Have you noticed how practically every storage vendor of late has been adding new interfaces to their storage offerings? I'm talking about the front-end interface. Almost all the FC vendors have added an iSCSI front end, either as a separate model number or as an option to their FC box. NAS vendors too have started adding SAN interfaces, such as iSCSI or even FC. What you see happening is what I think may be the first step towards NAS/SAN convergence.
These two trends are about to merge to create universal storage. Think of a storage system that is built on the grid principles described above and implements all the popular NAS/SAN protocols in the same box. NetApp is heading in that direction. They're not there yet, but I suspect they will lead the market in commercializing this concept of universal storage. Let's look at the pattern developing here.
Their hallmark filer product line has been synonymous with NAS for the past decade. The backend has been FC drives for the past few years. The first thing they did was to allow pure FC block traffic come into their backend without going through the WAFL file system. That gave NetApp the right to say they endorse NAS/SAN convergence -- a baby step but a step, nevertheless. Then, they added the iSCSI front end, adding to their boasting rights for leading the charge towards NAS/SAN convergence. But it is the next chess move that NetApp will make that will align it with the convergence I'm talking about.
Behind the scenes, as we all know, NetApp is working to integrate the Spinnaker technology it acquired in 2002. This technology is what brings in the grid flavor. If you recall, the Spinnaker product line was based on a clustered file system and grid architecture. The basic molecule one could buy was a node, made up of an Intel server running Linux, a clustered file system and NFS/CIFS server code. Inside the same box, was a fixed amount of storage. You started with one box and deployed the NAS box much like any other. If you ran out of storage space, you added another similar box to the network and the second box simply merged into the first one. This increased the amount of storage available without requiring another file system to be mounted. NetApp wants to bring in this smooth global namespace and scalability to their filer product line. It is a non-trivial exercise and explains why it has been three years and we have not seen a product from them yet. But the effort is close to paying off, and I expect NetApp will announce the converged product line sometime soon.
In any case, I fully expect that what NetApp did to the filer they will do to the new product as well. It will most likely have an iSCSI and an FC interface in addition to NFS/CIFS. The backend would scale using the grid principles of Spinnaker. However, I expect that the product would be architected from the beginning to be universal storage rather than tacked on like the filer. One would be able to partition the backend storage into volumes that could then be utilized via FC, iSCSI or NFS/CIFS. But they would be managed as if they were one storage unit. They would allow the size of these volumes to be changed on the fly, without disruption. This would come very close to the vision of a modular, converged NAS/SAN, scale "forever," multi-tier, easy-to-manage storage system with support for global namespace, NFS/CIFS/iSCSI/FC front-end and FC/SAS/SATA back-end disk drives. If you have a few more buzzwords, go ahead and throw them in. You will probably be correct … at least on paper.
Another approach I expect some companies to take can be exemplified by the combination of HP, PolyServe and LeftHand Networks. HP has a separate relationship with both PolyServe and LeftHand Networks. I can certainly see the possibility of running a clustered file system on standard Intel-based nodes and creating a NAS/iSCSI combination that could quite possibly perform very well, especially with SAS drives on the backend.
In my view, of the legacy vendors, NetApp is the farthest along with this next generation idea of storage. But I suspect all the majors are working towards this. Amongst the startups, the leadership relating to the merger of these concepts is coming from players such as LeftHand Networks, Intransa and Compellent. I would also include less-known players like Reldata, who is just entering the US market.
Assuming I am correct, and all this transpires in the next six months, then at least these vendors would have taken a huge step towards incorporating the benefits of homogenous virtualization. So, what about heterogeneous virtualization, you ask? When would one be able to get a true enterprise class product that could virtualize across heterogeneous storage boxes? Don't be silly and don't ask for too much.
As for me, I would be thrilled even for this to come to pass. I think you would be, too.
About the author: Arun Taneja is the founder and consulting analyst for the Taneja Group. Taneja writes columns and answers questions about data management and related topics.