The enterprise flash juggernaut is gathering steam in 2013, leading many analysts to question not whether it will be implemented on core systems but how rapid and thorough the transition will be.
For the most part, it seems that flash and mechanical storage are facing off in an uneasy truce. Traditional media – disk and tape – will still make up the vast majority of bulk storage, even on the cloud, while solid state tiers will hold the most current and frequently accessed data. The ramifications of this change to enterprise infrastructure will be significant, but for the moment most eyes are watching how this will affect some of the storage industry’s leading players.
For industry watchers like UBS, the most pertinent question is who will be the first to integrate flash directly into their storage platforms. Firms like EMC and NetApp have been toying with various flash modules and add-ons, mostly for high-speed caching applications, but none have gone for an all-flash array just yet. Word on the street has it that such systems could be in the channel by year’s end, but no matter who is first out of the gate, disk storage firms like Seagate and Western Digital could find themselves behind the curve very quickly without flash solutions of their own.
Clearly, there is no shortage of industry maneuvering in the flash arena as top vendors seek to position themselves for a radically changed market in the next year or so. With IBM buying Texas Memory Systems and Dell signing up with Skyera, it seems that no wants to be left without a flash solution in this digital version of musical chairs. The latest tie-up has Toshiba and Violin Memory collaborating on a new generation of PCIe cards that aims to bring the cost of flash closer to that of hard disk. Already, Violin’s Velocity family comes in at around $3 per GB, and the hope is that tighter integration with Toshiba’s hardware and software platforms will cut the price even more.
For on-server deployments, EMC’s new XtremSF card is the latest offering that accommodates high-performance applications over the PCIe bus. The device delivers more than 1 million IOPS and is available in capacities up to 2.2 TB in eMLC configurations and 800GB in SLC. The card is supported by the XTremSW Cache software, built largely on the VFCache acceleration platform, but updated to work with both multi-layer and single-layer designs. Look for virtualization support and memory and DAS management tools to be added to the package later this year.
Deploying new flash tiers is the easy part, however. Much more challenging is the need to ascertain what data belongs on which tier. That’s the focus of a recent announcement that LSI’s Nytro WarpDrive PCIe card has been certified for NetApp’s Flash Accel software. The move creates an integrated caching solution designed to automate the placement of hot data used by critical business applications into high-speed flash storage. The solution is said to boost latency and throughput by 80 percent to 90 percent while at the same time lower cost per IOPS and cost per GB through reduced power and cooling costs and a smaller physical footprint in the data center.
In any great technological transformation, observers try to gauge whether it is disruptive or complementary to the current state of affairs. At the moment, flash seems to encompass a little of both. Without doubt, flash provides faster performance than spinning media, although it still lacks many of the management features that traditional SAN and NAS architectures provide. And as I mentioned, bulk storage will always gravitate toward the lowest cost denominator.
For the vast majority of enterprises and the vendors who serve them, flash will most definitely upset a few apple carts, but only if there is strong resistance to change. By embracing flash and integrating it into existing platforms, users and suppliers will find themselves in good position to capitalize on its higher performance while still providing a touch of the familiar for those who don’t want to change too quickly.