Today, IT consultant and blogger Chris Evans published a fun post dissecting the recent IDC All-Flash Array vendor market share numbers, and The Register did their round-up as well.  Whenever these reports publish there’s always a lot of speculation about them, as everyone digs in and tries to look through a cloudy porthole at what insight they shed on the industry.  Many of you probably read these reports, but not everyone gets the chance to be on the vendor side of the equation, so in this post I’ll share my perspective, and tell you what I know about the process for creating these reports from the “inside” of the industry, and perhaps shed some light on what I believe these reports can and can’t tell us.

Some observations:

1 – First off, thanks to all the Pure customers and partners!  I’d just like to say we’re pretty proud of where we landed in recent IDC and Gartner market share reports, as well as industry reports like the Gartner Magic Quadrant (where Pure was in the Leader’s quadrant) and 451 Research’s Heat Index in their Storage Wave 18 study (where All-Flash Arrays were the #1 hot technology, and Pure was the #1 in-plan vendor).  The reports differ a little bit based upon analyst insight, but fundamentally all of them paint a picture of the all-flash array market being on fire, and of Pure, EMC, and IBM being the clear market leaders.  From Pure’s perspective, that’s good company, and we feel proud to be executing extremely well in this market as a best-of-breed leader, amidst competitors with literally 10X-20X the number of sales people, channel partners, etc.  There’s a subset of the market who values buying best-of-breed technologies from innovative vendors, and we’re serving that subset extremely well.  Ultimately analysts build their view of the market by talking to many, many end users and partners, so thanks to everyone who has shared their Pure success stories with the analyst community!

2 – Understand where these numbers come from.  Let’s be clear here: analysts have an extremely tough job.  They live in a world of imperfect or (mostly) missing information, and “analyze” (hence their name) all the smoke signals they can find to produce these reports.  What I think many people don’t understand, is that vendor information on bookings and TBs shipped is confidential, and these reports represent educated analyst guesses.  Vendors vary in their internal policies about what information they do or don’t share (in Pure’s case, we have never shared revenue or unit volume information with ANY analyst firm), but in most cases it is a guess from analysts informed by varied levels of hinting and disclosure from vendors.  Regardless, analysts have to rely on the vendor accuracy of the information that is shared, which is often challenging due to the complex internal allocation methods vendors use to attribute large, multi-product deals between product groups (more on that below).  Net net: it’s a murky process, and analysts rely heavily on their market knowledge to try and fill in the many missing pieces or rationalize conflicting data points.

3 – Don’t confuse “factory revenue” vs. “street price” or “total solution price.”  Both Chris and a number of others took the opportunity to do simple math to divide the estimated total revenue of a vendor by the estimated TBs shipped for a $/GB price.  While this is interesting, keep in mind a few things:  First, these numbers are “factory revenue” – i.e. what the analysts believe the vendors actually got paid at the end of the process – not street prices.  This means that they don’t necessarily include mark-up for distribution, channel partners, international uplifts, bundling by converged infrastructure solutions, and the myriad of other things that might stand between the “factory” and what the customer pays.  Said differently – these numbers likely underestimate street price pretty dramatically.  Second, these numbers are for the flash systems themselves, but don’t include the “accoutrement” of other things required by vendors to make their flash solutions work.  Buying IBM FlashSystem and need SVC to add actual array features?  Not included.  Buying XtremIO and need VPLEX and RecoverPoint to add resiliency and manage your data migrations for upgrades and expansions?  Not included.  Need professional services to make it all work?  Not included.

4 – Raw vs. Effective $/GB.  In today’s world, comparing $/GB raw flash prices is essentially pointless.  While some all-flash arrays come with beefy controllers running sophisticated software like data reduction capabilities (Pure for one), others are simple flash devices with little advanced services, and a smaller price to match.  The results is that there is a real cost difference between vendors when translating between raw and effective.  A simple waterfall looks like:

  • Start with $/GB raw
  • Add in costs for RAID/mirroring overhead, flash management, over-provisioning, and metadata (this increases $/GB raw)
  • Reduce costs for the value of deduplication and compression if offered (lowers $/GB)
  • Increase costs for the added HW and SW cost of services such as snapshots and replication (this increases $/GB raw)

The net net: for many vendors the $/GB effective is typically more expensive than the $/GB raw (as has been the case for decades in the disk world), and for some vendors (like Pure), the $/GB usable cost is dramatically lower due to efficiency technologies.  Said differently – if you are comparing $/GB raw between flash vendors, you are comparing apples and oranges.

shell game

5 – Large Vendor Revenue Allocations.  This one’s a biggie.  For a single product best-of-breed player like Pure Storage, asking what percentage of our revenue actually comes from all-flash sales is simple: 100%.  For larger, “Big Storage” companies, this is much harder.  A large deal is booked that includes some maintenance extensions of old arrays, some new disk arrays, and maybe a new all-flash array.  How much revenue is allocated to each?  Are all transacted at the same discount?  If there was a heavy discount, was the new stuff or the old stuff discounted, and what is the fair revenue allocation?  The scary part is that there are incentives up and down the chain in BS companies to allocate revenues to the “right” products: quotas on emerging products for individual reps, stronger quotas for district managers and VPs, higher discount approvals available for deals that include AFAs, etc.  The result: Big Storage companies can create revenue acceleration for emerging products leveraging their market might, whether that product is bought, sold, given, used, or un-used in the actual market.


In Conclusion…

So…where does all this leave us?  Regardless of the showmanship of vendor jockeying for the top market spot, I think the real news here is the all-flash array market’s clear and unmistakable march forward.  I think we’ll all remember 2014 as the year that AFAs went mainstream.  It’s pretty clear from the recent analyst reports who the handful of leaders in the market are (some leading via technology, others via market might), and you owe it to yourself to kick the tires on all the leaders, and to make your own choice on who is the best fit for your environment and the best long-term partner for your business.  Whoever you choose – a choice on flash as we head into 2015 can’t be a bad one ;).