IDC recently produced a report showing the sales and shipment of all-flash and hybrid storage arrays.  The report shows EMC as top dog and not surprisingly, EMC have provided a free downloadable excerpt, available here.  I always like to dig into numbers into a little more detail as you can start to uncover interesting trends or anomalies (I guess that’s my Maths degree background for you).  In this instance I am struggling to work out how the figures translate to a comparable $/GB figure.


All Flash Revenue ($M) Shipments (TB)  $/GB
EMC 112.3 13404.5  $8.38
Pure Storage 90.9 7557.9  $12.03
IBM 82.9 22773.0  $3.64
NetApp 45.0
Solidfire 35.6 7525.7  $4.73
Nimbus Data 34.3 7501.3  $4.57
Violin Memory 5583.9
Other 95.3 19130

First of all, let’s look at the all-flash figures.  The summarised numbers are shown in the first table.  IBM is way ahead on capacity (TB) shipped, but in relative terms their revenue is much lower than the other vendors, giving a nominal $3.64/GB figure.  Compare this to EMC, the revenue shipping leader, at $8.38/GB, almost double the cost.  Pure Storage comes out even more expensive at $12.03, whereas SolidFire and Nimbus are at similar levels of just under $5/GB.  Reading the small print in the report, bundled, non-itemised software isn’t included in revenue totals, so presumably IBM and EMC use the BMW method of pricing where every feature is a priced option.  From memory, Pure, SolidFire and Nimbus are pretty much inclusive on cost in terms of features, which makes the latter two much better value for money.

What seems odd in the report is that NetApp appears in the revenue table but not in the TB shipped, meaning the amount of capacity it shipped must be lower than 6th place Violin Memory.  This translates to a figure of at least $8/GB for an all Flash NetApp array and also is counter to their claims to be a leader in flash shipments (see comment on this post).  Similarly, Violin Memory shows up in the shipments, but not in the revenue, making their revenue less than Nimbus and translating to a figure of at least $6/TB.


Hybrid Revenue ($M) Shipments (TB)  $/GB
EMC 1575.8 45273.7  $34.81
NetApp 891.8 22209.5  $40.15
Hitachi 521.2 37490.3  $13.90
IBM 408.7 18368.5  $22.25
Dell 211.3 44637.7  $4.73
HP 113.5 7511.9  $15.11
Other 715.6

The hybrid numbers are a little more difficult translate into meaningful use.  It’s no surprise to see EMC coming out top on both revenue and shipments as a hybrid could be any array accepting flash and/or HDD.  However the figures specifically exclude the HDD capacities shipped with the arrays yet show the cost of those systems; only the flash capacities are provided.  This is marketing obfuscation at its best, as it provides no way to fairly evaluate each vendor on the $/GB metric. However I did the calculation anyway.  This produces some interesting results.  NetApp appears 10x more expensive than Dell and way more expensive than an all flash system.  I can only take this to mean that the percentage of flash shipped in every NetApp system is much smaller than with other vendors.  This wouldn’t surprise me; all-flash disk shelves make no sense and Flash Cache is a read-only technology.  At the other end of the scale, Dell have figures similar to all-flash systems provided by Nimbus and SolidFire.  Does that mean that Dell’s hybrid systems are mostly flash?

If we take a single vendor example, such as EMC and use the all-flash figures as a guideline, figures of $8.38/GB (all-flash) and $34.81 (hybrid) imply around 25% of the cost of a system is flash (but not necessarily capacity).  We know flash drives are significantly more expensive than HDDs, so this implies only a small fraction of hybrid array capacity (in the EMC example) is flash storage.  This assumption is confirmed with NetApp’s higher $40.15/GB figure.  Looking at Hitachi and HP, we see more modest figures ($13.90/GB and $15.11/GB respectively), implying both of these vendors are shipping hybrid arrays with relatively more flash installed.

The Architect’s View

When analysing data, the well known adage “lies, damn lies and statistics” has to be at forefront of the mind when making any assumptions.  Figures can be presented in a way that best suits the writer.  In this case, the IDC report is abridged with less data provided than in the paid-for version of the report and some of the costs comprising typical shared storage functionality aren’t included.  That said I’m sure the source of the data is genuinely provided and in no way am I suggesting any impropriety.  Ultimately, the more data we have, the better informed we can be, however as usual, Caveat Emptor applies!

Related Links

Comments are always welcome; please read our Comments Policy first.  If you have any related links of interest, please feel free to add them as a comment for consideration.  

Copyright (c) 2009-2018 – Post #84F3 – Chris M Evans, first published on, do not reproduce without permission.


Written by Chris Evans

With 30+ years in IT, Chris has worked on everything from mainframe to open platforms, Windows and more. During that time, he has focused on storage, developed software and even co-founded a music company in the late 1990s. These days it's all about analysis, advice and consultancy.