|
Data
Storage
So-Called
‘Small Disasters’ Can Equal Big Trouble
By RON LEVINE
Forget about storms,
earthquakes, and fiery explosions that are the images that first come
to mind when we hear the word “disaster.” The chances of one
of these catastrophic events actually hitting and destroying your individual
IT operation are infinitesimal.
What should be worrisome are the far more common minor mishaps that
occur every day and yet are capable of bringing businesses to their
knees. Everyday occurrences like coffee spills, power surges, communication
outages, or an employee mistakenly deleting or overwriting a key file
trigger most IT disasters.
Why? Because it’s these seemingly small mundane events that attack
a business’ most valuable asset – its stored data.
Headline-generating disasters that make the 6 o’clock news can
wipe out complete information systems and network environments, but
they seldom occur at any individual company and rarely strike the same
site twice. On the other hand, recurring small disasters (usually never
heard about outside the company’s IT department) continually cause
havoc and financial losses to unprepared IT dependent businesses. These
more common types of IT disasters usually interrupt processing operations
because they block access to, cause corruption of, or destroy critical
data.
Data Disaster
Contingency Planning
Since organizations today depend so heavily on access to and availability
of stored information, it is imperative this information be fully protected
and easily recoverable, as anything that affects these critical files
can quickly change a small incident into a major company-wide disaster.
A well prepared IT administration (one with a data disaster contingency
plan in place and tested) can fully protect itself from damages caused
by most types of data interruption or destroying mishaps. Data replication
techniques provide the cornerstone of this protection.
High-risk incidences such as the introduction of a virus destroying
or corrupting files, a crashed disk on a server, a malfunctioning or
blocked SAN or NAS, a storage unit upgrade which goes amuck, a communications
glitch causing lost customer transaction data, or an employee’s
careless mistake – all can be guarded against with a solid data
replication solution.
The typical everyday IT emergencies don’t make very suspenseful
movie plots, but these seemingly mundane incidences have the potential
of placing an enterprise into Chapter 11. A data replication component
is an essential subset of the company’s overall disaster contingency
plan. And if history is a guide, chances are it will be this component
of the overall plan that is most likely triggered in response to an
event, thus averting a potential disaster.
Basically, data replication is the process of creating a duplicate copy
of the data at a secondary location. The copied data then can be made
available to an application in the event of a loss of the primary data
source. The choices include local data replication, for example RAID
1 mirroring or distance-based data replication, a replication technology
that can move the data over distance to a remote site. While both provide
instant online availability of the copied data, only distance-based
data replication guards against a catastrophic loss of the entire data
center.
After deciding the local vs. distance-based data replication issue,
planning should focus on the level of data replication required by assessing
data protection and recovery needs; each business (and even individual
site needs within a business) has unique requirements. After completing
the needs assessment, the proper options for meeting those needs can
be selected and installed. The final step is testing the data recovery
solution with a live run-through.
Data Recovery Needs
Each business is different, but almost all rely on stored files processing.
One chief technology officer at a storage management company begins
the needs assessment phase by answering the following questions in order
to help determine the level of data disaster recovery necessary at client
sites:
• Does the computer system
(and its stored data) directly generate income for the company?
• Do day-to-day business functions rely on access to critical data?
• Is there an impact on customer service if the data cannot be
accessed and timely information and answers provided?
• What would be the financial impact on the company if data access
is interrupted for a day, a week, a month?
• Is the company involved in emergency services that rely on data
availability?
• Is the computer storing quality assurance data? What quantity
of the manufactured products would be rendered useless if that data
were unavailable?
• Is there critical research being done which would be compromised
or destroyed if the storage sub-system malfunctioned?
The responses to these questions
provide answers to the type and speed of data recovery necessary to
protect the company.
Choosing The Right Option
With answered questions in hand, the executive begins working with the
client to develop what he calls an “application continuity curve”
(ACC) to select the right level of data replication protection and recovery
needed. The client plots the ACC, which illustrates the relevant importance
to the company of each type of stored data/files.
For example, the customer may start out with a normal backup and recovery
solution for low priority data that would not be part of the disaster
recovery effort (such as MS Word’s My Documents files, appointment
schedules, e-mail messages). Then, move onto mid-level data (e.g., marketing
information, proposals, profit projections) that would be recovered
more quickly as part of the disaster recovery capability. And finally
select top priority data the company can’t operate without (customer
lists, order processing database, human resource files, tax data) for
continual access and high-availability solutions.
The curve assigns relative weight to data that is to be protected/accessed
in the event of a disaster. This pictorial representation clearly shows
the direct correlation between the value of the data, and the investment
that can be justified to protect that data. For example, the My Documents
directory on a laptop will not justify the same level of investment
as an order processing application. As the value of the data increases
and its recovery time decreases, the cost of recovery increases. With
the curve completed, the client can decide how long they can remain
“down” within each scenario and select the recovery options
that are most effective and cost justified.
Data Replication
Data replication allows companies to have an exact duplicate of their
data on a secondary device, usually at a remote location. The replication
technique can be either hardware- or software-based and the data can
be replicated at the volume, file, block, or byte level. Which one is
right for you is dependent on your specific IT environment. Some factors
to consider are:
• Installed hardware;
• Available WAN bandwidth;
• Data type (files vs. database);
• Budget;
• SLA (Service Level Agreement) requirements.
Each approach brings different
benefits and consequences. An independent storage services company can
work with you to provide a balanced review of the options and help you
choose which one is best for your organization.
Once the decision to replicate a particular data element has been made,
make sure that the replication data path can support the amount of traffic
that will be generated moving data from the primary (source) location
to the secondary (target) location. The importance of this step cannot
be overstated. For example, database applications will generate more
traffic than file serving applications and, therefore, will need more
bandwidth availability.
Overhead associated with the replication process itself is another extremely
important consideration. Check with your supplier to make sure that
there is sufficient processing power and memory in the storage device
and/or server to replicate the data without affecting overall subsystem
performance.
Finally, decide whether to operate the replicated site in a hot, warm,
or cold state. A hot state is used for the most critical of applications
– those requiring contiguous operation and zero downtime. With
this option, the data is replicated and the application generating the
replicated data is in an online and active state. This option is the
most costly, but also most invisible to customers.
A warm state option is employed for applications that are important
but not considered critical. A copy of the application is loaded but
not running in an executable fashion. This level of recovery works for
applications that have a few minutes to a couple of hours of downtime
leeway. It is less costly to implement than the hot state recovery technique.
A cold state data replication function is preferred by businesses that
can sustain a defined extended period of downtime. It is the least costly
option as the replication site is operated as a data repository only,
without any applications loaded at the target location.
When it comes to safeguarding your data and implementing a data replication
technique, there is no “one size fits all” solution. That
is why it is prudent for most businesses to employ the services of an
experienced storage consultant to address the data protection and recovery
aspects of the company’s overall business contingency plan.
Completing The Data
Disaster Recovery Plan
After installing the physical data replication and recovery solution,
spend some time on the easily overlooked plan particulars. Write down
the names and numbers of every person responsible for each aspect of
data administration in the company and ensure each knows what they are
to do in the event of a data disaster. Make it specific as who will
notify who, who has keys to the buildings, the computer rooms, and data
storage vaults, who the back-up will be if some key person is unavailable,
etc. Make sure names and phone numbers are reviewed and updated regularly.
A major part of an effective data disaster recovery plan, and one most
likely to be overlooked, is the testing or rehearsal. Plan for a disaster
well in advance, and include the recovery solution vendor(s), every
employee, and all who might be involved in a true data disaster situation
(don’t forget the communications vendor).
Arrange a mock run, skipping no steps from original data disaster discovery
through recovery to online business as usual. If using an automated
recovery solution, ensure that it all functions as expected. Have everyone
report to any off-site facilities that they would go to if the emergency
were real, and ask that some pose as customers to make sure that the
backup data all works correctly.
Review the outcome of the test to determine what went well and what
needs improvement. Once changes are made, hold another rehearsal. During
one test, pose the scenario in which the disaster occurs during a normal
workday. During another, have the disaster occur off-hours or during
a holiday when many of the utilities and other contacts are closed.
The idea is to be prepared!
Ron Levine is a computer information
systems instructor at Santa Barbara City College and frequently writes
for Disaster Recovery Journal.
To comment on this article, go
to 1504-16 at www.drj.com/feedback.
«BACK
to the Articles Index
|
