[CONTINUED FROM PAGE 1]
The first strategy is pretty obvious. It combines the original and current strategy of data tiering, ie, moving data within, as well as across, storage arrays. A unified management layer would enable data to be managed using a single interface. For now, it would be a single pane of glass to manage data within each platform but over time, I imagine that you could use the tool to easily and automatically move data across these platforms.
The best preview of this today is EMC Unisphere. EMC has amassed an impressive portfolio of storage technologies over the years (Symmetrix, VNX, Isilon, Atmos, DataDomain, Rainfinity, ExtremeIO, etc). Since most of these platforms were acquisitions, they have different and incompatible interfaces.
Unisphere is designed to be a single, unifying, web-based interface with plug-ins. And when EMC buys more companies, they can quickly write plug-ins to manage those platforms too with this tool. It would be great if we could do this across a multi-vendor environment but we’re not quite there yet.
There are a number of startups that are starting to use a Sun Microsystems-developed, open source file system and volume manager called ZFS to create storage appliances that store data on independent, fast, cheap disk. There’s no reason why this couldn’t also be done with block-storage. This goes a long way in answering the perennial question, “why does my storage vendor charge me $1,000 for a 2TB SATA disk when I can get that same disk for less than $100 at my favorite big box store?”
What if an IT organization can buy a storage appliance that performs all the advanced functions they expect from multi-million dollar arrays but allows them to store that data on whatever disk they choose? Sounds compelling but I doubt that many enterprise IT organizations will go to their local Fry’s to buy disk to store their organization’s crown jewels, its information, regardless of how redundant their RAID configurations are.
More likely, they will purchase reliable, cheap, fast disk from a company that specializes and supports this type of hardware. The market leader in this market today is a company that was once called Engenio. It was acquired by NetApp (NetApp CEO Tom Georgens used to be the CEO of Engenio) and is now sold as the NetApp E-Series. The E-Series provides backend disk to Oracle/Sun, IBM, Dell, Teradata, SGI, etc. So NetApp will be a big winner if Software-Defined Storage starts to take off (which I think it will).
When I worked at VMware, we frequently gossiped about Cisco buying EMC or NetApp or some other major storage vendor. It seemed obvious that, since Cisco dominates networking and it is making huge strides in computing with its UCS platform, their next step was to build or buy a major storage vendor. Then they would have a complete portfolio of network, compute and storage to supply the data center. Then they could build their own converged infrastructure.
A more likely scenario is that Cisco will introduce “network storage controllers” that integrate with their Nexus network architecture. So customers can continue using whatever storage array they prefer to store their inactive data but, to paraphrase a famous Scott McNealy quote, the network is now the storage controller.
Using tiering algorithms, these network storage controllers will direct high-demand blocks (or files) to the server (or compute node) that needs it. Cisco could pioneer this market if it wanted.
When I first started socializing these ideas to my clients about two years ago, it sounded a bit radical but now it is already starting to happen. For Unified Management, EMC’s Unisphere shows tremendous promise.
For Software-Defined Storage, Check out Violin Memory, Pure Systems, Virident/SanDisk, FusionIO, Nimble, Tintri. These and countless other startups promise some pretty cool technology in this space.
For Network-Centric Storage, Cisco has the potential to lead this segment. But there are others already making this happen. Just look at NetApp’s Flash Accel (partnership with FusionIO) or EMC’s VFCache. IBM recently acquired Texas Memory Systems so expect them to weigh in with a similar offering.
What do you think? Did I miss something? Am I completely misreading the tea leaves? Please leave comments. And, as always, thanks for reading the blog!