Categories: Analyst Blogs
Storage vendors of many sizes now offer all-flash storage arrays. Indeed, some of the dominant enterprise storage vendors now lead customer presentations with their all-flash arrays in new storage purchase discussions with customers. It’s the right thing to do strategically because, we believe that in three to five years, most if not all enterprise storage consumed for primary business applications will likely be some form of solid state. However, today’s storage marketplace is still split between buyers of solid state and rotating disk.
One group of buyers has come to the conclusion that all enterprise storage available from vendors will be solid state and that the cost of solid state, all cost factors considered, will be lower than rotating disk. Therefore they’re starting the conversion in their data centers now by replacing disk arrays with solid state.
Another group has yet to be convinced that their futures are all solid state. They believe that the cost of rotating disk on the basis of cost per unit of capacity will always be lower, therefore there will always be a use case for it. They are adding flash to their environments, but it is being used primarily to augment existing disk storage environments as opposed to replacing disk arrays. They are buying flash to add I/O capacity in order to accelerate specific applications and application environments—virtualized server based applications for example. The difference is not only in enterprise storage buyers’ strategic thinking about their futures. After talking to a number of enterprise storage end user professionals about the future of solid state in their data centers, it is clear to me that senior IT management is driven by two different cost appraisals.
The group that focuses solely on the cost per capacity of array-based storage adds flash to existing disk arrays and/or adds all-flash arrays to their data centers. These moves are taken to incrementally increase I/O capacity and to implement flash where performance is needed the most now. They will also be buyers of hybrid (flash plus disk) arrays.
The solid-state-is-our-future group works with a different—and broader—set of numbers. They include in their total cost of ownership (TCO) calculations the long-term cost of maintenance for example. Five-year, all-flash array warranties are becoming more common vs. the standard two and three year warranties for disk arrays. Solid state consumes less in data center environmental resources like power, cooling, and floor space. Adding-in these factors to both sides of the comparison brings the cost of flash more in-line with disk. Array vendors have also added compression and data deduplication in some cases to narrow the difference in cost for unit of data stored.
One thing both groups have in common however is not only the need for performance but a need to understand more exactly how, where, and how much flash can be used to accelerate performance at the lowest cost and with the greatest impact. They also have budget. The Flash Memory Summit coming up at in three weeks at the Santa Clara Convention Center will be a good place to start.