EMC tried to make a bit of a splash with flash last week with the announcement of its VNX2 storage platform with six new hybrid arrays and an all-flash model.
Forgive us if we’re not impressed…
These are not the long-term solutions enterprise customers need for next-generation applications built around virtualization, cloud computing, and big data, all of which place a premium on storage performance and elastic scalability in mixed workload environments. Why not? Let us count the ways:
Organizations worldwide are experiencing massive data growth. They need simple, agile, and elastic IT infrastructures to meet demanding enterprise SLAs, improve productivity and impact both the top and bottom line. Storage I/O has been one of the prime bottlenecks. While flash changes the game versus traditional HDD drives, without the right deployment architecture it’s a battle half won or dollars wasted. Does VNX2 have the right architecture to maximum flash? Nope.
The underlying architecture for VNX2 remains optimized for sequential media. This is the wrong approach for customers, but perhaps the right approach for EMC’s bottom line — it protects its legacy base. The solution fails to deliver the consistency, predictability and elastic scalability needed by enterprise workloads. If you’re a customer or potential customer the choice is yours: do you prefer a battle half won or do you prefer dollars wasted?
The EMC VNX2 features controller and software enhancement designed to fill in an original VNX gap by leveraging the multicore processing power of powerful CPUs. But no matter how much they dress it up, it is still a workaround approach that does not capitalize on the random access media capabilities of flash.
Fundamental to the discussion of the right storage software architecture optimized for flash is the debate on scale-out vs. scale-up. Scale-out always wins the debate. Why?
Scale-up architecture has intrinsic scale and elasticity limitations. While the controllers can be upgraded, there is an overall limit on capacity and performance scale.
Scale-out, on the contrary, provides the agility and elasticity needed for any workload and application, including mixed workload environments. Organizations can start small and grow with the business demands. This balance and flexibility can drive significant CAPEX and OPEX savings.
Examples: A retailer doesn’t need to stack up infrastructure ahead of its seasonal peaks. A financial services organization can be agile and elastic to dynamic market conditions — acquisitions, mergers and consolidation.
Working with customers and watching industry trends of virtualization consolidation, cloud computing, and Big Data, we believe that all-flash arrays with scale-out architectures — designed from the ground up for random-access media — are going to be the de facto Tier 1 storage array solution.
A scale-out architecture, with the right balance of compute, networking and capacity (along with variable-block-size support for mixed workloads) is an excellent choice for enterprise customers, providing industry-leading agility, simplicity, and elasticity at the lowest price/performance for today’s dynamic data centers.
EMC knows this – that’s why they bought XtremIO over a year ago. Unfortunately, it seems that EMC has been unable to stabilize XtremIO’s alpha product and get it to a point where it can actually sell it to customers. Why? Scale-out, like anything that is truly worth doing, is really, hard to do well!