|
A staggering number is that for every dollar spent
on storage hardware, companies will spend three dollars to manage the
information stored on those devices. Given these realities, it is no wonder
that storage management cost containment - or the cost of ownership of data -
has emerged as a primary IS issue.
Increasingly, IS managers have turned to storage automation technology to
solve their data management problems. Advances in storage hardware and software
have made solutions such as Hierarchical Storage Management (HSM), archiving and
disk grooming automatic functions that are designed to optimize the investment
in storage hardware while at the same time protecting valuable corporate
information.
A key component in any automated storage architecture is a tape library.
These libraries are typically equipped with tape drives capable of storing
dozens of gigabytes of data on a single cartridge. Advances in drive
technology, robotics, and software have so greatly improved access to
information in these devices that tape libraries are increasingly being utilized
to store active user data instead of the traditional backup and archival
applications. Indeed, most storage automation schemes rely on moving
less-frequently accessed files to secondary or tertiary storage - most commonly
in an automated tape library - where the cost of storage is significantly less
than with disk drives.
There is little argument about the cost benefits of automating storage
installations. However, there is a great deal of discussion about how tape
libraries fit into the picture, ranging from which tape drive technology to
which library designs to utilize. When it comes to total cost of ownership
(TCO), the cost benefits of today's scalable library designs stand out.
The return on investment in storage automation hardware and software
technology can come quickly, given the rising expenses involved in managing
information. According to a report from Strategic
Research Corporation, total PC LAN data management costs averaged $206,000
per year, while UNIX data management costs were $147,000 per year. And, fully a
third of the expense was due to managing backup. The findings for the PC LAN
market were even more dramatic, with those managers spending an average of 954
hours per-site per-year on backup administration - defined as media handling and
error handling - making it the most time consuming component of the entire
backup process.
Clearly, the role of storage automation will become increasingly important
as companies move to reduce these data management expenses and implement a
proven cost-saving technology.
While automation is the cornerstone of controlling overwhelming data
management costs, administrators are faced with a series of new issues when
deciding the best approach to automating their storage environments. They must
define an optimal mix of storage technologies using on-line disk and near-line
tape, (and possibly optical disk) to maximize data protection and security while
minimizing costs. Automated tape libraries have become an indispensable tool as
part of this automation solution. But once again, with an almost overwhelming
maze of choices, administrators need to be aware of several factors in making a
tape library buying decision, including capacity, performance, and the
appropriate tape technology - 4mm, 8mm, 3490, Magstar or DLT. And one of the
most critical buying factors is also one of the most often overlooked - is total
cost of ownership.
Quantifying the cost of ownership can be an imprecise process, but in any
analysis there are several key components that must be considered when computing
the cost equation. For tape libraries these include initial price of the
library hardware, media costs, warranty, maintenance, repairs, cost of
upgrading, cost of downtime and the cost savings generated by automating storage
management operations.
Most of these cost components are straightforward and easily understood, but
the last three items - automation cost savings, cost of downtime and the cost of
upgrading - are perhaps the most critical and least discussed factors in the
cost of ownership equation.
REDUCING THE COST OF UPGRADING:
Stringent new library up time requirements, as well as customer demands for
a solution to the cost-of-upgrading problem have inspired some companies to
create a new tape library architecture focused on a scalable, high availability
design. The result is that highly scalable DLT libraries are now available.
Some manufacturers such as Overland offer a radical departure from the
traditional "monolithic" library designs that restrict the user to a
maximum number of drives and cartridges inside a single enclosure. Monolithic
libraries are also exposed with several single failure points that can cripple
the library and eliminate user access to critical data.
The standard in expandable, modular DLT libraries has been available since
early 1996. Users can begin a library implementation with a "base module,"
with one or two DLT drives and an integrated, removable 10-cartridge storage
magazine. These high-performance base modules meet today's network backup needs
with native sustained transfer rates of over 36 GB per hour and native
capacities of up to 350 GB. Then, whenever and wherever storage requirements
increase, the scalable design delivers cost-effective modular expansion that can
be easily added as either capacity and performance requirements increase.
Why force an IS manager to scrap an existing investment in library hardware
and obtain unneeded capabilities simply to upgrade the library? Instead of
purchasing a complete new library as demanded with monolithic designs, scaling
today's high-performance scalable libraries beyond a single base module simply
requires additional modules, providing seamless expansion. Unique pass-through
technologies move tapes at high speed from module to module, allowing any tape
cartridge to be moved to any available drive in the system, or to any storage
slot. And each base module can operate in standalone mode as well, with easy
removal from the system when a 150 to 350 GB library is needed for important
department backup jobs.
The ability to modularly expand the capacity and performance in reasonable
increments is the most cost-effective feature of modular storage. But redundant
components also help meet high availability demands of today's network computing
environments.
Monolithic library architecture limits expandability to a fixed number of
drive and cartridges. Some of these products allow customers to start with a
minimum configuration of one or two drives then add more drives, but always
within the constraints of the single library enclosure. To expand the library
capabilities beyond these limitations, users must purchase another complete
library, even if their current storage needs require only a modest increase
beyond the current single library capabilities.
The scalable design of the modular systems places no such limits on users.
Customers can start as small as a single base module with dual drives then
expand the system in customized, modular increments up to capacities for meeting
changing storage requirements. The flexibility of this scalable model is
advantageous in two key areas: users can expand the library by purchasing only
the capabilities they require, and can easily calculate their cost savings vs.
multiple stands of discontiguous automated devices.
The expansion options offered by some manufacturers of scalable storage
architecture enables users to expand the system based on individual requirements
with capacity modules - containing only DLT cartridges to boost overall library
capacity - or with modules containing additional DLT drives - to enhance total
library throughput and enable more users to retrieve data more efficiently.
The cost savings of this approach can be significant. If a monolithic
library system is fully populated with drives and cartridge slots, the only
available option to expand is to purchase an additional separate library cabinet
with additional robotics, drives, control electronics, and cartridge slots,
hardly a cost-effective purchase if the only need is for additional throughput.
The customer has no other option, however, as these "all-or-nothing"
designs provide only a single, extremely expensive expansion option.
THE PRICE OF DOWNTIME:
It would require another long article to cover all of the issues involved in
determining the cost of downtime. However, it is safe to say that in the new
applications environment for tape libraries, zero downtime is the only
acceptable target, and a modular, fault-tolerant design that provides a level of
protection against failure and down time that is simply unavailable from
monolithic designs.
Estimates of the cost of downtime to an organization span a vast spectrum,
from $1,000 to $100,000 per hour - even $100,000 per minute for real-time
transactions. A Gartner Group report posed the question, "How do you come
up with a number you're sure is accurate?" Here's the short answer: "You
don't. The real cost of downtime is elusive."
To a large extent, the cost of server downtime is tied to the applications
environment, producing much higher costs for transaction processing and
manufacturing environments, but still expensive for any organization. The
Gartner Group also points out that downtime cost computations typically only
figure productivity loss to an organization; they ignore transaction loss, loss
of business, or customer dissatisfaction.
With the new applications for tape libraries requiring continuous access to
information stored on tape, the same zero downtime demands that have
traditionally been placed on network servers are now being applied to automated
tape libraries.
The highly reliable design of high-performance scalable libraries brings a
new level of reliability and fault tolerance to library designs with the goal of
minimizing costly downtime. Quantifying the reliability of tape library designs
is an issue that has been grossly misrepresented in some circles. Specifically,
vendors of monolithic library designs have depended on traditional standalone
reliability analysis in an attempt to discredit the reliability of modular
design. These detractors have typically referred to outdated reliability models
that have proven to be invalid when applied to tape library technology.
According to Strategic Research, "a well-designed library with multiple
drives will achieve a lower failure rate than just independent single drives
because the library presents a consistent and controlled environment to the
drives. Thus, the real world average failure rate for a tape library system is
lower than for a combination of single drives."
That assessment of library technology runs counter to the conventional
wisdom of design which proclaims as more parts are added to a system, the more
likely it is to fail, with all other factors being equal. However, all things
are not equal and, as noted by Strategic Research, this model for reliability
does not apply to library technology.
A multiple-module installation is highly insulated against a catastrophic
failure by having redundant components. Monolithic designs provide no similar
insurance against component failure. Single failure points - such as control
electronics or robotics - leave monolithic libraries much more susceptible to a
catastrophic failure rendering the library inoperable and interrupting user
access to data.
QUANTIFYING AUTOMATION COST SAVINGS:
As demonstrated by the Strategic Research findings, data backup is the
largest piece of the data management pie, and an area where automating the
process with a tape library can pay immediate dividends. In general, the
advantages of an automated tape library technology include:
- Unattended backup of all servers and critical workstation data across
multiple tapes; manual intervention typically needed only once a week.
- Automated media management ensures proper tape rotation and multiple
generations of backup, preventing possible disaster if a tape is damaged.
- User can restore their own data if permitted, without intervention by a
network administrator.
- Using tape for low-cost, near on-line storage is labor intensive unless
this is done with a robotic tape library and automated Hierarchical Storage
Management system.
The advantages of an automated tape library are even more pronounced in a
dual-drive configuration. There are four key points that make multiple drives
almost mandatory for today's network storage environments.
- In many backup applications with shrinking time windows, it is impossible
to complete a backup in the allotted time without dual drives, which can cut
backup times nearly in half.
- Having two drives allows for easy creation of off-site copies either by
mirroring during backup or by producing off-line copies after the backup is
completed.
- When backing up mission critical data, having two drives eliminates an
important potential point of failure.
- With two or more drives in a robotic library, Hierarchical Storage
Management runs more smoothly, with less chance that a user will have to wait in
line for the denigration of a file.
CONCLUSION:
Automated tape libraries are an indispensable component in the battle to
control storage administration costs. Buyers of library technology must take
into account a variety of factors beyond price to make an intelligent buying
decision. The only accurate way to ensure the best product is to gauge the
total cost of ownership, including factors such as the time and administrative
cost savings involved in automating data backup and retrieval processes, the
potential impact of library downtime, and the costs of upgrading, if in fact,
the library can be upgraded.
This analysis makes a strong case for modular scalability. No other type of
library offers the cost savings associated with upgrading the library by virtue
of adding only the specific capabilities demanded by unique customer
environments. For example, the user is not forced to buy additional drives or
robotics - which is the norm with monolithic library designs - when the only
need is for additional capacity. The innovative modular design architecture
also provides unprecedented protection against a catastrophic failure that
impacts library availability as part of the overall network storage environment.
Simply put, scalable is better, and the LibraryXpress product from Overland
Data represents the state-of-the scalable-library-art. ...Overland Data profile | 
