Today, the team at Pure Storage is wrapping up the first phase of our journey and embarking on the second. We call ourselves the Puritans, a moniker we are happy to share with customers and partners that join the cause. (Hopefully, our launch video thoroughly refutes the old adage that Puritans are those that fear that someone somewhere is having fun ;-)).
It has certainly been a wild voyage thus far. Our progeny, the Pure Storage FlashArray, is already proving itself on center stage: Customers have pitted us against a multi-million dollar, multi-cabinet disk array with honking DRAM and flash caches—configurations that are 10X larger and 10X more expensive than Pure’s FlashArray … and we’ve won.
Pure was founded in 2009. Since early 2010, our technology has been in heavy-duty use at select North American endusers. In those two+ years:
- We shipped more than 100 systems to customers;
- We supported production deployments for over a year; and
- We tested the product across a range of performance workloads, including virtualization (server and desktop) and database (OLTP and analytics), and across a broad spectrum of vertical industries (tech, financial services, cloud hosting, energy, pharma, manufacturing, advertising, legal, and government)
Along the way, we proved out the founding principles of Pure Storage:
- That inline deduplication and compression algorithms optimized for flash can deliver 4-10X data reduction and 100Ks of IOPS with submillisecond latency on virtualization and database workloads.
- That software (collectively, our FlashCare™ Technology) can be used to make economical MLC flash behave as reliably and as fast as more expensive SLC and enterprise MLC (eMLC). We’ve been called “the flash whisperers” for our ability to deliver consistently submillisecond latency and long-term robustness (annual SSD failure rates below 1/10 of a percent) with standard MLC.
- That this combination of data reduction and MLC enables Pure to deliver (alert: elevator pitch) all-flash storage that is 10X faster, more space and power efficient, and simpler than disk-centric arrays at a comparable price of $5-10 dollars per GB usable.
Proving the naysayers wrong on a couple of fronts has been particularly gratifying:
- “Deduplication is incompatible with high performance” or “Deduplication is too risky for primary storage” — Pure’s deduplication algorithms actually accelerate performance: with flash, writes are expensive, and random reads cheap, so it just doesn’t make sense to write the same data over and over again the way it is done on disk. And our implementation confirms dupes via byte-for-byte compares, so hash collisions cannot lead to data corruption. History tends to repeat itself: Not surprisingly, Riverbed heard similar rants in the early days of WAN optimization, but deduping your WAN traffic is now de rigueur.
- “There’s no market for all solid-state storage. Hybrids are the future, and every array will be mostly mechanical disk” — Analysts are now forecasting a multi-billion dollar market for all-flash storage. History again: DataDomain heard the same refrain about disk-backup losing out to tape supplemented with disk caches (remember VTL?).
Looking forward now, we could not be more enthusiastic about the opportunity in front of Pure and our partners. According to IDC, the impact of server virtualization has been an 83% savings in data center floor space, and 87% savings in energy costs. But virtualizing workloads puts a great deal more stress on the storage (check out the I/O blender). Solid-state storage is now poised to have a similar economic impact. Spinning disk typically sucks down more than 40% of the data center power budget. By delivering storage that is dramatically more space and power efficient, we can enable customers to expand their terabytes 5- to 10-fold on their existing power & space budgets. And because it is >10X faster (both in throughput and latency), we have customers that have been able to virtualize and consolidate I/O intensive applications, saving enough money that the storage effectively pays for itself. Most encouraging of all is the enthusiastic reaction of users to the dramatically faster performance, even for applications that are now going across the network to flash rather than accessing local disk.
Something similar is happening for our customer’s database applications. We have encountered workloads in which the batch load cycle takes so long that there is little time left for running analytics on the imported data. As a result, the business lives with severely constrained analysis, or else has to rely on older data. With our all-flash storage, we have cut database load times from >20 hours to <2 hours, and accelerated the analytic applications by 10X as well. The only question now is how to take advantage of all the new headroom, capacity that was procured for less than the cost of the storage Pure is replacing.
Solid-state storage is also more reliable storage. According to studies at CMU, roughly 1 in 12 hard drives fail per year. Good infant mortality testing can drop that number to 4-5%. In contrast, Pure has deployed 1000s of SSDs, both internally and at customer sites for 2+ years of hard labor. In all that time we have had one SSD get into a state that we couldn’t heal with our software.
In closing, the Puritan team is thrilled to have the first 2½ years of our journey under our belts, and our 2.0 release out in the market. But we are even more excited about what’s around the corner. We believe today’s performance storage customer is not getting their money’s worth for the >$15B they spend per year on 15K disk arrays, technology that’s increasingly antiquated by Moore’s Law. Our aspiration is nothing less than to lead the industry transition from mechanical disk to solid-state flash. In the hyper-competitive market of storage, we know every customer engagement will be competitive. No doubt EMC’s move into the all-flash array category via their purchase of XtremIO is going to add more fuel to the flash fire. We welcome that competition. Competition drives innovation and customer value. In competition lies our opportunity to prove the unique value in our architecture and our product. That’s after all why we spent 2+ years on our customer early access program: Because it’s storage, it’s got to be right.
But of course the biggest winners coming out of the next decade of storage will be customers, as they push 20th Century mechanical storage and all of the associated complexity out of the critical performance path of their business, and do so within the constraints of their existing budgets! With all-flash storage, faster, more space & power efficient, and far simpler than mechanical disk, why buy disk?