Taperemains popular for long-term
archival storage, but tape technology is hard-pressed to meet the
changing needs of business.
Storage volumes are growing at an alarming pace, and
tape-based backups are not always practical (or even possible)
within an available backup window. Large backup users routinely
cite cases where tape backups run beyond a weekend, actually
running into the workweek and impacting the production network.
Today's businesses are increasingly focused on tighter recovery
time objective (RTO) and recovery point objective (RPO). Faster
restorations and more timely recovery points can minimise the
threat of lost data.
Disk storage technologies are systematically replacing tape,
spawning specialised storage systems that bring better performance
and reliability to the backup process. One emerging data protection
technology that has attracted significant attention is
continuous data protection (CDP)
[see the SearchStorage.com Tech Closeup on CDP here].
Understanding CDP
Unlike a conventional backup that seeks to maintain an outright
copy of your data,
CDP works by tracking changes to your data -- often right to
the individual read/write event. By recording each change into a
digital journal on disk, a storage administrator can literally
"rewind" the server or storage array (or other storage system
protected by CDP) to a previous point in time; from a few seconds
previous to days earlier. Technicians can easily leverage this
granularity to recover from a myriad of problems such as lost
files, virus damage or data corruption due to network or server
faults. Some CDP products annotate a timeline of activity with
actual events, helping administrators to identify potentially
useful restoration points.
CDP can be implemented as hardware or software. Software-based
CDP is typically implemented through an agent running on each
server you're protecting (e.g. a database server). Hardware-based
CDP appliances are also available for use in-band (in the data
path) as well as out-of-band (outside the data path). Hardware can
often eliminate the need for agents, though some software may still
be required for out-of-band applications. It's best to discuss
implementation requirements with your CDP vendor before making a
purchasing decision.
 |
| Storage All-In-One Guides | | Learn more about storage topics like disk storage,
disaster recovery, NAS, and more in SearchStorage.com's
All-in-One Research
Guides. |
|
| |
|
Although CDP can support extremely granular restorations, it does
not protect against changes or transactions that occur between a
fault and its discovery. For example, CDP cannot prevent a virus
from infecting a file, but it can restore the infected file to its
preinfected state -- the problem is that any work done to the file
since the restored point is lost. Some amount of data recreation
may be necessary depending on the fault and its impact on your
data. Consequently, CDP does not alleviate the need for regular
backups.
It is also important to note that CDP is similar to
snapshots in some respects, but the two approaches are
different. A snapshot basically captures a system's state at a
particular point in time, much like CDP. The difference is that
snapshots are treated as an event, being taken perhaps once a day,
twice a day or maybe even once an hour. When a fault occurs, any
data generated between the last snapshot and the fault can be lost
just as with any conventional backup. By contrast, CDP is
approached more as an ongoing process, recording all activity in
real time and allowing restoration back to a precise point just
preceding the fault. Snapshots are often visualised as just one
"slice" in a CDP timeline.
CDP products
There are several key vendors in the CDP marketplace, including
Revivio, Storactive (recently acquired by Atempo), Asempra
Technologies, Mendocino Software, TimeSpring Software, Topio and
XOsoft. Each vendor brings its own unique emphasis to the
technology and its use in the enterprise.
Companies like Mendocino are touting the idea of manageability,
using event annotation to improve utilisation of the CDP timeline.
Rather than selecting a restore point based simply on a timestamp,
an administrator can select a restore point that corresponds to a
more significant system event. Mendocino calls this
"event-addressable storage." TimeSpring follows the idea of
manageability through "offline replication," allowing offline
testing and inspection of data without impacting the production
network. This approach also allows protected data to be used for
other purposes beyond backups, including business intelligence or
lab testing.
Numerous vendors are applying CDP technology to specific
applications. For example, Storactive has geared its LiveBackup
software product for Windows-based tasks like backup/recovery and
disaster recovery. Storactive's LiveServ software protects Exchange
servers for e-mail backup and recovery. TimeSpring's TimeData
software is available in several versions supporting SQL, NT file
system and Exchange environments. XOsoft provides WANSync software
designed to offer CDP features between remote offices.
Companies like TimeSpring, Storactive, Mendocino and XOsoft
implement CDP as software, but Revivio favors hardware
implementations in products like its CPS 1200 or CPS 1200i. The CPS
is a nondisruptive, block-based, out-of-band appliance intended to
protect mission-critical enterprise applications without impairing
application performance.
Applications of CDP
The applications of CDP are as varied as the vendors' products.
Some users employ CDP technology to avoid the time and trouble
normally associated with traditional backups that frequently ran
overnight into the workday, through a weekend into Monday or even
failed some way through the process, forcing administrators to take
more backup time -- or forego a backup entirely.
A substantial volume of valuable corporate data is often left on
laptops or remote locations that aren't routinely protected by any
backup strategy at all, so some CDP users focus on supporting
remote users. In most cases, CDP software can protect laptops and
remote systems across relatively slow WAN links. When a laptop user
experiences a lost or corrupt file, the system can be restored from
a CDP platform in the corporate data center. System administrators
often find the biggest problem is getting remote users to regularly
employ the CDP capability.
CDP is also employed to protect specific applications, such as
the corporate Microsoft Exchange server or a database like Oracle
or SQL. For example, CDP allows lost or deleted e-mails to be
recovered far more efficiently than systematically searching
through tape backups. An administrator can simply look back through
the CDP logs, find the deleted/lost message and restore the e-mail
directly from disk.
Finally, CDP is not appropriate for every enterprise. The
technology isn't terribly difficult to use, but it is expensive,
and the CDP paradigm requires a fundamental rethinking of data
protection. As a result, CDP is best suited for organisations that
seek a negligible backup window and RPO; enterprises with busy
transactional network traffic are often the best fit. Businesses
that do not require those benefits may find better value in other
disk-based backup technologies like virtual tape libraries or
snapshots.