Flash storage startup Fusion-io doubled capacity with its new product this week, and established storage giant EMC took to the stage at Oracle OpenWorld to give the audience a look at its server-side flash offering, Project Lightning. Following a flurry of investment activity over the summer, flash storage remains a compelling and potentially disruptive piece of the enterprise storage puzzle. How will incumbents respond to new entrants such as Fusion-io, and what do these developments mean for customers looking to incorporate flash into the way they store data?
Solid-state drives (SSD) have been available for decades, although cost and perceptions of unreliability and short lifespan meant they struggled to compete in an enterprise data storage market saturated with low-cost disk-based storage. However, SSDs are also fast and energy-efficient, which has led to their use in premium consumer devices (such as the MacBook Air and its competitors) and the exploration of how the positive characteristics of SSD can be incorporated into existing storage models without raising costs or reducing lifespan. In recent years, technical innovation and improved production processes (caused, in part, by the demand for flash memory to fit smartphones, tablets and energy-efficient laptops) have led to enhanced reliability and a fall in cost. SSD is not yet as cheap as a hard drive, but the gap has narrowed sufficiently for a number of startups to begin delivering products in which SSD plays an increasingly significant role.
Two startups — Pure Storage and SolidFire — have gone as far as to deliver big storage arrays full of SSD (with no hard drives at all). They argue that it’s possible for the total cost of ownership for their array to prove comparable with functionally equivalent disk-based storage. SSD now deserves serious consideration when customers are looking to purchase new storage hardware, especially in environments where space or power are limited.
Established storage providers such as HDS and EMC also recognize the potential of SSD, although their approach tends to treat SSD as a “Tier 0” within a multifaceted storage solution. Tier 0 stores rapidly changing and valuable data and uses the speed of SSD to deliver it quickly on demand. Lower tiers use increasingly cheap (and slow) disk drives. They may also include the lowest tier of data storage, which is actually the cheapest of all: data stored offline (and possibly off-site) on tape.
The Project Lightning card shown by EMC at this week’s Oracle OpenWorld is not intended for use within a large storage array. Instead, it is meant to be installed directly into a server, bringing the data much closer to the applications that require it and shaving fractions of a second off processing time by removing the need to transfer data from a physical storage array to the server.
Regardless of the specific merits that one product may have over another technically, startups such as Pure Storage and SolidFire face an uphill struggle to see their products widely deployed. Storage infrastructure lies at the heart of how any big organization manages its data, and any cost savings or performance improvements will need to be demonstrably significant before a CIO will consider ripping out so much of a company’s infrastructure and trusting her business to startups that lack a track record. EMC doesn’t have this problem, as the company is already well-known, respected and trusted. Its approach fits alongside existing infrastructure and ways of working. Fusion-io faces less of a problem, as it supplies hardware that is then used by third parties (startups like Nutanix, as well as established storage or server vendors like HP).
But for startups advocating use of their all-SSD storage solutions, the surest way to critical mass may be acquisition by an established storage vendor. Comprehensive adoption of radically new technology, sold by a startup, may simply be a step too far for cautious CIOs who would rather wait for a comparable solution to be developed — or acquired — by their existing storage providers.