Exploring backup alternatives in a SAN environment

However there are many factors that can impact the suitability of a SAN backup architecture in your environment, such as the technology requirements and the type of data that you intend to backup.

The industry trend towards consolidation and re-centralisation of storage has provoked changes in backup traditional architectures. We have seen enterprises move away from large-scale network backup solutions towards a far more consolidated model.

More recently, network-attached storage appliances have resulted in new backup methodologies. Additionally increased deployment and visibility of SAN environments has led to further developing of backup architectures.

This article looks at some of the newer backup technologies that address the growing presence of the Storage Area Network, and investigates how this new architecture can be exploited to deliver better backup performance and more cost effective manageability.

Backup Architectures
The wide varieties of environments requiring backup services and the numerous storage connectivity alternatives have resulted in several distinct backup architectures. These backup architectures have evolved with the changes in enterprise backup and recovery requirements and the availability of more sophisticated hardware and software infrastructure components.

Many of these backup architectures are applicable in a SAN environment with the availability of technologies such as IP over Fibre. In order to understand what best fits SAN environments in particular, it is worthwhile to first complete a very brief survey of the various backup architectures that exist, and to define terminology.

Local backup is the most common and straightforward mechanism for providing protection of data. Using this technique involves copying data objects (files, volumes, databases) from primary disk to secondary storage, most commonly tape media.

A local and wide area network backup architecture is an obvious alternative to the local backup architecture, and typically involves a client-server relationship.

The backup client (agent) produces a backup image from disk and sends the image across a network to the backup server, which directs this image to secondary storage.

Network Controlled Local Backup is an effective blend of the local and network backup techniques providing network controlled local backups. In this configuration the master server provides centralised administration and the media server provides local backup for itself and possibly network backup for other clients on the LAN or WAN.

The Network Data Management Protocol provides a means to perform backup of servers and appliances that are typically unable to host native backup clients or agents. NDMP backup architectures are used predominantly in environments where appliances and specialised fileservers are part of a larger enterprise backup environment.

Shared Tape Backup gives you the ability to share secondary storage devices between multiple hosts requiring backup protection. Typically fibre channel SAN's are use as the enabling technology for this type of solution.

Second-host backup is a technique designed to use shared disk environments to offload backup operations from an application server to a dedicated backup server.

Third-party copy backup techniques provide essentially the same functionality as second-host backup, the disadvantages of the third-party copy backup techniques include the evolving nature of the XCOPY specification is the loss of common platform functionality and resulting increase in management complexity.

SAN Backup Supporting Technologies and Requirements
Of the techniques discussed the two methods that offer the most significant benefits in a SAN environment are second-host backup and third-party copy.

Many of the technologies required to implement second-host-backup and third party copy in a SAN environment are also necessary (or at least useful) in non-SAN environments. In this section we look at the prerequisite technologies supporting these advanced backup techniques.
One such technique is Object Mapping.

Object Mapping is the process of mapping logical file information into absolute physical addresses. The mapping process must take into account various levels in the data management stack, including applications, database management systems, filesystems, and volume managers.

This technique is required to ensure that a backup product can understand and respond to changes in how the data is layed out and addressed on disk. Some maintenance and redundancy techniques, if they occur at a specific point in time, may invalidate what the backup product thinks it is backing up. The net result would be a corrupted (potentially undetected) backup.

Snapshot technology provides reliable SAN backup, the basic premise is to create a consistent point-in-time image of a given file system or data volume, and use this image as the reference point for the backup operation. Once a snapshot is in place, the primary data can continue being modified without affecting the backup operation.

The concept of Application Quiesce is closely related to the snapshot technology. In order to obtain a consistent and useful version of the data on backup media, it is vital to coordinate the snapshot (and the resulting backup operation) with the application that manages the data. Application Quiesce is the process of temporarily pausing application updates and flushing application buffers at a point in time where application consistency exists.

The high-profile benefit of such SAN backup techniques is the ability to export the actual backup copy operation from the primary application host to an off-host entity. The off-host copy service can reside in either a second host on the storage network or any other SAN element with visibility of both the backup source and destination.

Conclusion
After examining the various backup architectures and techniques, both SAN and LAN, and exploring the data environments, several observations and conclusions can be reached:

1. There is no 'one size fits all' backup architecture. Combining multiple backup methodologies can often provide the best solution.

2. SAN backups are not trivial operations. There are several key technologies required for successful and reliable SAN backup deployments. Although the potential benefits are high, the current reality is that due to the complex prerequisites, only a limited number of environments will be able to take advantage of SAN backups.

3. Snapshot technologies can drastically shrink or eliminate the backup window, but they require additional online disk storage, up to three times the storage if the mirror technique is used.

4. Compared to conventional backup solutions that are host-resident, SAN backup solutions may not have built in features like load balancing and clustering/failover. In this regard, solutions such as second-host backup need to be evaluated versus third-party copy data movers.

5. While SAN backup solutions with exported data copy services do drastically reduce (or eliminate) the I/O overhead on the application host, the potential for I/O contention between the application updates and the backup data copy still exists.

In summary, one must closely examine the variety of backup architectures and techniques being offered against the environments (physical and logical data) being protected. The potential benefits of SAN backup solutions are huge, but it is important to fully understand the caveats and limitations before deployment.