Experts predicted flash technologies will be so hot in 2016 that solid-state drives will cause the demise of high-speed...
hard disk drives and solid-state storage arrays will outsell disk-based systems.
Forecasts for 2016 on flash technologies also include predictions that 3D triple-level cell (TLC) NAND flash technology will go mainstream, storage-class memory will warm up and nonvolatile memory express (NVMe) will grow in importance.
Below is a sampling of predictions on flash technologies from storage industry veterans.
SSDs vs. HDDs
Marc Staimer, president at Dragon Slayer Consulting: People are going to get very excited over the cost and density of read-optimized flash drives this year when they're at 16 TB and 32 TB drives in 2.5-inch form factors. You're going to see, for all intents and purposes, the end of any significant sales of high-performance hard disk drives. And sales of SATA drives are going to slow down because of the read-optimized flash drives.
Jim Handy, semiconductor analyst at Objective Analysis: Solid-state drives are replacing high spindle-speed HDDs following a certain dynamic that we identified back in 2007. First, SSDs replaced arrays of six to eight high-speed HDDs based on very clear economics. Later on, it became smaller arrays of HDDs that were replaced, with an SSD taking the place of four HDDs in a RAID array. There are still lots of applications where a single, high-speed HDD is more cost-effective than an SSD, and this is likely to remain the case for a long while. So, the market for 10,000/15,000 rpm HDDs should continue for some time.
Bob Fine, director of storage product marketing at Dell: TLC NAND flash will become pervasive across the industry. The market share that 10,000 and 15,000 spinning drives have had will steadily decline over the next 12 months with that kind of cost advantage of TLC.
Flash storage technologies: System predictions
Vish Mulchand, senior director of storage product management and marketing at Hewlett Packard Enterprise: We went from storage customers telling us that investing in flash required justification to them telling us that purchasing spinning media requires justification. With the introduction of ultradense SSDs and systems capable of delivering truly enterprise-class resiliency, 2016 will be the year where all-flash array shipments outpace HDD-only.
Randy Kerns, senior strategist and analyst at Evaluator Group Inc.: The use of all-flash storage for business-critical primary storage will accelerate in 2016, with most enterprises having plans to have all primary storage on flash within three years. This will cause more examination of the data on primary storage and subsequent movement of less active data to secondary storage systems.
Rob Commins, vice president of marketing at Tegile Inc.: The NAND vendors -- the SanDisks, Microns [and] Samsungs of the world -- are putting two different product classes out there in their roadmaps. Some flash is going to be very low latency, high I/O-focused, and the other half is going to be what I call cheap-and-deep flash. Now, you're going to see serious offerings from the storage [system] vendors that are going to start leveraging this bifurcation of flash.
NAND flash technology
Handy, Objective Analysis: NAND flash prices will be flattening for the next year or two, as chipmakers struggle to make 3D NAND cost-effective. This will allow the HDD-SSD price gap to grow to eliminate any possibility that SSDs will replace capacity HDDs. In about three years, there will be a semiconductor price crash, and that will bring NAND prices closer to HDD prices, but SSDs will remain at least 10 times as costly as HDDs for some number of years.
IT managers are getting smarter about how to use NAND flash, finding that it isn't necessarily best to use it as an SSD or even as storage. Direct-attached NAND does a very effective job of reducing DRAM requirements in systems with shared storage. The industry will continue to migrate from SATA [and] SAS to PCIe, with the emergence of affordable medium-performance NVMe SSDs.
Ulrich Hansen, vice president of product marketing, SSD at HGST, a Western Digital Corp. brand: Flash will continue to expand during 2016 in both enterprise and cloud markets. Key application areas for 2016 will be conventional and NoSQL databases, and real-time analytics. Broader adoption of the NVMe protocol will establish PCIe as the SSD interface of the future for server-side flash deployments, and a spectrum of PCIe SSDs will emerge that spans from highest performance levels and capacities to entry-level configurations.
Alex Lam, vice president of enterprise business at Fujitsu America Inc.: NAND flash in the form of SSDs or PCIe cards is no longer [the] only game in town when it comes to fast storage. We will see a rise in storage-class memory (SCM) solutions as a high-end alternative to high-speed storage media for storage. While SCM will be niche in 2016, expect to see more use cases proliferate for this medium -- especially as NAND flash starts approaching its theoretical limits.
Scott DavisCTO, Infinio
Scott Davis, CTO at Infinio Systems Inc.: Storage-class memory will fundamentally change the industry even more profoundly than flash did. Its specs are orders of magnitude faster than flash, endurance is significantly better than flash and the cost per GB is lower than DRAM. It eliminates the asymmetric read [and] write quirks exhibited with flash, and most importantly, [it] will usher in an era of byte-granularity persistence. Storage-class memory will provide unprecedented storage performance and further drive the trend toward server-side storage processing, shaking up everything from storage economics to application design.
Handy, Objective Analysis: There's a lot of talk today about storage-class memory. Expensive DRAM-based nonvolatile dual inline memory modules are being used for software development, and Intel and Micron recently introduced their 3D XPoint memory, which aims to reduce costs enough to give SCM its rightful place in the memory [and] storage hierarchy. Over the next five years, though, I don't expect there will be enough software support for 3D XPoint to be used for SCM. Instead, 3D XPoint will make its mark as SSDs are doing today -- by reducing DRAM requirements in servers.
Flash technologies: NVMe developments
Dennis Martin, president at Demartek LLC: Look for NVMe to continue to grow rapidly in importance during 2016. NVMe will provide access to very fast solid-state storage in a variety of form factors and platforms, and by the end of the year, [it] may span from mobile phones to data center storage. For a glimpse into the future of storage networking, expect some very exciting NVMe over Fabrics implementations to come to market in 2016.
Greg Scherer, CTO at QLogic Corp.: NVMe over Fabrics will emerge as the transport of choice for SSDs as we enter 2018 and beyond. NVMe over Fabrics significantly improves SSD performance by eliminating the overhead of protocol translations and semantics used in traditional SCSI transports for spinning media. Fibre Channel and Ethernet can now communicate to NVMe-based SSDs natively using a much lighter-weight set of commands that greatly improve performance and reduce latency.
All-flash technologies were on the rise in 2015
Dell-EMC deal transformed the flash storage market
Top flash storage headlines of 2015