Yesterday EMC updated customers on their evolving flash strategy, you can read our CEO Scott Dietzen’s take here. There was the usual fanfare and drama one has come to expect from EMC launches, but let’s cut through the smoke and analyze what we really heard:
1. Server Flash goes web-scale, and the commoditization dog-fight begins. The first piece of news was EMC’s upgrade to it’s VFcache product line, and re-branding as XtremSF. We’re not server flash guys, so I’ll keep my comments here brief, but I saw two interesting things in the announcement. First, Fusion-IO has enjoyed a relatively open market in this space for the past few years, but that is no longer. EMC has doubled-down their focus, Violin just announced entry into the space, and Virident seems to be doing well…a nice battle is brewing as the technology standardizes and commoditizes, and I think most are realizing that it is all about the software. Second, and more interestingly IMO, is that the vendors are starting to admit that server flash is really a web-scale datacenter thing. It requires massive changes to the architecture that simply don’t lend it to playing well in the enterprise, and the vendors are realizing that this is where they need to focus.
2. EMC kills Project Thunder. This was no surprise, in fact we called it a year ago. EMC doesn’t need two flavors of networked flash, and it is pretty clear that the market for networked flash without data reduction or other advanced array services pretty much dried up in 2011.
3. XtremIO enters actual Beta. EMC has continued to slip their delivery timelines for XtremIO since the acquisition, and this announcement is just another step down that path. In fact, now that EMC and NetApp are boldly marketing their future flash products, we wanted to take a second to put the relative technology trajectories in perspective.
Pure Strorage has been shipping customer beta systems for over two years now, executed a broad Early Adopter Program (similar to EMC’s new ‘Directed Availability’ program) in 2011 where we first shipped units for revenue, hit our official GA date in May, 2012, shipped GA+1 in late 2012, and you can bet we have more coming in 2013. Shipping an enterprise-class storage system takes time to mature, and nothing can replace real hardening in the field in production customer deployments.
4. EMC Acknowledges Pure Storage as “The Leading All-Flash Array Startup.” Well, almost…they didn’t quite mention us by name, but if you look at the slide below they presented in their launch yesterday, it’s pretty clear who they were talking about. What’s amazing about this slide to me, however, is the extent to which they get it wrong…even basic stuff that they could have understood by reading our website let alone actually being in the market.
I’m not going to address these one-by-one, but a couple of things to highlight from my POV:
- Scale: scale-out vs. scale-up is somewhat of an academic discussion, customers should ultimately be asking about the max capacity a given cluster can support, and how much hardware / RUs / cost is required to get there. Pure was designed with a similar Infiniband/RDMA clustered controller interconnect architecture to XtremIO, we’ve just focused first on maximizing vertical scale before horizontal scale (vertical scale is more cost-effective than horizontal scale, horizontal scale becomes important when you need to expand performance beyond what vertical scale can deliver).
- Predictable Performance: this one’s hard to even discuss/refute it is so general. A huge portion of Pure’s technology goes into how we deliver predictable sub-millisecond performance from consumer-grade flash, but ultimately the proof here is testing flash products with your workload (not in the benchmarks), which is what I encourage customers to do. A hint for those evaluating XtremIO: try testing non-4K IOs and see how it compares to 4K. The perfect 4K IO is a mystical creature almost never seen in the wild, but more on that in later posts 🙂
- Compression: hmm…oddly missing. XtremIO doesn’t do compression, Pure Storage does. About half our data reduction value comes from compression, especially in database-centric use cases. Not having compression simply excludes XtremIO from about half the market opportunity for all-flash arrays.
- Inline deduplication: this one is a funny one for XtremIO to try and highlight as an advantage. Both Xtrem and Pure do data reduction inline, the biggest difference is that Pure’s operates at a viariable chunk size and detects duplicates down to 512 bytes in size, 8x smaller than XtremIO’s 4K architecture. This one’s simple, the smaller the chunk size, the more effective the deduplication.
- Thin provisioning: both have it, XtremIO at 4K chunks and Pure Storage at 512-byte chunks (again, 8x smaller).
- Space-Efficient RAID Protection: I can’t comment on XtremIO RAID since EMC has never shared any real details about it, but Pure uses a proprietary version of RAID-6 that we call RAID-3D. It protects against dual drive failures, acts at the data segment level vs. the drive level, is global across the whole array vs. being stuck in each brick, and has very low overhead. Check out a demo if you like.
- Snapshots: hmm…Pure Storage offers ZeroSnap snapshot technology, zero-overhead snapshots which enable every snapshot to function like a full clone in terms of performance and usability, but consume zero space overhead and require zero planning.
- iSCSI Support: hmm…been shipping GA in our array for almost 6 months now, roughly 30% of our deployments in the most recent quarter.
- VMware Integration: in addition to VAAI support, Pure supports a vSphere Web Client Plugin, and our 512-byte architecture auto-aligns perfectly with all VMs and applications.
And I can’t help but notice a few missing items from the XtremIO comparison list that customers might want to ask about (details are so light on XtremIO in this beta phase that I don’t know for sure about XtremIO support – you should ask):
- Non-disruptive code upgrades: it isn’t a real HA array without it. Pure performs NDUs without downtime or performance penalty.
- Non-disruptive capacity expansion: it isn’t a real HA array without it.
- Performance through failure: again, for performance storage it is important to maintain performance through a controller failure or drive loss. The FlashArray was designed to maintain performance through failure.
5. Finally, let’s compare today to today and tomorrow to tomorrow. When a larger, incumbent vendor is behind, a classic marketing maneuver is to attempt to “freeze the market” by marketing the future. If the vendor can only buy enough time by stalling customers, theory goes, they can catch-up and have a viable product before customers leave them. The problem with this strategy though, and why it usually back-fires, is that a) all this does is validate the market and get customers to realize that they really do want the future product now, and b) the disruptive vendors aren’t sitting still…so by the time the incumbent vendor “catches up” the disruptive vendor is yet again farther ahead. The most obvious recent example of “freeze marketing” is NetApp’s FlashRay announcement (our take), but EMC’s marketing strategy around XtremIO is just as lean forward.
EMC customers: We’re happy to have you compare XtremIO to Pure Storage. Our request is simple:
- If you want to compare XtremIO.today to Pure.today, that’s easy. Just ask us both to do a PoC and test them head-to-head. EMC is now reportedly ready for this with their “Directed Availability” program. Testing Pure is easy – just click here and we’ll contact you within 24 hours.
- If you want to compare XtremIO.next to Pure.next, that’s easy too….just get NDA roadmap briefings from both of us, and ask tough questions.
But please don’t fall into the “pre-marketing” trap of being asked to compare XtremIO.next to Pure.today…because by the time XtremIO.next arrives, we’ll be pretty far down the road 🙂