|
By Ravi Chalaka
 Implementing
a remote data replication policy is the first step toward a comprehensive
disaster recovery plan – which is no longer an option, but
a necessity for many IT environments. In addition to the basic replication
process, data mirroring can also be employed as part of the DR solution.
The process of mirroring involves the use of a "shadow"
disk that is updated in parallel with the primary disk, providing
a real-time or near real-time copy of the primary disk. Local mirroring
provides the first level of data protection with a mirror disk attached
to the host machine or an appliance located at the primary site.
In the event of data loss on the primary disk, the data is retrieved
seamlessly from the mirror disk.
There are two types of mirroring techniques
that may be used over a remote IP to establish a copy of the primary
site data - synchronous or asynchronous. Synchronous mirroring solutions’
performance depends on the link bandwidth (speed) and distance spanned
by the remote link. Each write transaction to the disk is sent to
the remote mirror disk and the application cannot continue until
that transmission is acknowledged from the remote location. Synchronous
mirroring solutions provide maximum data protection at the expense
of degraded primary site performance and reduced network throughput
on the link to the disaster recovery centre.
An optimum alternative that has minimal impact
on performance, while minimising data loss is asynchronous mirroring.
With this technique, multiple writes are transmitted without waiting
for individual acknowledgments. The use of asynchronous mirroring
offers the additional benefit of near real-time availability of
data; online standby of data is only a few writes behind the primary
site.
Data replication made easy
Data replication is the basis of all disaster recovery solutions
and involves periodically copying of a volume’s data onto
a secondary storage device, which can be located any distance from
the original, preferably far away. If the main storage device should
fail, data on the secondary storage device can immediately be promoted
to primary status and brought online. Replication is a continuous
process that begins by establishing a complete copy of data at risk
at the disaster recovery (secondary) site. With that copy as a baseline,
the replication process continues, recording any changes to data
and forwarding those changes to update the secondary site based
on watermarks with a user-specified policy.
Fast recovery
For true protection against major disasters, disaster recovery centres
require remote sites to be located hundreds of miles away from the
business, raising issues of data loss and synchronisation of data
between the production and DR site. The ideal disaster recovery
solution provides quick time-to-recovery (TTR), assuring continuance
of operations through near-uninterrupted access to data. Additionally,
the solution must minimise data loss for a graceful recovery, keeping
primary and DR centre data sets synchronised while minimising human
intervention to reduce errors during the recovery process.
For robust data protection, the remote replication solution must
also work in conjunction with options for multiple point-in-time
snapshots to provide full, incremental, or differential automated
instant backup to disk. This allows administrators not to have to
spend needless hours trying to recover accidentally deleted records
or virus-infected files from tape, even in the event of a disaster.
Snapshot agents minimise recovery time. When data replication is
used in conjunction with snapshot agents, the data has full transactional
integrity in addition to point-in-time consistency. This means the
replica can be immediately put into active use without going through
a ‘consistency check’ process that can be very time-consuming
for large databases. Simply put, disaster recovery and replication
solutions require flexibility, ease of deployment, scalability and
rapid recovery.
Reducing cost
With the IP-based remote replication, data is replicated from one
site to another over any existing LAN, MAN or WAN network infrastructure.
Data centres use Fibre Channel for their storage networks and a
simplified DR solution must provide interface to MAN and WAN routers
without the need for extra FC-to-IP converter boxes. These solutions
also need to be independent of the type of storage subsystems, application
servers and operating system platforms rather than work only in
a homogeneous environment. By eliminating the need to deploy matching
disk arrays, file servers or application servers at the secondary
disaster recovery site (except for hot standby scenarios), such
systems can offer unprecedented flexibility in creating disaster
recovery environments and allowing for low-cost DR planning by using
low-cost JBOD or ATA based RAID arrays at the DR centre.
Accelerating pre- and post-disaster
operation
During initial set up of pre-disaster operation, synchronisation
of data between the two sites can be done using tape backups or
local mirroring, shipped to the remote site, followed by a delta-sync
process to facilitate minimal transfer of data over expensive WAN
links. During an emergency, an IP link may be deployed to allow
emergency access of the data at the disaster recovery centre over
WAN by any server located anywhere. This IP connectivity provides
significant advantages, offering more ways for temporary offices
to access data during an emergency. A reverse delta-sync process
facilitates fast recovery of the primary site when the emergency
is over.
Flexibility in preparing for a catastrophe
To defend against site failure by providing automated off-site data
protection, such solutions offer fast remote data synchronisation
of data on disks — across the street, across town, or across
the globe. In case of a catastrophic failure at the primary site,
the system administrator must be able to quickly redirect application
servers to access data from replicas located in the backup data
centre. Administrators can specify a variety of policies to control
the replication process, giving them a very granular, policy-driven
mechanism for keeping an extra set of data off-site for disaster
protection.
Summary
Remote data replication is no longer an option, but a necessity
for corporate enterprises. The amount of data at risk and the cost
to replace that data – if in fact it can be replaced –
highlight the need for a data protection solution that extends beyond
a building, a campus or even a country. While most customers recognise
the need for this type of data protection, they are squeezed by
the continued tight IT budgets and the cost and complexity of existing
solutions.
Simplified disaster recovery solutions must
be based on a customer-centric philosophy, addressing the cost and
complexity issues that have kept many customers from successfully
deploying remote data replication solutions. The complexity problem
can be solved today by using fully integrated replication solutions
built on performance-optimized appliances. By deploying data replication,
mirroring, backup and multiple snapshot storage applications on
an affordable appliance platform, a comprehensive DR and replication
solution can address the key customer requirements of flexibility,
ease of deployment and scalability.
Ravi Chalaka is vice president of marketing,
MaXXan Systems, Inc.
http://www.maxxan.com/
DOWNLOAD PRINT-FRIENDLY
VERSION
•Date:
10th Sept.r 2004 •Region: N.America/World
•Type: Article •Topic:
IT continuity
Rate this article or
make a comment - click
here
|
