Pure Storage is finally out of stealth. As you can see, we have launched our product—the Pure Storage FlashArray and its tightly-coupled software, the Purity Operating Environment. We also have a brand new website, whitepaper, product datasheet, and bundle of press releases.
Until now, we have been talking from behind the veil of stealth about how disruptive solid-state flash memory will prove to be for data center storage (see Why Flash and the Top Ten Reasons Why Flash Changes Everything). We’ve been quietly building our product and testing it with innovative customers in some of the most challenging storage environments in the world. Customer feedback has helped us build a better and more reliable product, in fact by the time our product GA’s later this year, we will have spent almost two years in customer beta trials to ensure we meet the bar for enterprise-class resiliency.
With such a huge market (data center shared storage is approximately $20B/year) and substantial technology disrupter (solid-state flash memory), you should expect competition to heat up dramatically in the months and years to come. Competition drives innovation and lower costs, and thus is great for end-users. But with the proliferation of new companies and new products, it can be hard to pick long-term winners. Here’s what we think will be the winning recipe for solid-state storage:
- Lower cost than disk – Flash makes inline data reduction possible for performance storage (see this blog entry). Pure Storage’s “5-20X data reduction” means it takes that much less space to store the same dataset versus the typical disk alternative. High-performance data reduction can deliver an all-flash array at a $/usable GB cost that is well below that for performance disk. We expect many vendors will be introducing primary data reduction technologies in the coming years. Shop around, as data reduction effectiveness will vary dramatically, and ask your vendor the tough questions: Is your data reduction inline? Does it impact performance? Do you run your benchmark numbers with data reduction turned on?
- Enterprise array functionality – Choosing solid-state storage should not require endusers to compromise on the mission-critical reliability, compatibility, and scalability they get from a traditional enterprise disk array. The winners in solid-state storage will deliver standard interconnects like Fibre Channel and iSCSI; scalability from 10s to 100s of terabytes; and high availability (HA). HA has traditionally served as a litmus test to distinguish true enterprise arrays from mere storage appliances. Forcing a customer to buy two appliances for HA is just not cost effective. Look for active/active HA configurations, and pull drives, cables, and power cords to prove that the array self-heals in the face of such failures.
- Purpose-built for flash – Retrofitting SSDs into a legacy disk-centric array simply cannot deliver the 10X in performance, space & power, and simplicity afforded by a holistic redesign for flash. If you just treat flash like disk, you ultimately must provision more expensive flash (which is why it costs $40/GB and up). Flash is such a profoundly different media that it is deserving of a storage architecture optimized for its performance and failure characteristics—the same treatment that disk has enjoyed for decades.
As we said, it is crucial to launch a new venture only after you have hit critical mass – which, with a demonstrable product and customer results – we now have. It feels great to be out of the gates with Pure, and working toward our aspiration, which is nothing less than to build the next great storage company. Stay tuned.