Digging into IBM’s SAN virtualization software, Part 2

This week, we’re looking at IBM’s storage-area networking virtualization products.  In my previous newsletter, I described IBM’s SAN Volume Controller. Now it is time to look at the company’s TotalStorage SAN File System.

About four years ago at its Almaden Research Center, IBM began working on a virtualization storage project called StorageTank. The SAN File System (SFS), released during the fall of last year, is the latest incarnation of this effort.

SFS sits atop the virtualized storage pool created by IBM’s SAN Volume Controller software, and provides a shared file system with centralized, policy-based management across the storage pool. SFS is intended to supply a non-disruptive, adaptable and highly scalable environment, providing the basis for storage that adapts to business requirements. The intent is to provide a common namespace across a large set of heterogeneous clients within the storage pool.

A common file system is imposed across the SAN, which in turn allows disparate application servers to share storage. It also provides a centralized point for crucial services such as backup and recovery, virus scanning, and other management services. The SFS recognizes different categories of storage, and allows managers to assign storage and classes of service according to business requirements.

This varies from the services available on most SANs because many SANs cannot share data across applications and are thus required to replicate data (and add storage) for each app. 

SFS insulates the file system from changes to the infrastructure, so that storage and application servers may be added, brought down, and consolidated without reprovisioning or data migration.  This should enable non-disruptive storage scalability.

Much of this is currently automated. Much more, including data movement between different classes of storage, is definitely part of future plans.

SFS supports any SAN-attached storage system. Currently available are clients for AIX, Linux (Red Hat), Solaris and Windows.  Support for other flavors of Unix and Linux will be part of subsequent releases.

SFS will play a major role as IBM continues to roll out its on-demand (utility) computing strategy.  This in turn will certainly be tightly intertwined with the company’s strategy for information lifecycle management. 

Key to all this will be the policy-based assignment of data to different tiers of storage devices based on business need, and the ability – again, policy-based – to migrate to other tiers of storage as the value of the data shifts. This is not available now, but is clearly in-plan for future releases.

All this data movement will have to be transparent to the application servers, but transferring data within the bounds of a global namespace and with a system that manages physical and logical storage should make that an achievable next step.

In the last two weeks we have looked at two high performance file systems:  Lustre, aimed at engineering and scientific environments, and IBM’s Total Storage SAN File System, which is positioned for use in commercial settings.  In the future, time allowing, we may also look at other high performance file systems such as ADIC’s StoreNext and SGI’s CXFS.