What seems to be the problem? Often, pundits state that a lot of the key
information within enterprises is contained on the large number of
desktops and laptops in the organization. In fact, the disk space used
by all of those computers may be more than in the organizational data
center. Two questions arise: how to backup this data and how to ensure
that no key data is lost since now a lot of data is being protected. And
the answer to those two questions is the challenge that IT organizations
face.
What do you need to know? Backing up to an internal data center is one
answer, but not only is this costly, but RAID 5 may not be enough to
protect all the large quantities of data all the time. Berkeley Data
Systems (BDS) has another answer: Simply backup to an outsourced data
center that is accessible over the Internet.
In essence, this is what Carbonite has done for consumers and BDS offers
a similar service for an equivalent price. But BDS also wants to appeal
to large customers who have hundreds or thousands of individual PCs (and
in fact, BDS has announced that General Electric has signed a contract
with it).
With a large number of customers, the question is how to protect all
that data. And we are talking not just about large numbers of terabytes
of data, but rather about petabytes of data. And that is a mind-numbing
amount of data. And RAID 5 alone is not adequate to protect all that
data. Even though two failures in a single disk array before the array
is rebuilt from the first failure is rare, with such a large amount of
data a permanent loss of data is inevitable eventually.
Now the way that a standard data center would protect against two
failures in a single RAID group is through the use of tape. However,
tape would be a very expensive addition as well as adding in management
complexity to the restore process. Simply put the data stored is bulk
storage that could be petabytes in size. However, restores tend to be
very granular. Individual personal data tends to be individual files
rather than all the application data that might be needed in a data
center data restoration project.
A simpler way would be to use disk exclusively, but introduce additional
parity into the process. Carbonite does this through the use of RAID 6
(which has double parity), but BDS has its own protection strategy,
which it feels gives even greater protection.
Berkeley Data Systems uses what it calls a distributed Reed-Solomon
(DRS) approach rather than RAID. Reed-Solomon is a reliable and
well-known error-correcting algorithm that can recover original data
files even if there is some bad data in the file. Reed-Solomon is used
in a variety of commercial applications (CDs and DVDs), in data
transmission technologies (DSL and WIMAX), and in broadcast systems (DVB
and ATSC).
BDS uses the DRS approach to provide petabyte scalability, a high level
of reliability, and a relatively cost efficient solution. BDS feels that
this solution (relative to a triple mirroring solution) would give an
order-of-magnitude more reliability at a fraction of the cost. And that
lower cost is just what it takes to make it economically justifiable.
Business needs to protect business PCs, but cannot afford standard data
center prices to do so.
What can you do about it? Enterprises have a number of choices on
backing up laptops and desktops. Not doing anything is one choice, but
that option is becoming less and less acceptable. Enterprises also have
the choice of in-house or outsourcing solutions. The outsourced
solution, such as BDS and others provide, takes advantage of both the
economies of the Internet for communications costs and economies of
scale at the outsourced data center level. Now some enterprises may
hesitate to use outsourcing because of a perceived loss of control.
However, from a security perspective, the data is encrypted. From a
management perspective, only the backup data goes out of the enterprise
-- not the original copy of the data. Whether or not to choose an
in-house or an outsourced solution then really comes down to a make
versus buy decision. And that decision is about cost, not control.
Management control issues are likely to be irrelevant for private
enterprises (although perhaps not for governmental organizations).
Therefore, choose the solution that is financially attractive.
