A long time ago (at least from an IT perspective) flash made the jump from consumer memory sticks to the enterprise and the storage industry has never been the same. Thanks to technologies like 3D NAND we’re now seeing solid-state drives at 4TB, and soon to approach double-digit capacities. Even the lowly thumb drive that’s given away at trade shows is 4GB or more.
Based on the trajectory of non-volatile memory storage technology in the enterprise there’s a lot to learn about and the Flash Memory Summit (FMS) in Santa Clara (the 10th Anniversary) is the place to get that education. A new SSD interface and the remarkable growth of all-flash arrays are just two of the many reasons to go.
NVMe is a host controller interface designed to maximize the potential of non-volatile memory storage in storage and compute systems. Developed by a consortium of about a dozen companies that manufacture solid-state devices, storage and systems, the objective was also to provide an interface standard that would shorten time to market and reduce development costs.
NVMe provides lower latency than traditional SATA or even PCIe-based SSDs, accessing flash as a memory device instead of mapping it to a disk drive protocol. It leverages a streamlined command set and dramatically expands the command queue architecture to increase scalability, plus provides support for security and data protection functions.
While it uses the PCIe bus, NVMe drives are available in a 2.5” form factor that supports the new SFF-8639 or “U.2” connector type making it as easy to replace as a traditional SATA drive. NVMe also promises to deliver better flash endurance, lower power consumption, dual-port operation and other features central to enterprise storage design.
The NVMe standard supports all major platforms, including Windows, Linux, VMware and several UNIX variants. If you’re interested in this new technology, there’s and all-day track at FMS on NVMe and PCIe SSDs, plus eight other forums, keynotes and sessions that feature NVMe.
When they first appeared a few years ago, all-flash arrays were pretty far out of the mainstream. Limited to “bleeding edge” applications that needed high performance at any cost, you could count the number of viable use cases on one hand – and the number of vendors on the other. Things have certainly changed.
Effective per-GB costs have come down, due to cheaper NAND flash and improvements in data reduction technologies. Controllers have improved to better support low latency storage and accommodate the flash write and erase processes. And applications are routinely being written for the kinds of high-IOPS performance that non-volatile storage affords. All this adds up to an increase in the number of applications and use cases for all-flash arrays in the data center.
On the vendor side it’ll now take all your fingers and toes, and then some, to count the companies offering an all-flash array. In Evaluator Group’s current Solid State Storage Evaluation Guide we list more than a dozen vendors’ products, those designed specifically for flash storage and those adapting existing storage designs for flash media. But in addition to the major storage players and ‘established’ independents, there are a dozen or more up and comers joining this space, many of which will be exhibiting at FMS.
While their use cases are getting more mainstream, the specs for these systems are anything but ‘middle of the road’. Scale-out capacities for some reach into the PB range with performance well into the millions of IOPS. Appropriately, FMS has close to a dozen forums, sessions and keynotes that feature enterprise flash storage systems and applications, including all-flash arrays.
Many products have long lists of features that sound the same but work very differently. It’s important to think outside of the checkbox of similar-sounding features and understand how technologies and products differ.