EMC Unity – It’s DéJà Vu All Over Again
By Ajay Singh – VP Products

It’s like a scene out of the movie Groundhog Day – a big-name vendor promising a shiny new “re-designed from the ground up” product line. Marketing claims fly through the air like confetti, about how this changes the industry landscape:

  • Want flash? We’re totally optimized for it.
  • The complexity you’ve grown to hate? We’re completely re-designed to be simple – just look at the shiny new GUI.
  • The pricing models you loathe? This time we’ve really made them even more “almost-all-inclusive”.
  • Multi-threading? Entire sub-sections of the code are now multi-threaded!
  • Pointer based snapshots? Forget that we derided them for the better part of two decades – now we have them too.
  • Cloud based monitoring? We’ve got it.
  • And it all starts at $9,995.95, so we’re going to kill those “flash optimized” emerging vendors. Never mind what value you get for that price – all that matters is that it starts at only $9,995.95.

There’s just one problem – we’ve heard almost the exact same story and the same claims before. On more than one occasion! Remember EMC’s VNX introduction back in 2011? Or the VNX2 in 2013? Sure EMC sold a fair number of them based on almost the same claims, though there may have been some unhappy customers in their wake. The reasons are well understood today, and Unity promises more of the same.

Let’s evaluate Unity, not just in a single dimension, but against nine fundamental design principles for modern flash arrays:

  • Data reduction: Unity may well offer models ending in the letter “F”, but is no more flash-optimized than VNX, or its predecessor, the Clariion. Its venerable two-decade-old data layout was never designed for inline dedupe and compression (with variable block support, and without big performance overhead). This means customers will either face performance tradeoffs, or be forced to buy 2 – 5x more raw flash capacity than they would from Nimble to achieve an equivalent amount of effective storage capacity.
  • Endurance management: To get the most out of flash technology, an all flash array needs to implement an effective strategy for flash endurance management. The path EMC has chosen is to delegate endurance management to SSD firmware, forcing them to rely on more expensive SSDs with over-provisioned flash capacity. This will likely inhibit EMC’s ability to take advantage of future generations of more cost-effective flash storage.
  • Write performance optimization: Like all older architectures, Unity still promises performance that scales with spindle or SSD count, rather than with CPU cores like modern flash-optimized architectures (so that price / performance improves with every generation of CPUs).
  • Scalability: Unity offers no scale-out capabilities, and like its predecessors is expected to have no future performance scaling at all except through forklift upgrades.
  • Unified secondary storage: Even if Unity F arrays can replicate to their hybrid cousins (VNX-2), those are still weighed down by their archaic tiered architecture and complexity.Further, the pointer-based snapshots are expected to have the same performance overhead as with older VNX versions, limiting their utility for data protection.
  • Resilience: Unity still offers the old-school muddled menu of RAID choices – you can either get reasonable performance or capacity overhead or reliability, but not all three at once. And for an enterprise product line to still not offer triple-parity protection in this era of multi-terabyte disks borders on neglect. Nimble not only does this by default, but also adds intra-drive parity to further protect customers against sector failures.
  • Cloud-based predictive analytics: EMC’s CloudIQ is a token nod towards cloud based monitoring, but as their blogs say, anything more than “lightweight reporting” requires additional on-premises management software. This is a far cry from what Nimble delivers today with true predictive cloud based analytics and monitoring.
  • Security: Unity offers hardware-based system-wide encryption, which does have the benefit of not depending on special self-encrypting drives. However a modern software architecture like Nimble delivers completely granular encryption as part of the data path, providing secure separation between tenants’ data, granular data shredding, and continuity of encryption when replicated.
  • Storage lifecycle: If history is any guide, there will be one final insult added to injury – a forklift upgrade for customers of “legacy” VNX arrays. Modern companies like Nimble Storage offer a better way. We start with an architecture that allows performance and capacity to be scaled independently, and that allows seamless upgrades across multiple generations of hardware. Add to that a business approach (called Timeless Storage) that offers truly all-inclusive licensing, guaranteed flat support pricing, and the option of free controller upgrades every 3 years – and you can see how modern architectures can make storage lifecycle management a pleasure.

Once the glitter and hype of the launch event is done, we expect a growing number of customers to examine the new claims critically and join their peers in saying “won’t get fooled again”. We invite you to make the comparison with a fresh approach to data storage, by taking the Nimble Predictive Flash Challenge.