Integrated Solutions
by Ed Miseta
August 2004
Ensuring you have an effective disaster recovery plan in place means first determining the value
of the data and applications you are trying to protect.
The risks of not performing a proper backup of your data can be severe. Studies
have shown that computer disasters are one of the main reasons that smaller
companies fail. "If a company is running on marginal cash flow, and it
experiences a computer disaster and can't perform normal functions such as
billing, inventory, payroll, or accounts receivable, that can very easily drive
it out of business," notes Morgan Edwards, president of UltraBac Software
(Bellevue, WA). "Companies can easily get tapped out on their line of credit in
a short time."
Aside from properly backing up your data, which is the first step in any
effective disaster recovery plan, Edwards states that off-site backups are one
of the best ways to make sure data is protected and can be accessed in the event
of any type of data or application disruption. "How can you possibly recover
from any type of a disaster if you do not have a backup copy of your data stored
off-site?" says Edwards. "My most critical piece of advice for any company
considering implementing a disaster recovery program is to perform a backup that
is either hosted off-site or is carried off-site every 24 hours. By doing so,
end users can ensure that their loss will, at most, be limited to one day's
worth of data."
1. Redundancy Ensures Data Availability
Dick Benton, senior consultant for GlassHouse Technologies (Framingham, MA),
agrees that disaster recovery requires redundancy. He believes that in the event
of a disaster, a backup copy of the data must be sitting off-site. "The only
disaster recovery process that really works is one that is carried out to
multiple locations," he says. "With medium to large companies that are spread
out over multiple campuses, the disaster recovery solution that will not only
work, but will also minimize expenses, is the one that involves having the
mission critical applications and data at multiple sites. If one of the sites
were to become inaccessible, the other sites could still carry on."
Benton believes there are many disaster recovery issues that most companies
simply do not consider. He notes that 9/11 really changed the way we think about
restoring data. "Before 9/11, who would have thought that during a disaster,
administrators might not be able to get to their disaster recovery site?" he
says. "No one ever thought that roads and bridges might be closed or that
airlines could be grounded."
Fortunately, users don't have to deal with many 9/11-type disaster situations.
Still, the investment in disaster recovery solutions needs to be directly
related to the value of what is at risk. "We encourage clients to have a clear
understanding of the value of their data," says Benton. "There may be only a few
applications and a small amount of data that is actually mission critical, and
that is where the money needs to be spent. The investment has to be justified.
Less can be spent on data and applications that do not have to be back online
until much later."
2. Determine The Value Of Your Data
Of course, determining the value of data and applications is not an easy task.
Ask end users how much of their data is valuable, and they will normally say it
all is. Chris VanWagoner, director of product marketing at CommVault Systems
(Oceanport, NJ), notes that customers can better understand the value of their
data by looking at the cost of downtime. "If an end user has a transactional
data base that is processing orders, any downtime at all will have a huge impact
on the company," he says. "That application is far more critical than a
marketing or human resources employee not being able to access their files for a
few days. There are certainly costs associated with losing data, but there can
also be substantial costs associated with not having access to the data."
Even though end users think all of their data is vital, there still has to be a
process for determining its value. End users need to actually look at their data
and assign a value to each piece of data. They also have to rank all of their
applications based on which are the most mission critical. The mission critical
applications need to get the highest priority, and any disaster recovery plans
need to make sure those applications can be back up and running in the shortest
amount of time.
"The biggest issue end users have to face is balancing the risk of downtime
versus the expenditure needed to prevent that downtime," says VanWagoner. "There
are some simple and inexpensive things customers can do to provide needed
protection. Backups are great for disaster recovery, and most businesses are
doing that. But they must also move all of that backed-up data off-site to a
secure location. That provides the disaster protection component without taking
on a lot of additional operational issues and costs."
3. Mirror Data To Remote Sites
For those customers that do not want to move tapes off-site every night,
mirroring the data to a remote site is another option. However, Zophar Sante, VP
of market development for SANRAD (Alameda, CA), advises customers to look at the
performance of the connectivity when looking at remote replication solutions.
"The connectivity between the local site and the remote site is critical," he
says. "Many end users forget that most remote links are very slow. As a result,
a lot of disaster recovery implementations end up being fraught with problems
because of a slow link."
At a minimum, Sante advises that customers use a T3 (extremely high bandwidth
circuit) connection, which has a maximum bandwidth of about 4.5 MB/sec. Most
Windows servers read and write to the storage systems at 5 to 10 MB/sec. With
anything less than a T3, the data flowing from 5 or 10 Windows servers will
quickly saturate the wire. At that point it will no longer be able to support
the remote I/O (input/output) requirements. Sante has found this situation to
exist in almost 90% of the situations SANRAD is called in to.
Another solution is to use snapshots, which are pictures of the data at
specific points in time. "We always recommend that customers perform a backup of
snapshots to the remote site," says Sante. "By using snapshots of the data, the
backup can be performed off-line. Backups are also designed to deal with the
slow performance pipes that can exist with IP (Internet Protocol) storage.
Taking a snapshot of the data set and then backing that up to a remote site
takes care of the problem of dealing with the slow interconnect. Remote
replication takes a lot of time, and using snapshots allows the replication to
take place without any disruption to the production data set."
Click here to view article.
|
