EMC's all-flash benediction: Turbulence ahead
Who's ready, who's going to have to buy in?
Analysis A rising tide lifts all boats and the blessing EMC gave to all-flash arrays at EMC World will be gratefully seized upon by a string of established small players and start-up wannabees such as Pure Storage and SolidFire. But this tide could carry some boats into turbulent waters.
Shared solid state storage has been the preserve of a niche market, one where customers need an in-memory database for their financial arbitrage trading. Seconds, even fractions of a second, count when trying to make money from fast but small changes in market prices for financial instruments or whatever else can be traded on Wall Street or other exchanges. Disk latency is the enemy of such trading applications and DRAM and flash arrays from Texas Memory Systems (TMS), such as its RamSan products, have been the classic storage array product for the company, where lightning fast access from a large data set is the key to an application's success.
Another application area is national security, where the fastest-possible database lookup is needed by communication monitoring systems – whisper the Echelon name – and border control traveller identification applications. Now there are enough customers wanting such applications that EMC is going to offer all-flash VNX and VMAX arrays as standard configurations. The trend for primary data to migrate off hard drives onto flash drives is starting, creating more business for the established flash array players, opportunities for start-ups and problems for the hard disk drive array vendors.
We're going to cast our eye over the field and review the suppliers in it, and offer opinions about what how the established storage array suppliers will react. This is an area looking prime for accelerated development and events could happen quickly.
Shared solid state storage suppliers
TMS initially introduced DRAM-based RamSan products, hence the "ram" in the name, and then added a flash line of RamSans alongside the faster and more expensive DRAM-based ones. It continues to lead the area and its RamSan-630 has just posted whip-cracking SPC-1 performance numbers and price/performance numbers. No other supplier comes close in the SPC-1 rankings to TMS' product performance, for the moment. We think an EMC VMAX all-flash SPC-1 Kaminario makes the K2 DRAM product, for the same market areas that TMS supplies with its DRAM-based RamSans, in-memory database screamers. It is a mid-to-late stage startup, having announced a $15m C-round of funding in May. A flash version of the K2 technology is said to be in development.
- DataRam produces the XcelaSAN array, a DRAM device, which sits in front of SAN-attached storage arrays and acts as a large front-end cache.
- Nimbus Data has tried the all-flash route with multi-level cell flash since April last year with its S-class unified storage array. There has been an absence of independently validated performance data but a second version of the product has been introduced and Nimbus' pioneering might well pay off.
- Solid Access went for a UNAS 100 all-flash NAS product, eschewing block access. That could be a useful differentiator in the months ahead.
- Violin Memory has focused on shared flash arrays and is building a great relationship with HP as that company sells against Oracle's Exadata box. Both Toshiba and Juniper have invested in Violin, demonstrating that they see potential there which they could potentially exploit. There is an air of suppressed excitement coming out of Violin, a feeling that a tipping point is coming with sales about to rise substantially.
- WhipTail has focused on serving virtual desktop images from its Racerunner Virtual Desktop XLR8r, which uses in-line deduplication.
Early stage start-ups
- An early-stage start-up called Pure Storage was introduced earlier this month. The company is punting an all-flash array technology concept to hold enterprise primary data with a bulk disk store behind it. Little in detail is known about its technology architecture and product is likely nine to 18 months away.
- SolidFire is another very early stage start-up, based in Atlanta, with development taking place in Boulder, Colorado. It gained $11m in an A-round of funding in February and its pitch is that primary data should be stored in solid state storage with an array controller designed from the ground up for flash and not disk. Its target market is a cloud storage service supplier.
In blogs posted on the company's website the point is made that disk-based SAN array storage controllers can manage thousands to tens of thousands of IOPS whereas SSDs can deliver hundreds of thousands to millions of IOPS. Disk-centric controllers can't keep up and bottleneck systems.
A flash array controller needs: "An architecture built from the ground up around SSD technology that sizes cache, bandwidth, and processing power to match the IOPS that SSDs provide while extending their endurance. It requires an architecture designed to take advantage of SSDs unique properties in a way that makes a scalable all-SSD storage solution cost-effective today."
What kinds of controllers does it think TMS, Violin Memory and other shared SSD storage suppliers have been building?
SolidFire talks a good talk and we'll have to see how well it walks the product walk, how well it lets "cloud providers guarantee sustained performance to thousands of servers from a single storage system", when its technology becomes visible, possibly towards the end of the year.
El Reg imagines that EMC has that sort of vision in mind with its all-flash VMAX. Will the tide of EMC's all-flash array vision float SolidFire's boat or smash it against the rocks? Or will it be acquired?
What will established suppliers do?
Let's cast our eye over the established storage array players and ask what they will do in the face of the primary data, all-flash array tide washing their way.
- BlueArc does not have an all-flash offering yet is focused on high-performance NAS boxes with its hardware acceleration technology. It will find its hard-won position towards the top of the NAS performance tree contemptuously over-taken by all flash-array suppliers – like VNX flash – and will have to develop or acquire its own technology – like UNAS NAS – not least to preserve its HDS reselling deal. El Reg plumps for an acquisition as time could be short.
- Dell does not have its own technology and is quite willing to acquire storage technology these days. We vote for Dell acquiring the technology and acquiring it with relish and speed as it sees another way to take storage share from its competitors.
- Fujitsu is building up its Eternus disk array line and now will see the primary data storage needs of its customers going to other suppliers unless it can provide all-flash arrays that do the job. If it sees enough time to do then it will probably have the acquired Fujitsu Siemens Computer storage people do the job. If it doesn't, then buying in technology is the only alternative.
- HDS does not have its own in-house flash array technology and faces losing high-performance array business to EMC, TMS and the startups unless it gets it. Not known for its fast-reaction to new market developments but noted for its focused acquisitions – Archivas – and reselling deals (BlueArc), HDS could look to a reselling deal as its first response and might do a deal with an established player like TMS or Kaminario.
- HP, El Reg can imagine, may be having some trouble with the idea of another storage technology it has to get to grips with before it has even fully integrated what it already has. Buy baby, buy: it's the short cut to getting a product line and technology you can integrate. We think a promising late-to-middle stage start-up with great technology will likely be in HP's sights, one with product that HP's channel can sell straight away. Is that the sound of violins we can hear?
- IBM once again finds its famous research capabilities have missed the boat, and the company will have to dip into its wallet and develop its own technology, producing say an all-flash V7000, an all-flash SVC, or pulling a rabbit out of its DS hat with an all-flash DS8000. Alternatively it could buy its way in to the technology, a much faster option that it has employed before. That's what we think it will do, and like HP go for a mid-to-late stage start-up with product that Big Blue can pump through its channels – or an established niche player like Kaminario or TMS.
- Now to everyone's favourite storage puzzler, NetApp and its likely choices. We think that NetApp could be one of the traditional suppliers Solid Fire has in mind, a classic disk array storage controller company with a singular focus on disk drive I/O problems. The company could surprise us with its own technology development, unlikely we think. But perhaps an all-flash array technology is buried in its Engenio acquisition. LSI is no slouch in anticipating technology developments and is involved with Seagate's SSD initiatives. El Reg thinks an Engenio all-flash technology could emerge and, if it doesn't, then NetApp will buy an established player or mature-ish start-up.
- Oracle, with its flash-experienced Sun engineers, could readily build Exadata-type all-flash systems. We don't foresee an acquisition but, hedging bets, don't rule one out.
Data protection suppliers and wily disk dogs
There are other array suppliers who face the same problems and come from a data protection background. We're thinking of Overland Storage and Quantum. They might well buy an early-stage startup, cheaper than an established player. Overland, with the fertile mind of Geoff Barrall in play, could even pull its own all-flash technology out of its development labs.
- Both Pillar Data and Xiotech have experienced disk array people involved: Mike Workman, Pillar's CEO, and the Richard Lary/Steve Sicola team at Xiotech. Both companies have solved the frailties of disk drives in their own way to produce high-performance systems and both are late-stage start-ups hopefully transitioning towards profitable independence from their funding agencies. Both are also faced with reacting to all-flash arrays as JBOSS boxes – Just a Bunch of Solid State – can do the high-performance primary data storage of their products better, in raw performance terms, than their own carefully-designed technology can.
El Reg thinks Pillar needs to provide its own all-flash Axiom, and our intuition tells us it will build its own technology. Xiotech needs to do the same. JBOSS boxes threaten its storage blades' success in the primary data storage market and Xiotech needs an all-flash ISE (Intelligent Storage Element) – which will turn it into a direct competitor to TMS, Violin and every other provider of primary data solid state boxes, thus destroying its differentiation. It will have to find new ways of differentiating itself from the main group of solid state primary data array suppliers.
- Lastly there is Huawei-Symantec, a quietly very ambitious storage supplier whom, we think, will produce its own technology or source one from China. Maybe it will surprise us all and buy a US start-up though. We think that's less likely. Fellow Chinese storage supplier UIT could go the same route.
Totting up, we have six to seven potential acquirers, four mid-to-late stage start-ups or mature suppliers (TMS), and five early-to-mid stage start-ups. Will any of the start-ups make it and become mature, independent businesses? El Reg thinks it unlikely.
We reckon that there will be a host of all-flash array technology company acquisitions over the next 12-24 months as EMC's competitors react to its forcing of the market pace, and they scramble to catch up. EMC has opened the floodgates and, by the end of 2015, the idea of primary data being stored on spinning disk will seem simply archaic. ®