Vendors of enterprise storage arrays (EMC, HDS, HP, IBM, NetApp) are all riding the Software Defined Storage (SDS) bandwagon to one degree or another. IBM says that its enterprise storage product line will over time become increasingly software defined. EMC’s ViPR SDS platform is now a flagship product while HDS (SVOS), HP (StoreVirtual) and NetApp (FlexArray) are positioning offerings in this space.
Despite the rush to market, defining SDS is still a matter of public debate and is by no means a settled issue. The Storage Networking Industry Association (SNIA) offers something that is less of a concise definition and more of a long list of attributes. But pinning down a definition has little value here. What really matters is whether or not customers seevalue. And here, the jury is decidedly still out.
Consider EMC’s ViPR that doesn’t actually store a bit’s worth of data. Rather it forms an overarching virtualized storage layer above existing storage boxes—generally the large ones that actually contain the enterprise’s data. ViPR software allows administrators to apply a growing list of data services and management functions across the arrays it can “see” below. And it’s important to note that these arrays are already running their own data and storage management services that are defined in software as well that customers are generally paying extra for.
So here’s the big and as yet unanswered question: Will customers see value in SDS? Whether or not customers see value as these products reach them will depend first on what they do that is new and that they can’t already do with what they have. It’s not likely that they will turn off the array-based features they’ve already paid for, at least not right away. Indeed, vendors of SDS are wisely trying to avoid forcing their customers to make that choice. After all, array-based software has been a cash cow for years. However, based on what I’ve seen before, there are in fact some things that can be done using SDS as an over-arching virtualization layer that can’t now be done with array-resident software. These include:
Second, new and unique capabilities by themselves won’t be enough to get SDS to measure-up to the bandwagon noise. Potential customers still need to know how much they’ll pay, and how they’ll pay it. One common method seen so far is for vendors to charge on the basis of some unit of stored data—a GB for example. Fair enough. They have to make money somehow with this. But this is a new space for both customers and vendors who are feeling their way around here. Charging on the basis of GB stored seems logical until you consider the face that these products don’t actually store anything. They interact with the actual storage devices that customers are already paying for. And these customers are used to paying up for things that store mission-critical data vs. those that store big volumes of secondary data. Customers tier their storage environments and expect vendor pricing models to reflect this reality. However, what I’ve seen so far from the SDS vendors makes no such distinction. From a pricing perspective, all data is of equal value and this I believe will inhibit sales.
As time progresses, one of the promises of SDS is to allow customers to run software defined storage services on commodity hardware. In theory, this gets customers out of having to pay for the big expensive storage boxes. But the pricing and value-for-investment questions have to be answered first. And in realty, vendors are a considerable distance from giving customers the ability to throw away those cash cow arrays.