The decision to implement SAN storage at this firm was based on the need for cost efficiency and additional functionality for application environments. To support a major migration effort from VAX to Alpha in 2001, the storage support staff realized new technology was required.
The firm became an early adopter of SAN technology almost three years ago with its initial implementation of Fibre Channel (FC) switches to support shared, clustered file systems for their VAX-to-Alpha migration. This first SAN environment was built to allow booting directly off of the SAN fabric (with no local host storage required), and it eventually grew to 20 fixed Brocade switches. A year later, the support team began moving Windows and Solaris systems to Fibre Channel SANs attached to high-end external storage arrays from Hitachi Data Systems™.
Over time, the proliferation of SAN islands and the manageability of multiple SANs across several discrete datacenters became a source of inefficiency. To reduce the number of SANs in the environment, the team began consolidating SAN islands onto the Cisco MDS platform. In addition to consolidation, however, the staff also saw an opportunity to augment the resiliency of the SAN architecture.
Today, local, direct-attached storage is still supported (there is no wholesale migration planned to get rid of DAS) when the application requirements can be met within the constraints of DAS functionality. There are three primary application requirements that dictate the need for a SAN-based infrastructure:
Replication
Multiple redundant paths
Shared data storage (such as required by clustered file systems)
If any of the preceding features is required by an application, the application is built on SAN storage no matter the size. In other words, the architecture of the application is the most critical piece of the storage networking decision matrix.