The solid-state storage market is still a work in progress. A prudent approach will ensure that a flash deployment fits well with an overall storage strategy.
Over the past year and a half, the market has been flooded with products designed to solve storage performance problems. Taneja Group has recently completed field research with several hundred storage practitioners regarding storage performance -- specifically what products and technologies enterprises are looking at to enhance data storage performance.
It was clear from this exercise that IT priorities remain pretty straightforward. In the midst of all the buzz about cloud, server virtualization still rules the roost. After server virtualization, a basketful of storage infrastructure initiatives surfaced, followed by client virtualization. Cloud infrastructure was a distant fourth, with the prevailing impression that cloud is still too young, distant and/or opportunistic for it to be on the IT strategy radar.
As we dove deeper into the study data, we found this set of IT priorities aligned fairly well with where these IT pros said they were experiencing storage performance challenges. Many reported storage performance issues, with server virtualization, client virtualization and data warehousing lining up as the prime suspects. Interestingly, storage acceleration still doesn't rank very high as an independently funded IT priority. This suggests that despite the proliferation of performance-oriented solutions, users still don't have a firm grasp on how they're going to get a much-needed storage performance boost.
We also see users actively seeking solutions, but where their storage dollars will end up is not yet entirely clear. For instance, some are focusing on server-based flash offerings using SAS or PCI Express (PCIe) flash as direct-attached storage. That type of solid-state implementation has received a lot of attention with Fusion-io's success, and its big customer wins like Facebook. But interest hasn't necessarily translated into action: We see far fewer users who have concrete deployment plans.
The takeaway is that a server-based flash deployment looks like a viable solution for specific problems, but a clear path for integrating those products with an overall enterprise storage strategy is still lacking. Deploying server-based flash presents its own challenges, as it fragments storage management, requires the installation and maintenance of multiple devices, and a lot of storage pros are unwilling to give up traditional storage features for the sake of performance. For the typical enterprise, server-based flash remains a narrowly deployed solution that is only installed when there's a great need for more performance. That need may be defined by specific applications, or it may be a relatively inexpensive fix when there's not enough budget for a broader solution.
Meanwhile, those who like the idea of server-based flash may be convinced to take another route to solid-state performance that would either enable acceleration without breaking enterprise storage strategy or be more easily deployed without having to crack open a bunch of servers.
Companies leaning toward those approaches are likely to meet up with two types of vendors. Server-based storage accelerators (SSAs) promise to cache data while keeping the authoritative data centrally stored; this provides tremendous acceleration without breaking the enterprise storage strategy. Examples of those vendors and their products include EMC with XtremSF, Marvell and its DragonFly products and QLogic's 1000 Series FabricCache adapters. Software vendors may also be able to deliver required performance improvements. The list of software vendors in this space is fairly long, but some examples are Intel's Cache Acceleration Software (CAS), Proximal Data's AutoCache and SanDisk with FlashSoft. There are many other small vendors active in this space, and even VMware is working on an offering.
Popular alternatives to server-based flash technologies are hybrid storage arrays that accelerate storage performance using a combination of flash and hard disk drives. Most users recognize the difference between traditional arrays that "up-tier" hot data to solid-state media, and newer-generation arrays that more seamlessly interweave flash with hard disks by using flash as the primary storage and de-staging data rather than using an up-tiering strategy.
The net-net is that it's still very early stages for the solid-state storage market. We still lack a clear understanding of how to identify, quantify, value and address storage performance issues, but at the same time the performance pressure on the storage infrastructure is growing. Some early adopters hacked together short-term solutions that may not fit well with their enterprise storage strategy or be viable longer term solutions. If you're up against storage performance demands, you may want to dig a little deeper into what the solid-state market has to offer so that your chosen path will be more strategic and manageable.
About the author:
Jeff Boles is a senior analyst at Taneja Group.