There was one common prediction when we asked a panel of technologists and industry analysts to weigh in on the...
future of enterprise flash.
Flash storage is a mainstream option, and its use will continue to escalate for active data as costs continue to drop.
"There really is no longer much, if any, debate about the inherent value and desirability or even relative reliability and longevity of flash," said Mark Peters, a senior analyst at Enterprise Strategy Group (ESG). "Instead the questions are now about the type, speed and extent of adoption. Simply put, it is not if flash will be adopted, but where and how soon."
New ESG end-user research showed flash storage is becoming widely understood and almost unanimously seen to be good and to represent the future. But, most of the surveyed organizations said they use flash storage only as a point solution where absolutely needed, and not as a storage tier.
"The only thing holding flash back from broader adoption is cost," Peters said.
Straight from the sources, here are the predictions.
Flash predictions: Adoption, pricing
David Floyer, CTO and co-founder, Wikibon: For active data, the disk drive in 2015 is fundamentally dead. The vast majority of active data is going to be on flash over the next five years. Flash has reached a critical point where it's cheaper for active data. The price is going to go down because there's lots of volume driven by consumer demand for mobile devices, thumb drives, and cameras. 3D NAND is in place. It's going to happen. The investment is being made in it. And each generation is going to be denser, cheaper and faster.
Jeff Janukowicz, research director, International Data Corp. (IDC): There is no doubt that flash is transforming the storage landscape. In 2015, we expect flash shipping into the enterprise to continue to outpace HDD growth, with enterprise SSD capacity shipped expected to increase by more than 75% year over year. Much of the growth will be attributed to pent-up demand for higher performance, lower latency solutions and lower prices, which will enable customers to leverage more flash in hybrid storage arrays, all-flash arrays, and directly attached to servers.
Tim Stammers, senior analyst, 451 Research LLC: Data center use of flash in all forms will increase, with adoption of all-flash arrays showing the fastest growth. The main reason will be the falling cost of flash. On a per-GB basis, enterprise flash drive prices fell hugely between 2012 and 2014, and a large reason for that was the move from SLC to MLC and consumer-grade MLC. That transition is in its end stages, which means that we may not see such steep price falls during 2015. But, increasing density of raw flash chips will help keep prices pointed downwards, and we will see enterprise flash products based on 3-D flash -- the next evolution of flash chip architecture -- keep per-GB prices falling.
Jay Kidd, CTO, NetApp Inc.:I've heard people talk about the notion that the data center will go all-flash, and I believe that is utter nonsense. The future of all-flash array companies is not all-flash. We foresee the gap in cost between the least expensive flash and least expensive hard drive staying at about a 10 [-times] difference in cost through the end of the decade. Not all of customers' data is valuable enough to them at any given point in time to be held on flash, which will be very expensive storage. There will be a combination of flash and disk in the enterprise data center, and the successful companies will be those who can seamlessly and easily blend those two together. The combination of flash and disk is the future of enterprise storage.
Andrew Walls, IBM Fellow, CTO and Chief Architect Flash Systems: The flash tipping point is really happening and accelerating. Many large data centers will be switching to all flash for active data in 2015 because the TCO strongly favors flash. And in fact, the price per GB does too, when taking into account data reduction. I also think we'll see more enterprises have one set of storage do all applications instead of having one set for VDI, one set for OLTP, etc.
Flash technology predictions: Chips, drives
Bob Brennan, senior vice president, Memory Solutions Lab, Samsung Semiconductor Inc.: We believe that, moving forward, flash itself will become a multi-tiered technology. You'll have a superfast flash in PCI Express. Then, you'll have bulk flash -- write once, read many, also called WORM flash -- which will be SATA and SAS [drives]. After that, you'll have your tier of disks for archival or cold storage. In this way, you'll have a hierarchy of disks, which go from DRAM on the very top of the stack to PCIe [Non-Volatile Memory Express] NVMe SSDs at the next layer, to SAS and SATA at the next layer.
George Crump, founder and president, Storage Switzerland LLC: The continual decline in the cost of flash will be driven primarily by 3D NAND, which both Samsung and Intel will be delivering probably in the latter half of 2015. The second thing is triple-level cell (TLC) NAND, which means we're writing three bits per cell instead of two bits with MLC and one bit with SLC. Obviously, that'll drive down price, but TLC will wear out faster, so we might see the emergence of a flash archive tier or something like that. At the opposite end of the spectrum, I think we'll see a lot of work in non-volatile memory (NVM) in the 2015 timeframe.
Some of the hybrid array vendors might be able implement a system that leverages MLC flash on the front end and TLC flash as their tier at rest. And then going forward, you might even see some guys do non-volatile DRAM as their front end and MLC or TLC as their backend tier. So, I think we'll start to see a tiering of memory-based storage starting in 2015, which will be really good for the industry.
Floyer, Wikibon: By 2016, flash drives are going to be bigger in capacity than the traditional magnetic disk drive. You're going to have 16 TB performance flash drives and probably 10 TB disk capacity-centric drives. Flash drives are going to last for at least 10 years. Pure Storage uses MLC, the cheap stuff, and they guarantee their drives for 10 years. Everybody is going to have to guarantee them for 10 years. Replacing drives will be a thing of the past. You're just going to put them in and leave them. I'm predicting that nobody will need to replace flash drives. Old flash and old controllers will be repurposed in the data center, and software and hardware maintenance will be much lower than current disk drives.
Jim Handy, semiconductor analyst, Objective Analysis: Samsung will be joined by most of its competitors in the 3D NAND SSD market. Despite this, 3-D will account for a very small portion of overall NAND flash shipments, as the technology defies attempts to move it to mass production [and] 3-D involves semiconductor processes that have never been used before. This means that they have to learn how to control them before the chips will behave predictably.
Flash will move from today's tightness to shortages severe enough for chip makers to be driven to decide who will receive product and who will be left without. This will put the squeeze on some flash array and SSD makers, forcing them out of the market. Meanwhile, SSD prices will not drop appreciably over the course of 2015.
The Diablo, SanDisk, IBM approach of Memory channel flash will gain widespread adoption because the performance of this technology is unencumbered by the delays typical of conventional I/O channels.
Jon Bennett, co-founder and CTO, Violin Memory: 3D NAND will be made available to storage subsystem companies from the flash semiconductor suppliers in late 2015. This will result in implementations by SSD vendors in the 2016 timeframe and incorporation into storage subsystems and availability in 2017.
Now that the industry has had more than a year of triple-level cell (TLC) in consumer-grade SSDs, we can expect the enterprise SSDs to get economic benefits from TLC NAND. Initial rollouts may see higher-than-expected prices due to a higher amount of overprovisioning by SSD vendors.
Flash-based system predictions
Marc Staimer, president, Dragon Slayer Consulting: I expect to see usable flash capacities -- before deduplication and/or compression -- ranging up to 500 TBs in a single rack unit before the end of 2015. As for storage systems, differentiation among all-flash arrays and hybrid storage systems in 2015 will be in the storage software capabilities they offer, application integration and capacity per rack unit.
John Hayes, co-founder and Chief Architect, Pure Storage: It's now generally accepted by storage vendors that AFA is the new Tier 1. Storage-end users have been more hesitant, and AFA products don't yet cover every use case. In 2015, these adoption barriers will improve. But, perhaps a more interesting change will be the movement towards implementation of risk sharing between storage vendors and customers. This dynamic will ensure storage vendors act more like service providers, and storage vendors and their customers will form a critical and long-term relationship.
Robin Ren, CTO, XtremIO, EMC: We expect to see next-generation agile data centers designed around the use of all-flash arrays. This technology won't be about making what you have run faster as much as it will be about enabling the creation of new business processes and competitive advantages that don't exist today.
Floyer, Wikibon: We're going to see a completely different way of managing active tier-1 data in the data center. What you want to do is get rid of the processor in the traditional storage array and move that to the server. You have lots of cores in the server, and they can look after the storage. The flash will be much closer to the processor, and the data will be close to the application. The really interesting thing will be what's going to happen to new applications that come along. At the moment, disk drives are the bottleneck to application design. With flash, you remove the biggest possible bottleneck in the whole application design.
Tracking the rise of all-flash array marketing
Considering the role of enterprise flash in a storage system