SANs (storage area networks) have failed to work as advertised and could be replaced by application-centric direct-attached...
systems, according to a recent report from Forrester Research.
The report, Do you really need a SAN anymore?, written by Forrester senior analyst Andrew Reichman, questions whether SANs will be the dominant storage model in the future. Reichman maintains that SANs haven't accomplished what they were designed to do: improve performance while lowering cost and complexity of managing applications and databases.
"It's been the conventional wisdom of the past 10 years that to provide the best performance, protection and capacity utilization for applications and databases, you need a robust storage array in a storage area network," Reichman wrote. "But with low capacity utilization, the inability to prioritize application performance, long provisioning times and soaring costs, SANs haven't lived up to their promise."
The report said the purpose of SANs was to reduce cost, complexity, SAN islands and incompatibility, yet "instead of a consistent, shared pool of resources for all applications, storage in many shops is a thorny, complicated and costly mess." Reichmann pointed out some of the problems of SANs: utilization rates commonly in the 20% to 40% range, application islands caused by a fear of putting all applications on one SAN, incompatibility between different vendors' storage and lack of good tiering schemes because block storage systems know little about the information stored on them.
The report claimed that the storage system of tomorrow will be application-centered, clustered and direct attached instead of networked. Instead of centralized storage, they will concentrate on a specific application and be managed by business applications, such as Microsoft Exchange or an infrastructure application like a database or hypervisor.
Reichmann predicted these systems will have utilization rates in the 60% to 80% range and will be easier to manage and cost less than SANs. The report claimed that this type of approach has already begun, pointing to the Oracle/Hewlett-Packard Exadata system, Microsoft's recommendation of DAS for Exchange 2007 and VMware's increased storage management capability.
The report stated that application-centric storage systems may not be right for everybody, but "the possibility of bypassing SAN architectures for a fundamentally cheaper and simpler alternative is in the offing."
The Forrester report doesn't go deeply into network-attached storage (NAS), except to predict that file storage will remain the domain of storage vendors. Reichman told SearchStorage.com that he sees NAS systems separate from those that run applications in large shops with file and block storage in one system for smaller organizations.
"In larger environments I'd picture a silo for departmental file storage and/or special purpose apps that create files," he wrote in an email to SearchStorage. "So, picture 50-plus terabyte silos of file storage that would be separate from silos for Oracle or Microsoft or VMWare apps.
"For smaller shops, below 50 TB or so total data footprint, I would picture them wanting a single box that does application storage and NAS in the same system. For them, DAS is less appealing, as they want to be able to pool the storage and replicate it, and aren't likely to want many separate silos across their data."