This week, the storage industry turns its focus to one of the industry’s flagship events – EMC World. With so much uncertainty around how the new Dell/EMC structure will evolve, you can’t help but wonder if this will be the final EMC World of its kind. As we prepare to potentially bid farewell to EMC World as we know it, we decided to put some thought into the theme of this year’s conference – Modernize.
We love to talk about modernization. We, too, are firm believers that modern storage drives massive business transformation for customers. And like EMC, which has doubled down on flash in recent months, we believe that the underlying basis for modernization is an all-flash storage environment. So it seems ironic that the theme of EMC World – at a time when the company is being taken private and introducing retrofitted technology under the guise of innovation – is Modernize.
With its EMC Unity/VNX3 announcement today and recent EMC VMAX F announcement, EMC would have us believe that we can modernize by retrofitting 20-year-old architectures with all-flash. Can you modernize through retrofit? Fundamentally, it just doesn’t compute.
Imagine you’re in the market for a hybrid vehicle, because it gets 50 miles to the gallon. It’s not just the hybrid engine that contributes to great gas mileage. It’s the entire car, from aerodynamic design to weight to horsepower. If you simply drop a hybrid engine into a 1960s Mustang, for example, 50 miles to the gallon is no longer on the table. And, quite frankly, that’s not innovation – it’s retrofitting.
EMC has now pivoted back to retrofit VMAX and VNX arrays as their primary all-flash selling motion, so why will this retrofit strategy be limiting? It’s mostly about the metadata. Modern, all-flash arrays rely on rich, always-fast metadata architectures – it’s that metadata speed and granularity you need to implement advanced data reduction, copy services and flash management. The metadata is core to the inter-workings of the array, and is difficult or impossible to retrofit. Here are two tangible examples of how metadata shortcomings manifest themselves in EMC’s retrofits:
- Neither VMAX and VNX/Unity support data reduction today, leaving IT buyers to continue to buy 72 percent more flash than they need. Data reduction on all-flash arrays enable customers to buy less (a lot less). For example, it used to take 125TBs of capacity to store 100TBs of data. Data reduction has made it possible for Pure Storage customers to buy 35TBs of flash to store the same amount with an average of 5-to-1 data reduction. That’s 72% reduction in the amount of capacity that has to be purchased upfront and over time. Both EMC VMAX and VNX/Unity architectures were developed in the 80’s/early 90’s – well before all-flash storage was a practical consideration. The metadata of these architectures was optimized for mechanical disk, not flash. Retrofitting these arrays with all-flash means weaker or no data reduction. Sure enough the VMAX and VNX/Unity don’t do anything to dispel this claim… neither supports data reduction today, leaving IT buyers to continue to buy 72 percent more flash. That’s not modern.
- Retrofit arrays don’t allow for continuous hardware evolution. Compute and Flash continue to advance annually, consistent with Moore’s Law. Any modern all-flash array needs to be future-proof, i.e. built to allow the customer to non-disruptively upgrade their array to take advantage of the performance, density, and cost gains that come with annual technology advances. The Pure Storage FlashArray, via its stateless architecture and flexible metadata, has a proven history of enabling customers to take advantage of both flash and compute advances – non-disruptively. In contrast, mix-and-match of v-Bricks (performance/compute building blocks) is unsupported in the VMAX F, locking customers into current compute technology. Mix-and-match of flash types is a similar story: VMAX-3, the architectural underpinning of VMAX F, doesn’t allow mixing of 2D NAND and 3D NAND flash on the same array. Similarly, VNX3/Unity locks you into the performance of the current and specific controller model that is purchased… there’s no non-disruptive upgrade to other Unity controllers (current or future). Various constraints in the VMAX and VNX3/Unity architectures mean the customer is either left “frozen” in today’s technology or re-purchases new hardware and software and migrates 100% of their data to put the annual technology advances to work for their business. This lack of flexibility comes from a rigid engineering approach – and inflexible metadata. That’s not modern.
When the bright lights and machine-made smoke clear from Vegas, two simple truths emerge in EMC’s new flash strategy:
- EMC’s two flagship all-flash arrays, VMAX-F and Unity/VNX3, are both retrofits. While filling legacy arrays with SSDs is a fast patch job, you can’t retrofit efficiency, simplicity, or true flash hardware/software integration.
- EMC is now selling four highly-overlapping all-flash arrays, each of which has fundamental compromises. Some are good at performance, some resilience, some at efficiency, and some at simplicity. But what if I don’t want to choose?
Four All-Flash Arrays, No Ideal Solution
Part of what’s made all-flash arrays so attractive to the market is that they are potentially a gigantic leap forward in multiple dimensions:
- Performance: game-changing latency to make storage not the problem
- Scale: virtualized mix-workload consolidation at a new level of scale
- Efficiency: leverage inline data reduction to actually reduce the cost of storage
- Resiliency: nothing mechanical means less to fail
- Simplicity: shed decades of accumulated complexity, tuning, and manual administration
When customers buy into the promise of an all-flash array, this is pretty much the shopping list. And to be fair, EMC does deliver these attributes – but unfortunately not from a single product, which is why they’ve likely chosen this 4-product strategy. Below is our relative assessment of the attributes across these EMC products. As a customer, how do you choose?
Want resiliency? Pick VMAX. But what if I want resiliency and efficiency? Want the best performance while achieving simplicity? The choice becomes less clear. And this is only the beginning – don’t forget the Dell options (all-flash Compellent, all-flash EqualLogic), the all-flash hyper-converged options (all-flash VxRail) and software-defined all-flash options (ScaleIO, vSAN).
At Pure, our strategy is to deliver a purpose-built all-flash architecture from two products – one optimized for applications, databases, and VMs (FlashArray), and one optimized for large-scale file and object environments (FlashBlade). Both purpose-built for the use cases they serve, and neither making a customer choose between performance, resiliency, scale, efficiency or simplicity. We’ve built our arrays from the ground-up, investing in hardware and software that have only lived in the era of flash. And we’re only starting to exploit the advantages of more closely-coupling flash-optimized hardware and software – just check out what’s possible with our new FlashBlade product.
Some Recommendations for Evaluating a Modern EMC
So, if you’re looking to refresh your VMAX2, VNX2 or XtremIO, here are a couple of tips for you:
- Conduct a proof-of-concept of each of the EMC all-flash options so you can find out which option is best-suited (i.e. which compromises you can live with) for your real-world environment.
- Bring in Pure Storage FlashArray to do a proof-of-concept along with the EMC all-flash options. This way, you can experience the difference among the EMC all-flash options and Pure, instead of taking our word for it. Either way, you are likely facing a “forklift migration.” With Pure, you can make it your last.
Modernization is not about getting incrementally better. It’s about getting exponentially better, all of the time, by innovating with purpose-built solutions designed to maximize the benefits of flash today and seamlessly integrate future technologies. That’s simply not something EMC can offer, and a problem that certainly won’t be addressed with retrofit arrays.