Storage Pipeline: First Person: Global File System or Global Name Space?

With file copy sharing falling by the wayside, IBM has pushed out its first-generation Storage Tank, complete with a universal file-sharing system. This system requires agents on every server, where they intercept normal server file-sharing functions to place all files into IBM's proprietary SAN file system--preferably on IBM-branded storage platforms aggregated into storage pools behind an IBM Storage Virtualization Controller.

Not to be outdone, virtually every NAS vendor is pursuing a strategy of NAS head clustering, combined with some sort of proprietary "clustered" or "extensible" file system. Of course, this solution will work only if you buy the vendor's NAS heads.

Meanwhile, there are rumblings in Redmond that we should abandon file systems altogether and use a database instead--preferably, Microsoft SQL Server.

Frenetic Engineering?

Basically, we're being told that it's time to get all our files singing on the same set of semantics, and it will take a disruptive rip-and-replace overhaul to our server file systems, with standardization around a single vendor's technology, to make this happen.IBM is working on a wide-area file-sharing system called Distributed Storage Tank. Interestingly, this effort is taking Big Blue into the realm of open standards as the vendor seeks to develop a master LDAP name-space server for a file name-space service leveraging untapped capabilities in Network File System 4. Microsoft, for its part, is developing an open, object-oriented database replacement for the file system it provides with its servers. Whichever approach becomes dominant eventually will be supported by the NAS/SAN folks, despite their work on proprietary schemes.

The Real Issue

These "improvements" mask the real issue: the inability of FC SANs to store files. With nearly 70 percent of corporate data in file format, you can imagine the acid reflux that develops when users are reminded that their expensive SANs aren't optimized for files at all.

Should we replace our server file systems? From a pure engineering perspective, this tack is probably better than the current practice of using surrogate global name-space servers from vendors like NuView and Tacit Networks to broker access to back-end files. But we must consider the disruption--and the cost--of fixing the architecture. In these days of "good enough" solutioneering, the expedient fix--a surrogate global name-space server--might be a better fit than the engineer's fix. At least it will buy time until the strategies under development have time to gel.

Jon William Toigo is a contributing editor to Storage Pipeline, CEO of storage consultancy Toigo Partners International, and founder and chairman of the Data Management Institute. Write to him at [email protected].0