Conjectures and rants about enterprise storage evolution

Yesterday I spotted this article on the register. It shows a graph with the trend of storage sales growth for some of the major vendors.
In this case, storage means external storage and there are no doubts that NetApp and EMC (and HDS?), pure storage players, are doing better than HP, Dell and IBM (generalists) but, at the same time, I’m not sure that looking at “external storage” is still the right way to measure the real success of storage vendors… So I’m going to add some personal considerations.

A different way to measure success

Last week I attended HP Discover. During one of the meetings, David Scott (former 3PAR CEO and now the man that is leading HP Storage) clearly stated that HP 3PAR is selling very well (1B$ annual run rate) but overall numbers are below expectations because DAS sales (direct attached storage like MSA/P2000/JBODS stuff) are weak. In fact, DAS sales are strongly related to server sales.
Talking in this terms, Scott is implicitly saying that HP storage is doing well: long story short, they are earning money selling very high margin enterprise storage instead of low value (and low margin) boxes… and it’s not hard to believe!

Starting from this point, I formulated some conjectures about actual enterprise storage that I would like to share with you. Unfortunately, since I do not have access to detailed storage sales numbers from array vendors and disk manufactures, it will be very hard for me to verify what I’m going to say.
I think that Dell, HP and IBM have three different stories to justify their poor external storage sales and they aren’t all masked successes.

HP is doing well

I’ve already wrote above about HP but there is more.
If you think at server+DAS sales there are two key factors to consider: HP is losing some market share on x86 servers and new server models can install many more disks than in the past (up to 60!). This confirms David Scott’s thesis, but I have a concern on the old EVA (P6000) because I don’t know if EVA customers are migrating to 3PAR or not… and if this is actively contributing to the 3PAR success or to the overall fail.
Also the P9500 (rebrand HDS VSP) is a big question mark: HP pushes on 3PAR now but how many XP24000/P9500 customers are willing to change their consolidated stuff (especially if there is a mainframe around… )?
In any case, thanks to 3PAR, HP is probably doing better than sales figure suggests!

Dell is so and so

PowerEdge R720xd Rack Server - 24 HDDFor Dell we have a similar scenario but the results are even more confused.
Dell is gaining market share on servers and they are still more present in SMB than Large Enterprise. Their DAS attach rate is high and now they can be even more competitive using internal hard drives on latest models. The probable consequence is that Powervault entry-level arrays/JBODS are not seeing a good momentum.
On the other side Compellent is not a clear success and, during the last Dell Enterprise forum, I don’t have noticed major announcement on Compellent or Equallogic or storage in general. Their recently acquired startups are still struggling to show their potential and lack of a complete and/or integrated offering is not helping. For example, all primary vendors have a 100% flash array now while Dell doesn’t!
They are probably doing well in the SMB but they are failing to gain traction on the high end with the consequence of an overall bad sales figures.

IBM is doing bad

If DAS is important for them too, they are losing on both sides: weak server sales means that they don’t have servers to attach cheap external storage.
Furthermore, my opinion is that IBM can sell storage only to its customers: if a customer buys servers from IBM (especially in case of iSeries or pSeries deals) he probably buys IBM storage, but it’s hard to find end users running servers from other vendors and IBM storage… isn’t it true?
I also find that some of IBM storage is less interesting than other stuff and, looking at sales, probably I’m not alone to think that.

Disks matter

If DAS, and internal disks, are important we shouldn’t look only at the revenues but also at the quantity of installed disks. I would like to know how many disks the three server players buy/install in comparison to the 3 pure storage players.

Many applications and new technology are really changing the way “enterprise storage” is consumed. For example, I’m seeing end users that are running business analytics on standard x86 servers full of 600GB/10K internal hard disks and a PCIe SSD cards and no shared storage!

Scale-out architectures work without an ordinary shared storage and ISPs and enterprises are looking at them with more interest than in the past: Filesystems like Ceph, VSAs are are only the first examples that come to my mind. Software-defined Storage and commoditization are somehow happening…

Bottom line

The final picture is very complex.
The first thought that comes to my mind is that pure storage players will continue to increase market share for a while,
NetApp could suffer more than others because its OEM sales to IBM and Dell, but it is on the first group.

On the long term things will drastically change: commoditized software-defined storage can severely hurt traditional low-end to mid-range external storage sales. Here the winners will be the storage software vendors (for example, I think that it’s not a case that EMC is buying in that space).
Cloud storage and object storage will become more important, and also in this case we often talk about servers with local storage and software on top…

Enterprise storage is quickly evolving, are traditional storage vendors ready to keep up the pace?

Every comment is welcome.

  • Prasanth

    Enrico, your conclusion seems pretty spot on (not surprisingly, we’re a software defined storage company!)

    >> On the long term things will drastically change: commoditized software-defined storage can severely hurt traditional low-end to mid-range external storage sales. Here the winners will be the storage software vendors (for example, I think that it’s not a case that EMC is buying in that space).

    We believe there are important distinctions within SDS vendors and what exactly they offer – and if they provide guaranteed QoS (IOPS, throughput) and on-demand performance, software-defined and software-only storage can replace even the high-end storage arrays. SDS is not just about the hardware flexibility and affordability, but it’s also about allowing datacenters to easily scale to thousands of applications and meeting their performance demands. As they say, ‘software is eating the world’, and storage is no exception :)

    Prasanth
    CloudByte
    http://cloudbyte.com

    • http://juku.it/ Enrico Signoretti

      Prasanth,
      thank you for your comment.
      I agree with you that SDS can also be considered in the high end, but only for new projects/infrastructures.
      My point is that High End tier-1 storage systems (VSP, VMAX, DS8K) are often connected to legacy systems too (Unix, mainframe, and so on)… it’s very difficult (and costly) to touch those silos! :)

      Enrico

  • Chris Rodinis

    Here is a video overview of the Dell PowerEdge R720xd: http://www.youtube.com/watch?v=eQEjY2Xa6g0