Introduction to SANs: Technology, Benefits, and Applications - Page 7

Storage Consolidation
A major benefit of SANs is the consolidation of storage resources. SAN software and hardware have the ability to provide a centralized storage solution for servers. Interoperability issues due to operating system incompatibilities (most OSs dont like sharing storage with each other) and file system incompatibilities make the sharing of storage resources a difficult proposition, but the emergence of advance storage virtualization and management packages will eventually make the heterogeneous SAN a reality. Future articles will explain in-depth the challenges of building mixed SANs and the available solutions.

LAN-Free Backup
LAN-free backup first came into existence over 2 years ago and to this day, is still one of the biggest selling points of Fibre Channel SANs, especially in the workgroup to midrange class networks. As the name implies, LAN-Free backup moves backup traffic away from IP LANs and onto Fibre Channel SAN links. Another important feature of these products is that they facilitate the sharing of expensive tape backup libraries and autoloaders.

Two major vendors selling LAN Free backup solutions are Veritas Software with their NetBackup solution (with the Share Storage Option) and Computer Associates with their ARCserve Backup.

Backup over a SAN configuration

In tests, both of these solutions were able to move backup traffic away from our IP network and onto our SAN. Both of these solutions also featured intelligent management software that was able to specify which server had control of our shared tape library. A key component in any shared storage solution is an intelligent media manager, which not only keeps track of data, but also knows where that data came from. Without intelligent media management, the danger of servers writing over each others tapes is very real, and potentially deadly if the need to do a full data restoration ever occurs.

Serverless Backup
The reality of 24/7 IT shops has made Serverless Backup an enticing option. In a "Serverless" backup scenario, the CPU cycles that are normally burned doing a network backup are offloaded to intelligent SAN components (most often done by adding intelligence (an extra CPU) to SCSI/Fibre Channel router units).

On a SAN using Serverless backup, a physical volume can be ordered to back itself up or self-initiate a backup and tell other devices on the SAN that the physical device is unavailable, and any modifications made to its files will have to be temporarily recorded elsewhere until the backup is complete. This allows backup to take place alongside regular user activities. The high bandwidth typical of SAN networks shortens the time necessary to run a backup and to merge changes made to a volume after its backup is complete.

When devices on a SAN have this level of intelligence, they can operate in conjunction with SAN management software to do other operations on data besides mere backup. For example, storage devices can migrate their own data according to predetermined migration rules without having to wait for a server to initiate migration. A RAID system can migrate little-used files or data to archival storage. A tape library can notice if a file or block of data is being heavily used and migrate it to faster storage. All this activity takes place without input from administrators, servers, or users, freeing up server resources.

It also means that the SAN is not dependent on servers for performing housekeeping tasks. Servers are merely SAN storage resource access points for users and administrators.

Serverless backup typically uses "snapshot technology" to take a digital image of a RAID subsystem, which the Serverless backup components can then backup to an offline storage unit like a tape library. The use of the snapshot enables the servers, which are still utilizing the live data on the RAID subsystem, to continue to use that data without interruption, while the Serverless backup components quietly direct data through the SAN from the disk arrays directly to the tape unit, without going through the server. With this combination of technologies, IT managers can finally backup important "nonstop" data (which is required to be highly available), without having scheduled downtime or server performance degradation.

Legato, Computer Associates and Veritas all have Serverless backup solutions.

High Performance Caching (using SSD)
High performance storage caching is an emerging market that focuses on pushing up the performance of ultra high-end applications like enterprise-class databases. Well see these high-end storage caching units placed in front of SAN-enabled database clusters, allowing IT managers to boost overall performance of the clusters, provided the Cache hit ratio (I/O requests fulfilled by cache / Total number of I/O requests) is high. It is unlikely that this technology will be implemented in low to mid range shops anytime in the near future.

SSD (Solid State Disk) storage units that are built out of high performance RAM are being pushed as the top caching repository solution. SSD can support data throughput rates and IOPS (input output per second) several times faster than anything possible in the hard drive market, but this performance comes at a significant price. For example a RAM-SAN 8GB solid state, fibre channel drive from Texas Memory Systems has a list price of $50,000.


In closing, this wraps up our first foray into the land of SAN, and we hope it was informative. We expect to dive deeply into many other SAN technology areas in the future, as highlighted throughout the article. If you want to give us some feedback or share your successes or struggle with implementing SAN technology, let us know in our community forums.