Share

Where the Internet leads, with DNS as a shared directory of domains, so corporate networks need to follow - with shared directories of files and other resources.

At least, that's according to Mike Schmitt, the product marketing manager for Brocade's Tapestry file services group. His domain includes, among other things, WAFS software bought in from Tacit (now part of Packeteer) and the StorageX distributed file system acquired earlier this year with NuView.

Whether you give the resulting concept an analyst-friendly name such as File Area Networking (FAN) or a more accurate but less snappy one like global namespace, the idea is the same - you take a bunch of disparate and heterogeneous systems and centralise their file directories, so that any user or application anywhere on the network can - with the right permissions - access files without needing to know where they really reside.

"People want the ability to manage files on the back-end, so they need a shared directory, not just a shared file system," Schmitt says. "It's a bit like a DNS for files - when you type the address, you don't have to worry where google.com really is.

"So you need transparency and to separate off the back-end management, so the management tools can do whatever's needed - ILM, archiving, de-duplication and so on - and the user has no idea that anything has changed."

To do that, there has to be some sort of intermediate layer - in effect, a server that virtualises all the different subsidiary namespaces and merges them into one central directory. It also provides services such as file locking.

Of course, Brocade is not the only player here. Others include Microsoft with DFS, Rainfinity (now owned by EMC), NeoPath, Attune Systems and Acopia. Network Appliance also has a Virtual File Manager providing a centralised directory, but this is based on Brocade's StorageX.

"Then there's lots of niche guys with point solutions, for example Documentum [also owned by EMC] can provide the versioning and log files in and out," says Schmitt. "At that level it starts to blur as to whether it's competitive or complementary."

Brocade's version is software that runs on a Wintel server. "It is a central directory," Schmitt says. "It could be in or out of band, as it doesn't do anything to files, it just maps them.

"It might give different views to different users though, or do the initial cataloguing of metadata. You can also have a personalised namespace and directory service - that's great for security.

"It's tied to Active Directory - it assumes a Microsoft or NetApp architecture. By comparison, Rainfinity is more Unix-tuned, it's also in-band which lets you migrate files on the fly, and it's hardware as well as software. Brocade can migrate on the fly, but only closed files."

He adds, "We get compared to Rainfinity because it's the only other global namespace, but they actually target different things. It's the same for the others - Microsoft DFS isn't scalable and it needs manual management, but it's free with the OS."

The next vital thing is a standard, or at least a set of APIs for file system virtualisation. That way, any compatible file management tool should work with any compatible global file system or shared namespace.

"We want to enable others to plug in and use our shared namespace to do what they need to do. The namespace is the critical part," Schmitt says. "The world is more standard, so it's easier to standardise - it's all CIFS and NFS.

"So we're shooting for a standard for how the namespace works. Vendor interest in a standard API has been overwhelming - even Microsoft has endorsed it.

The next step is up to SNIA, which has already proved its worth with successful specifications and standards such as SMI-S, he adds.

"SNIA has finally said 'We need to bring the central part into the standards' - that's how to create a common interface for a global namespace. Then the customer can say 'I need de-duplication and archiving tools', and they'll work together. We will start to see more about how they all play together."