In part one we learnt about Brocade's file area network initiative and its core global namespace attribute. We also found out that the FLM product which files from a primary to a secondary storage tier was not part of the global namespace, that such movements were stub-based. A benefit of FLM is that it moves inactive files off primary storage and thus reduces the amount of primary storage data needing to be backed up.

Wide Area File Services

The seminar then moved onto the topic of wide area file services (WAFS). There was nothing earth-shattering here. CIFS is a very chatty protocol, too chatty for file access over the WAN. So stick a core WAFS device at the edge of the data centre and an edge box at the remote office with an IP link between them. Then use TCP/IP acceleration as in the Brocade-approved WAFs devices (from Tacit networks via Packeteer)

Where is Brocade's FAN gear going?

Simon Panton, a Brocade representative, agreed that a logical end-point for all the various FAN products we heard about at the seminar was a single device or appliance on which the FAN products ran. We might envisage, and this is speculation, a FAN controller sitting in the (IP) network and connecting file storage devices on the one hand with applications in servers needing to access the storage on the other.

This is a similar position to a SAN director running block storage management applications. That should be no surprise as Brocade is an 'in the network' supplier. In the global namespace arena that puts it in a similar class of supplier to Acopia which has a network switch-based file virtualisation and global namespace product conception.

It is logically opposed to filer suppliers such as Isilon, NetApp and BlueArc, who are building clustered filers sharing a global namespace. The problem with filer-based approaches is that the supplier typically wants you to buy their drive arrays. The expectation with appliance and switch-based suppliers like Brocade and Acopia is that you get your drive arrays for file storage from wherever; they are decoupled from the FAN software technology.

That implies, and this starts getting both exciting and technically difficult, that you can have heterogeneous file storage managed by the FAN appliance/switch. What happens when the heterogeneous file storage has a virtualising, global namespace capability of its own?

We mustn't expect heterogeneous FANs to emerge anytime soon. It is far, far too early to even consider the possibility of a Brocade FAN appliance including a Isilon file virtualising and global namespace cluster inside its own file virtualising and global namespace capabilities. This begs the question about how global is global?

We needs should have a single and really global namespace, not a multitude of them. For a global namespace to be, in reality and actuality, global, then that is the logical end-point of FAN technology activities. It doesn't exist and no-one is talking about such an idea.

Brocade and Acopia are probably the two most advanced FAN suppliers in terms of their ability to understand and present FAN concepts and deliver FAN capabilities. There are, as yet, no FAN standards. No one is going to be able to buy a complete FAN product set yet but it is worthwhile understanding what FANs are, could be, should be, and may become. File-based data storage is growing massively and any sensible and powerful technology to better manage it will be a good thing.