Global file namespaces, tuning your file servers with Attune

Wednesday at TechTarget's StorageDecisions, I spent some time talking with Attune networks.  Attune came on to my radar by way of a session given by Brad O'Neill from the Taneja Group.  They have an interesting product for managing distributed windows file servers with better unification. Through that management, they enable some interesting capabilities for shifting files around by different criteria while making the underlying infrastructure transparent to end users.  One of the nice features is that they do this without using pointers to moved files, instead managing relocation through the use of metadata and remapping of file locations.  This seems to be a better long term approach.  Frequently, I'm left wondering about the long term viability of data migration through file stub approaches, especially if you're forced into a position of changing platforms.  Additionally, there are a number of customers on the market for which current high end file service aggregation appliances just aren't affordable.  Attune's real attractiveness is the high level of integration with Microsoft Windows file servers, which are frequently a bigger struggle for growing enterprises, and Attune may offer some hope for unified management without an approach which involves forklift upgrades of the environment with high-end appliances.

But the Attune system leaves you in a position of introducing an in-band and windows centric system into your environment.  If you're rolling up lots of high I/O file servers, then effectively front-ending them with an in-band solution like Attune might introduce some limitations.  If you're in a bigger environment, it's worth asking a few questions about how much load the Attune system has to support given the current load of your file serving environment.  Another concern might be backout of the system - specifically how you reverse engineer and break out your file serving enterprise in the future if you need to move away from the product for one reason or another.  Moving away from the product actually seems easier with this approach than with traditional stub approaches - it may take you a while to figure out how to present your data but at least you aren't reversing file stubs and other sorts of stuff - but this is something to keep in mind when you engage on any significant file namespace endeavor.

Even as information centric as we are today in the enterprise, we're only becoming more information centric, and there's a huge gap between our ability to engineer block performance and our ability to manage information in a context aware way.  There are a number of tools coming on to the market to try to engage this struggle, ranging from more content oriented storage systems to global file namespace approaches. It's nice to see some alternative approaches coming up in the global file services name space areas. 

The Attune approach has a lot of great immediate merit for enterprises with lots of smaller distributed file servers.  This is a pretty common struggle, and if you're debating the headaches and trade-offs associated with consolidation, Attune is worth looking at.  But for bigger file serving enterprises, it might be worth investigating, but do your homework on current custom hardware and potential scalability issues with bigger I/O loads.