and latterly ExaGrid, in
particular doing very nicely thank you in the mid-market—with tape being squeezed
out altogether in many of their accounts.

On
the other hand, large enterprises have a huge legacy investment in tape for backup
and archive storage. Enter VTL. The backup application thinks it is writing to
tape but VTL redirects it to disk; so initial backup is faster while the data
can be retained ‘near-line’ for much longer—allowing rapid on-line access as
well as fast recovery. The transfer to the physical tape library drives then
happens as a more leisurely background task. Despite this, making an ROI case
with VTL alone may not be easy.

However,
VTL that incorporates de-dupe is a different matter. This combination can use
only disk for backup and add remote replication—leaving only the archive and
remote backup tape libraries untouched. Now the major hardware and backup
software vendors are offering VTL/de-dupe solutions alongside, or in
competition with, specialists such as Sepaton and Data Domain. Even z/OS
mainframes are now catered for with an offering from Data Domain partner Shoden
Data Systems
.

Despite
this, VTL with de-dupe is actually a stop-gap solution (albeit for the medium
term). It avoids an enterprise being faced with disruptive system changes in
order to replace tape with disk for backup and gain the storage capacity savings;
but it also carries an overhead in formatting the output to one of the tape
formats—which is especially inefficient when there is no tape at all at the
back-end of the process (as VTL solutions allow as an option).

So
tape will survive for now, especially for off-line long-term deep archive
storage; but even here management headaches need addressing, including the changing
tape formats and drives causing old data
to need refreshing by being copied to newer format tapes—or maybe disk—on
replacement equipment to remain readable. Moreover, de-dupe can be applied to
the archive and typically achieve a 4:1(80%) data reduction as long as tape is not
used.

I
would be very surprised if the tape storage vendors had a viable answer this
time around. So, while it may be a slow
death, the end for tape for backup and storage is coming nearer with the demise
of tape archive only just beyond the horizon.