We here at Pure would be remiss if we didn’t offer a few thoughts on NetApp (NASDAQ:NTAP) announcing plans to enter the all-flash array category as Project Mars morphs into the FlashRay sometime in 2014. After flirting with a couple of potential acquisitions, NetApp ultimately stayed in house—putting some of the WAFL band back together to tackle the impending shift from mechanical to solid-state storage. While NetApp may be a bit late to the all-flash party, their market position in disk arrays and their multi-petabyte history in flash caching, makes their entry notable.
But what’s far more notable is that NetApp appears to be targeting the same all-flash storage recipe that Pure Storage first brought to market in 2011 within our FlashArray. EMC of course is following a similar path, with expectations that they will launch Project X (dervied from the XtremIO acquisition) sometime later this year. (With FlashArray and FlashRay now taken, I wonder what the EMC naming gurus will cook up.) While today there are a confusing variety of flash startups in the marketplace, yesterday’s announcement confirms that the storage market leaders believe the holy grail of next-generation performance storage follows the Pure recipe:
- Hardware comprised of commodity components (standard MLC SSDs, CPUs, and interconnects), combined with
- Storage software purpose-built for flash from scratch, providing
- Submillisecond, inline data reduction (deduplication and compression);
- Flash-optimized RAID6-style protection (no reliance on data mirroring or RAID10 schemes that don’t make sense on flash);
- Enterprise high-availability (HA) – self-healing from hardware/software failures without mirroring entire systems;
- Online-expansion and non-disruptive upgrade (NDU);
- Scale up and scale out; and
- Comprehensive data management, including ECC, snapshots, encryption, and so on.
Pure Storage was founded in 2009 to pursue precisely this recipe. Our belief then and now is that what is holding back flash adoption in data centers is price and plug compatibility with existing storage infrastructure. The above formula is what allows Pure to offer all-flash storage that’s a direct replacement for traditional disk-centric storage arrays (typically from EMC and NetApp) at the same or better price (in terms of $/GB stored), but >10X faster, 10X more space & power efficient, more reliable, & 10X simpler. No surprise that such a “no brainer” value proposition is going to spur others to strive to reproduce the recipe. The challenge is actually getting the product implementation right. Other vendors have spent years attempting to reproduce the offline dedupe functionality of NetApp, and now it is NetApp’s turn to try their hand at the more complex challenge of inline, submillisecond data reduction. For many of the flash upstarts reproducing this recipe will require dramatic changes to their technology and teams. We don’t expect more than a handful of vendors to ultimately get this right: while there are a great many aspirants, implementing the above represents person centuries of effort of the very best software engineers, and there is so much more work to be done beyond this baseline (although we’ll not shed any light on our roadmap just yet). So the race is on to see which of the incumbents and new storage ventures can deliver products based on this formula, and how far we will have been able to progress Pure’s FlashArray by the time they get here.
The other compelling takeaway from the news is that for next-generation performance storage, all flash trumps hybrids of flash and disk. When the two market leaders in caching and tiering both get on the 100% solid-state bandwagon, something intriguing is going on. As we have remarked elsewhere, caching is great when you’re expecting disk performance and hit flash cache, but is wholly unacceptable once you expect flash performance and have to wait on mechanical disk. The same shift in expectations ultimately doomed Virtual Tape Libraries (VTL) and Hierarchical Storage Management (HSM). And when Data Domain employed inline data reduction to make disk backup appliances price competitive with tape, the market voted with its feet. The storage market leaders in EMC and NetApp have now both endorsed the view that filling up a disk-centric VMAX, VNX or FAS array with flash just cannot deliver on the promise of solid-state. In retrospect this should be obvious: no one ever suggested that Data Domain should have started with the software from a physical tape library for building their disk appliance, and yet tape has more in common with disk than disk does with flash.
We at Pure sincerely welcome the NetApp announcement as we welcomed EMC’s before it. Competition is essential to growing the solid-state storage market, and we always expected EMC and NetApp would compete with us in the category. All a new company like Pure can ask for is a 2+ year head start and the big guy’s endorsement of the recipe we originated, meaning the competition is coming to our playing field. We now have both (2 years, after all, is a long time in tech), and the onus is on us to show that we can stay ahead of the pack.
Customers and partners most of all should welcome the competition to provide the best product in a increasingly well-defined category—the shared product architecture protects investment and makes it easier to sort the long-term winners from the losers, while competition within the category ensures customers derive maximal value for their dollar.
Perhaps the question I get asked most is why isn’t Pure looking to get acquired by one of the incumbent storage players given our original product vision has been anointed by the market leaders (we have turned down several such overtures). The answer remains simple: We believe the best outcome for our customers, partners, employees, and shareholders is to create the next great storage company. The transition from mechanical disk to solid-state is the most profound to ever hit the storage industry. Dislocations of this scale inevitably launch one or two long-term independent companies, and our aim is to parlay our pole position in the all-flash array race into an IPO down the road. More importantly, our fundamental premise is today’s disk array buyer is not getting enough value for their money. Attempting to disrupt the storage market from within an incumbent is a classic Innovator’s Dilemma, with the inherent risk that solid-state solutions will be held back rather than empowered to displace the mechanical cash cows.
No doubt the coming years will prove even more exciting for flash storage with enterprises and service providers coming to the same realization that consumer sites like Google, Apple, and Facebook already have—that the legacy of mechanical disk should no longer hold back performance-intensive applications. But one thing is clear now: If you want to see what EMC and NetApp all-flash products aspire to be, sign yourself up for a Pure Storage proof of concept today.