(866) 683-8844 Phone
(609) 683-9664 Fax
338 Wall Street
Princeton, NJ 08540
(Directions)
This
page is
printer friendly.
TechNote:
Replication and Disaster Recovery
Asynchronous Mirroring & Snapshots for Replication and Disaster Recovery Readiness
Remote data replication is a critical component of any plan that will allow an enterprise to rapidly recover from a major interruption. Disaster Recovery (DR) and Business Continuance (BC) have certainly been concerns for CIO's as a requirement of good business practices for a long time. However, the increased security concerns of recent years have served to focus attention on the need for more robust data protection and operational recovery strategies in DR and BC planning.
Legislative compliance with the Gramm-Leach-Bliley Act (GLBA, 1999) for financial data protection, Sarbanes-Oxley Act (2002) for accounting reform, and the Health Insurance Portability and Accountability Act (HIPAA, 1996) are well-known drivers for data replication and protection.
The 2003 blackout that struck the northeastern US and parts of Canada served as a reminder that having a DR site in the same power grid can undermine the best DR readiness planning. As a result, CIO's are contemplating trans-continental or even trans-oceanic data replication.
Synchronous
mirroring of data is good internally for the data center, and externally for short distances, perhaps as great as 10km (6 miles) over dedicated, and often costly, "dark fiber" gigabit links; however, the impact on performance at the primary site can become a significant consideration as the distance increases. In addition, synchronous mirroring often (but not always) requires the use of identical storage infrastructures on both ends of the link.
Asynchronous
block-level mirroring of data offers a number of advantages over Synchronous mirroring when the mirror is located at a remote site including the following:
- No distance limitation; supports longer inter-site distances now recommended by DR experts
- No performance degradation
- Uses any standard WAN TCP/IP connection
- Bandwidth elasticity accommodates network outages and standard TCP/IP QOS protocols
- Hardware independence; suitable for heterogeneous, SAN storage systems, & multi-OS environments
- Scalable from small to large enterprise environments
- Significantly lower overall cost than Synchronous Mirroring for remote operations
- Multiple individual remote servers can asynchronously mirror to a single data-center site
InfraStor's StorDOMAIN™ Modular Storage Servers, powered by DataCore Software's SANsymphony™ and SANmelody™ products (www.datacore.com), allow either Synchronous or Asynchronous mirroring to be implemented with complete freedom of choice in storage hardware. The Asynchronous IP Mirroring (AIM) option offers enterprise storage managers the options of bi-directional, many-to-one, or one-to-many mirroring between sites that can be at any distance from each other. When combined with wire-speed encryption devices, the security of any asynchronously mirrored data is ensured while it is in motion across wide-area connections.
Use of Snapshots with Asynchronous Mirroring
The Snapshot option can be used to create instantaneous (and if desired, repetitive) point-in-time snapshots of the data on both local (synchronously mirrored) and remote (asynchronously mirrored) volumes. Automation of the point-in-time snapshots using InfraStor's SANsnap!™ utility can ensure that recovery can be made from a very recent copy of the data, and not just from the last backup that may be hours, or even days old. The snapshots can also be triggered by data synchronization markers in the remote replication stream, and used for remote backups or rapid disaster recovery.
Because a snapshot is a reference to a state of the volume and the data at a specific point-in-time, it can facilitate recovery from corrupted volumes by "rolling-back" the volume to a previous state. Even data that has been deleted (and therefore is absent from each mirror) can be recovered by mounting the snapshot as a new volume and copying the deleted data back to the production volume in a matter of minutes.
Snapshots can be automatically mounted by a dedicated backup server and used as the source for backup to disk or tape, reducing the need for backup software client licenses, eliminating backup windows and the impact on LAN performance that is encountered with conventional backup schemes.
Economy of Operation
Asynchronous Mirroring connectivity costs scale on a per-site, rather than individual device basis. For example any storage device managed by a StorDOMAIN™ storage server can be mirrored. This open-systems approach eliminates any vendor lock-in with specific hardware requirements. Recent advances in the use of Serial ATA disk arrays that provide high capacity and performance at low cost mean that the remote site can be implemented with a high ROI.
Conclusion
Meeting data integrity requirements in a hardware-agnostic, cost-effective manner can be accomplished with a combination of Asynchronous Mirroring and Snapshot capabilities that are part of today's storage provisioning technology. Recovery from a disruption of services at a primary data center can be accomplished without the need for replicating proprietary hardware at the remote site.
Key Point Summary
Asynchronous Mirroring coupled with Snapshot capabilities:
- Enables remote backup, Business Continuance, and Disaster Recovery
- Facilitates the centralizing of DR capabilities for multiple sites
- Can be implemented over any IP network connection
- Makes use of WAN bandwidth elasticity to replicate data to a remote site
- Tolerant of interruptions in connections
- Eliminates any requirements for identical hardware at each site
- Allows lower-cost hierarchical (tiered) storage classes to be used at the local and remote sites
- Can be combined with automated real time snapshots for enhanced data recovery even if a volume becomes corrupted, or data is accidentally deleted
- Snapshots enable "zero" backup windows and LAN-free backups
