There are sound reasons for an enterprise to integrate its SANs. An FC-based SAN is truly its own island of computing. Each dedicated network comprises a fabric that links numerous servers, multi-terabyte disk arrays, and other devices. The fabric, centered around one or more FC switches, ties all of these devices together, allowing reliable, high-speed access to allocated storage resources from the servers connected to the SANs.
But therein lies the rub: In a SAN, direct access to the disk arrays is possible only to those servers directly connected to the FC switch via HBAs.
For a datacentre with only a single SAN, the SAN presents tremendous opportunities to manage storage more efficiently. Storage space is removed from physical servers and placed into a centralized location where it can be allocated, reallocated, backed up, migrated, and carefully managed. A SAN is extremely efficient, allowing each server to be given access to only the storage it needs. There's no wasted space, and modern management tools give storage admins all of the datacenter's SAN storage.
But when you bring multiple SANs into the picture, many of those benefits disappear. In a multiple SAN environment, servers must be connected directly to each SAN in order to be allocated storage resources on its disk arrays. That means multiple HBAs and multiple connections to various FC switches. Spare disk space on one SAN can't be shared with the other; shared disk volumes can't span multiple SANs; each network and its storage must be administered and maintained independently.
If one SAN is older or uses slower or less-expensive disk technology than the other, a storage administrator might want the flexibility to move data around to provide each server or application with the appropriate mix of storage capacity, performance, and cost. In a multiple SAN environment, those options are limited because each SAN is an island, protected by its FC moat and proprietary communications protocols.
Thus our challenge to the four vendors: Integrate the SANs. Make them look and behave like one SAN. Simplify administration so that one console can manage all of the combined disks and switches. Make it possible to allocate storage to servers across the combined storage network. Allow disk mirroring and backups to span the combined SAN. Tear down the walls between the islands, so that companies can gain the maximum benefit and efficiency from their servers and storage.
Sound easy? It wasn't. For the most part, the participating vendors were unsuccessful in the SAN integration challenge. When it comes to SAN integration, there is still a long way to go. And we need more openness in terms of protocols and cooperation among vendors to get there.
As it stands, vendors manage their storage arrays using proprietary protocols that make interoperability a challenge and require either continual cross-licensing or reverse-engineering by competitors. With a common set of protocols, such as the CIM-SAN1 protocol being developed by the Storage Networking Industry Association, and the means to test and certify implementations of those protocols, future integration tests should be easier and more successful.