The rate of adoption of flash in the enterprise storage systems market today points to not IF flash will take but rather WHEN and WHERE. When considering the growing use of flash in both hybrid and all flash arrays, there are four key drivers:
A declining $/GB. On a blended average basis enterprise SSDs are expected to go from $1.27 per GB in 2014 to $0.27/GB by 2019. This decline of 24.7% on a compounded basis, influenced by factors such as overall NAND supply, the increased use of 3D and TLC NAND and specific type of NAND flash used in the SSD. Systems vendors are augmenting the $/GB declines with use of a range of NAND technology from SLC to MLC to eMLC. This allows systems vendors to find the right balance between cost, endurance and performance and incorporating IP to accommodate a given architectural tradeoff.
Use of efficiency and optimization techniques. The use deduplication and compression made disk, rather than tape, viable for backup and operational recovery. The same is true with flash, and with enhanced techniques for managing copy data, which constitutes up to 60% of all enterprise data, even more efficiency can be gained on modern all-flash platforms. The effect of efficiency approaches with flash memory not only reduces capacity consumption but also aids in reduced wear. And external Systems suppliers have breached the effective $2/GB threshold continuing to make flash more attractive for a broader set of workloads.
Market transitions. Cloud services play a role in nearly every data center today. The result is a lot of bidirectional movement of workloads (and data) between on premise and cloud locations, or between cloud locations. This increases the stakes of infrastructure suppliers to land or protect the customer footprint and can results in strong discounting at the system level. And cloud service providers, both SaaS and IaaS providers alike, are also moving quickly to flash.
Immediate IT. In a society of immediate gratification, the same is now true to IT. Users will not wait for a page to load, DevOps won’t wait to release code and business units won’t tolerate sluggish application performance. All this leads to fact that performance often trumps all. The use of flash started out small, as a cache or tier representing 5-7% of the overall capacity footprint. That number keeps increasing and when I speak with hybrid storage users, many expect flash to reach as much as 50% of their overall capacity footprint in the next 2 years.
IDC data serves to illustrate that Flash is taking over the data center. By 2019, IDC projects that less than 25% of the external storage systems spending will be consumed by all HDD-based arrays while 75% of spending will be on hybrid and all flash arrays. Will spinning disk ever go away? Probably not in the next decade or two, but its role will certainly change.