Few organizations today would deny their absolute dependency on
both their computer systems and on the information they store. Exponential data
growth continues, and despite the current economic climate, the growth rate of
storage requirements has not abated. Trends such as increased user-created
content coupled with mounting pressures from regulatory compliance, are causing
us to store more information over longer periods of time. Simply adding more
storage capacity to keep up with this data growth is no longer an acceptable
strategy with organizations facing ever tighter constraints in budget, physical
floor space, power, and management resources. Against this backdrop, storage is
continuing to evolve and diversify to keep pace with data growth. Technologies
such as virtualization, deduplication, and solid state disk are adding to the
range of options available. While much is being made of the potential
opportunities opened up by disk-based solutions for data protection, as opposed
to traditional tape, the question for the majority of businesses is not about
whether to choose disk or tape, but about how to use both technologies to best
effect in modern tiered-storage architecture. Put simply, decisions on storage
should be made in the context of a balanced equation which includes the business
need they meet and the cost involved to maximize the return on asset (ROA) of
each storage device. This whitepaper describes five best practices for backup
and long-term data protection and explores the aspects of disk and tape
solutions that, when combined into a disk-to-disk-to-tape solution, provide the
most robust answer to backup and archival for the data center.
Unstoppable data growth
Organizations of all sizes are grappling with the need to keep
larger and larger volumes of ‘business critical information’, as well as ‘other’
information, for longer and longer periods of time. A recent InformationWeek
Analytics/ Byte and Switch Storage Management Surveyi reported that over 50% of
organizations surveyed were actively managing between 1TB and 99TB’s of data,
capacities that would have been unthinkable just a few years ago. The reasons
behind this extraordinary data growth are many and varied, but are largely
1. Business requirements to hold more detailed information on
products, services and customers that will shape business strategy, drive growth
and increase service levels (business applications and data warehousing).
2. User-driven growth in unstructured content, through the
increasing use of images, video and audio alongside more traditional content.
3. The need to mitigate risk influenced by both internal
policies and compliance with externally imposed regulations, which necessitates
the storing of data for longer periods of time.
At the end of the day, IT Managers are less concerned with the
causes of this phenomenal data growth as to what to do with all the data they
need to store. This is especially taxing in the area of backup and archival
storage, where company data is in effect ‘stockpiled’. The challenge for
organizations becomes balancing out the need to minimize storage requirements
and budgets with the business imperatives of compliance and protecting company
data from the many threats that can result in catastrophic data loss (see figure
Keeping pace: storage evolution
The evolution of storage in terms of capacity, access speed, and
cost has been evident since early computers employed disk drives the size of
washing machines, and tape reels as large as bicycle wheels. Until recently
there were few architectural choices to be made and consequently both disk and
tape systems were deployed in parallel. However, over the last few years
significant developments in disk technology have resulted in substantial
reductions in unit price. Furthermore, technology advances such as storage
virtualization, data deduplication and low-bandwidth data replication bring new
opportunities. By expanding the use of disk-based storage in a data protection
role, one that was formerly purely the domain of tape-based solutions,
enhancements can be made to SLAs through improved data recovery objectives.
The first tape drives were introduced almost sixty years ago,
and since then tape technology has also continued to evolve in parallel with
advances in disk technology. Storing up to 1.6TBs of compressed data, the latest
LTO-4 Ultrium tape drives hold 8 times the capacity of the first LTO tape drive
launched in 2000. In addition, long before it was offered on disk, tape
technology offers Write Once Read Many (WORM) capability as well as
Despite the evolution in tape, as disk-based backup solutions
have gained in momentum attempts have been made to promote the idea that tape is
“dead”, or at best a legacy solution. However, published research supports a
very different picture, with only a minority of organizations adopting a
disk-only approach to the storage of data. The InfoStor Annual Reader Survey
published in March 2009, indicates that tape is still a valuable component of
the storage hierarchy, with three-quarters of those surveyed stating that they
used tape either exclusively or as a primary element of their backup and
recovery infrastructure (as shown in figure 2). Additional research by Fleishman
Hillard reported that over two-thirds of respondents currently using disk-only
backup were looking to add tape back into their storage infrastructure, while
40% of managers in tape only environments are forecasting the increased use of
It would seem that tape technology continues to offer an optimum
balance of price and performance for storing data with lower access
requirements. Additionally, companies looking to store vast amounts of streaming
data (e.g. databases, digital imaging, multi-media etc.), can also benefit from
tape's high performance backup speeds. With native transfer rates of up to
120MB/s, the LTO-4 tape drives stream up to 864GB of 2:1 compressed data per
hour, easily equaling or surpassing the backup speed of the majority of today’s
disk drives in data streaming applications.
The considerable demand for tape is further evidenced by the
fact that tape drives and tape libraries are offered and promoted as part of the
storage ecosystem by all major system manufacturers. In fact, the branded tape
drive market was worth $3.47bn in 2008 with shipments of over one million drives
across all market sectors. Furthermore, major manufacturers including HP, IBM,
Quantum and others, continue to invest in further development of successful tape
technologies such as LTO and DAT.
The key to understanding the best use of disk and/or tape for
data protection lies in recognizing their place in the storage hierarchy. The
following best practices for data backup and longer-term data retention provide
a guide to choosing storage strategies and technologies to optimize your
organizations data protection.
Data retention for compliance and e-discovery (deep archiving)
- separate from archival of more unstructured, infrequently used data is the
need to retain information for compliance and business governance reasons. There
are now over 10,000 regulations in place throughout the world that require
records to be held for certain periods of time., for example; Sarbanes Oxley
(US), European Union Data Protection Act (Europe), Electronic Ledger Storage Law
(Japan) AIPA (Italy), HIPPA (USA) to name but a few. Companies that do not
comply face hefty financial penalties, bad PR and even imprisonment for key
board members. The same considerations as those for choosing active archiving
solutions may be used for longer term data retention and compliance.
Data backup and restore - ensuring the timely restoration of
data following a user error, system failure or other occurrence. Critical
decisions to determining which storage technology to choose include:
• RTO and RPO
• Physical data storage and security, especially for data in
Business continuity and disaster recovery - in the event of
a significant system failure due to malicious act or natural disaster, what
provision needs to be in place to get the business back up and running?
• RPO and RTO
• Determine off-site storage locations, it is vital to remove
data from the same threats posed to the data center using different tiers of
storage to meet these various backup and data retention requirement can
significantly enhance storage efficiency and reduce costs.
Best practice 3: Mitigate risk
Don’t put your data at risk - there are a number of best
practices concerned with minimizing the risk to your data from threats
previously described in this paper.
Have multiple levels of protection - it is advisable to hold
at least 3 copies of data in different locations, including one of these stored
in a remote region for disaster recovery purposes in the case of fire, flood,
earthquake or criminal attack. This latter event impacted a UK flight simulator
site as described in the left column at the top of page six.
It should also be noted that having data copies held on
different forms of storage media, for example a mixture of disk and tape, can
also avoid system or media process issues.
Another method in off-site data protection is the use of low
bandwidth replication (LBR) - particularly relevant with virtual tape libraries.
This technology offers the benefits of automating the transfer of data offsite,
using deduplication to transmit only changed data and hence reducing the amount
of bandwidth required to effectively replicate virtual tape cartridges between
sites. However, LBR does not protect against viruses, sabotage and human/system
error. A corruption at the source site could be replicated to the target site.
For this reason multiple iterations of the backup data is recommended along with
an original copy of the data to physical tape. Older iterations should be moved
to tape to free up the more expensive disk space.
Keep at least one copy offline (system isolation) - avoids
intentional or unintentional corruption. Some IT departments are looking at disk
only solutions for backup, restore, archive and disaster recovery. This may
include replicating data from disk to disk which can help protect it, however if
there is a virus, sabotage or system error, then the replicated data may also be
corrupted. The data was not isolated from the system. On the other hand, a
backup copy held on tape, offline away from the system, protects against
intentional or unintential corruption, physical disk failure and provides a
logically “clean” copy of the data. (See sidebar story “illustrating the need
for system isolation”)
The low cost per Gbyte of tape storage (approximately 4 cents
per Gbyte for an LTO-4 cartridge), means that multiple iterations of the data
can be inexpensively stored on tape in the event a point in time copy is
required. Removable disk solutions may also be implemented for removing data
offsite; however these are available in much lower capacity points than many
tapes and considerably more expensive in terms of cost per Gbyte of storage
(approximately $1.45 per Gbyte for a removable disk cartridge).
Protect data at rest and in transit - every month we hear of
new horror stories about loss or theft of sensitive data. With mounting
legislation around the world, protecting data at rest is becoming a new business
imperative and as a result data encryption is a hot topic. Data encryption can
be performed using software, dedicated encryption devices and also using
device-based encryption. A number of tape drives feature hardware based data
encryption that delivers a low-cost, scalable solution that does not impact on
Implement the right technology mix - hardware and software
failure is an inescapable fact of life, and in reality both disk drives and tape
drives do fail on occasion. However, neither technology is inherently
A number of myths have grown up around tape reliability. In
reality, the vast majority of failures that created this myth have been
attributed to other sources: human administration errors, software or system
issues. In the rare event of a tape error, the data can typically be recovered.
In the event of a disk error, the data is typically unrecoverable. Today’s tape
drives include technology features that help overcome these issues to provide
outstanding data integrity and reliability. Let’s take LTO technology as an
example. Advances in the coating of tape film, read-after-write data
verification and powerful error correction codes provide confidence in the
integrity of data stored on tape. These robust tape cartridges are coupled with
drive technology that features simpler tape paths and servo tracking systems to
promote error-free tape handling. Consequently, LTO drives are specified with an
impressive mean-time-between-failure rate (MTBF) of 250,000 hours at 100% duty
cycle, that’s 700% more than the MTBF of tape technologies created a decade ago.
To enhance reliability, disk drive manufacturers developed RAID,
and some would point to this to suggest that disk is more reliable than tape.
However, RAID is more about “uptime” than it is about data protection. If a
drive fails, you simply keep running from the other drive(s) in the array, if a
spare drive is available. However, if you have a virus that destroys data, if
files are accidentally, or deliberately erased, should you suffer from a
catastrophic OS failure that requires reformatting and reinstalling, then RAID
does not offer protection. Only an offline backup is going to really protect you
from losing data.
When it comes to minimizing risk, not one but all of the above
factors should be taken into account. It's not just about the reliability of the
technology you choose or the security of your location, but about the overall
strategy for holding multiple copies on different media, online and offline,
secured and protected.
Best practice 4: Consider TCO
More than acquisition cost - In meeting today’s SLAs. IT
Managers need to consider all aspects of the value of a solution, not only with
regard to backup window and recovery times, but also the total ongoing cost of
delivering the service.
While disk invariably provides for faster data access, the
introduction of lower cost disk such as SATA drives has caused some people to
believe that disk is a cheaper option than tape. However, the acquisition cost
of a data protection solution is only a small part of the total cost of
ownership and therefore any evaluation of storage technologies for backup and
longer term data retention should include the following factors:
Energy costs - the fact remains that tape systems will
always require less electricity than disk systems with similar capacity. In
general, disks must be kept spinning whether data is being accessed or not. Heat
emission and the need for cooling add to the energy requirements for disk
systems. With tape on the other hand, energy is required to write the data onto
the tape media and from that point on it sits on a shelf, or in a tape library
slot, where it requires no further energy apart from the occasional instance
when the data may need to be retrieved. In a data archiving TCO study, The
Clipper Group examined the total cost of ownership over a five year period for
the long-term storage of data in a tiered storage archiving environment. The
analysis compared a disk-to-disk solution (D2D) to a solution consisting of a
mixture of disk and tape (D2D2T). After factoring in acquisition costs of
equipment, media, electricity costs and data center floor space, The Clipper
Group found that the total cost of archiving solution based on SATA disk was
about 23 times more expensive than an LTO-4 tape library archiving solution.
When comparing the energy costs for the competing approaches, the energy costs
for the disk only solution jumped to about 290 times that of the tape solution.
As previously mentioned, data deduplication technologies for
disk can allow a significant reduction in data backup volumes over a period of
time that allows organizations to store more backups to disk, or store backups
on disk for longer periods of time without requiring additional disk purchases.
However, when the same Clipper Group Study described above considered the use of
data deduplication with disk, the TCO was still estimated to be 5 times more
costly than the tape solution. Furthermore, best practice would suggest that
having at least one full copy of critical data in a ‘non-deduped’ state is
advisable in case the disk or data deduplication system pointers become corrupt.
Tape provides a low cost, portable solution that eliminates the requirement for
a deduplication appliance at the disaster recovery site.
Physical storage space - with floor space in computer rooms
and offices at a premium, the high capacity and small footprint of tape provides
a cost advantage that's often overlooked, particularly in longer-term archival
and offsite storage scenarios.
Management - the people costs to manage a storage solution
should also be considered within the TCO. With the advent of advanced
backup/archive management software and physical or virtual automated libraries,
the resource costs required for managing and maintaining disk or tape based data
protection solutions are roughly the same.
Scalability - as data volumes grow, scalability of storage
becomes an important factor. How much does it cost to expand your storage
system, will it add to management overheads and complexity? In particular,
solutions such as Tape Libraries are infinitely capacity scalable; just add in
another cartridge within the library.
Calculating the TCO of different technologies for your backup
and archival scenarios should also include, hardware maintenance, software
acquisition, software support and maintenance after warranty and a look at the
need for management software and disk controllers, in addition to the cost of
data migration. Typically migration to a new generation of disk system needs to
occur every 3 to 5 years, whereas tape technology has a longer cycle of 7 years
or more. Generally the cost of migration effort with tape is much less for tape
than for disk. All of these factors have an impact on the true cost of your
Best practice 5: Ensure you can restore
This is the bottom line - when all is said and done, the
proof point of any data protection solution is in the ability to dependably
restore files, or entire systems, when the need arises. At this point, you don't
want to discover that something has gone wrong. It is therefore important to
consider the following factors:
Regular checks - with all archived data, regardless of
technology used for archive storage, frequent testing of restore capability is
recommended, if not essential.
Shelf life - if you are planning on keeping data for a
number of years in deep archival for governance or compliance reasons, then you
need to ensure that the storage medium selected has sufficient expected
shelf-life. In general, tape offers between 4 and 6 times the life expectancy of
disk, with media manufacturers specifying up to 15 years for DAT and up to 30
years for LTO tape media.
Efficient restores - while the random access of disk
supports fast file access for restore, for batch restores of large amounts of
data, tape may prove equally efficient. Powerful indexing through software can
also facilitate individual file restore from tape.
Disk may be necessary, but is not sufficient - new disk
technologies such as disk mirroring and replication have helped to expand the
use of disk into a primary backup medium. However, should something happen to
the original data due to error, system failure or virus attack, data loss may be
propagated through mirroring or replication to the backup copy. Used as a
complementary technology in a disk-to-disk-to-tape solution, tape provides a
secure, offline copy of the data and the ultimate recovery mechanism.
Recommendations for a balanced approach
In the data center, when it comes to disk and tape it is not a
question of ‘either-or’. The real discussion should focus on ‘where’ and ‘how’
to best use the available storage technologies as part of a comprehensive,
tiered storage architecture while meeting the needs of RPO, RTO, budget and the
best practices listed above.
For mission critical data that requires the fastest RTO and RPO,
primary disk mirroring may be necessary. Backups should be made to tape storage
for offline, offsite data protection and long term archive.
For daily short term backups users can implement a disk to disk
solution such as a virtual tape library (VTL) to allow for fast recall of data.
The data should be copied to tape for off-line, offsite long term data
protection and to lower storage costs. As the data becomes infrequently accessed
on disk it can then be removed to free up the disk space for new data. Many disk
to disk solutions offer data reduction technologies such as data deduplication
to help reduce disk storage requirements.
These technologies are briefly described below:
• Primary Disk Mirroring - a method in which mission critical
data is written to two duplicate disks simultaneously. If one of the disks were
to fail, the system can instantly switch to the other disk without any loss of
data or service.
• Virtual Tape Libraries (VTLs) or Disk to Disk - backing up
from primary disk storage to a secondary backup disk. A common method is a VTL
where data is backed up to a disk that emulates many features of a tape
drive/library, but with the random access recall speed advantage of disk.
• Standalone Tape Drives or Automated Tape Libraries - tape
based backup and restore for backup, archival and data protection.
• Disk to disk to tape solutions (D2D2T) - starts out backing up
to disk, then offloading data to tape for longer term data retention.
In summary, primary disk mirroring and high-performance
disk-based backup in a disk-to-disk or VTL solution will be needed in
environments requiring the most stringent RPO and RTO. Tape will complement
these D2D solutions; migrating data from disk backup to tape backup provides
secure, offline data protection and data archival to mitigate risk. This
Disk-to-diskto- tape (D2D2T) architecture combines the data access speeds of
disk with the low TCO of tape for best effect. Applications that do not require
stringent RPO’s and RTO’s can be backed up directly from primary disk to tape.
In addition to low-cost backup, tape is ideal for active archival of data such
as human resources records, emails, or last year’s patient’s records, for
example, with low access frequency. With a substantially longer shelf life, tape
is also the ideal medium for archiving of compliance data. It is also used as
the ultimate defense against collapse in disaster recovery solutions.
The roles of disk and tape have evolved in the data center and
are complementary in fulfilling best practice strategies. In many environments,
tape is no longer the primary backup target for mission-critical applications
that have stringent RPO and RTO objectives. Yet, it continues to maintain its
role as the primary backup target for applications with less stringent
requirements and for data protection and TCO. Tape has a prominent role in
disk-to- disk-to-tape strategies. It is a cost effective media for migrating
older backups from more expensive disk-based targets. Tape should also be viewed
as the low-cost, dependable, data protection media to store archived data for
long periods of time.