Last week NetApp launched a new line of All-Flash Arrays called AFF. It’s based on a specialized version of ONTAP, which has added peculiar optimizations to read/write data paths. It’s not the notorious FlashRay yet, but it’s not that bad and it has something to say… especially if you are already a NetApp customer.
New entrants do it better…
There is no doubt we are not talking about the top notch implementation for a Flash Array. Startups like Pure, Solidfire, Kaminario and others designed their product from the ground up just a few years ago. They started from scratch without constraints and without a legacy to maintain!
They have a modern design, in some cases more scalable, efficient in the way they manage flash and easier to use. Sometimes they have some particular features that make the difference. If you start building a brand new infrastructure, the advantages can be really relevant and even the concern to work with a startup will/could vanish!
Unfortunately, in most cases, this isn’t what always happens. Enterprise IT infrastructures are usually already in place and moving to a flash storage is a necessity for just a small part of existing applications which need to be supported by much more predictable, fast and agile storage systems. I’m not talking about pure speed in this case, fast and agile could mean many things like faster provisioning, faster resource reallocation, faster migrations and so on.
…but Flash is just a feature
Flash is an enabler to have a more agile storage infrastructure but it’s not the only one! And in many cases infrastructure and organization complexity transform Flash into a mere feature.
At the moment, on a capacity basis, only 4% of storage sold is Flash. The rest is Disk. Four percent is much less than what is usually esteemed to be active data in a traditional enterprise (it runs in the order of 10 to 20%).
It will take some time before Flash becomes as important as disk in the datacenter and in the meantime a lot of things will happen!
Market landscape has changed
Today the world is hybrid (Flash+HDDs) and tomorrow it will probably be hybrid too (with faster+slower types of memory). Not the single system maybe, but infrastructures will remain hybrid.
Most storage infrastructures involve various storage tiers (including cloud) and end users want to move data between them seamlessly. The best way to do it is to have the same storage OS and a rich set of features. In fact, some vendors offer different solutions in terms of storage media but with a common set of functionalities and features throughout the entire product line (and NetApp is one of them).
Vendors like might not have the best AFA, but I can’t say the contrary yet either, but if the project is complex or the Flash Array has to be integrated in an existing complex infrastructure it is just easier to adopt a single vendor that can grant an overall good solution made of similar pieces. The single piece of the puzzle is less efficient than what you could obtain from a new entrant but now, after the latest announcement, it could be good enough to satisfy end user needs and keep overall infrastructure TCO under control. My argument could be disputable for smaller infrastructures where complexity is not a major issue, but again, for the end user, it could be easier to continue buying from the same vendor if satisfied with what is being offered…
2 years ago the All-Flash startups were particularly well positioned due to the way they managed Flash memory, but now this is just common ground for everyone (for example also NetApp offers a 7 year warranty on Flash now, a sign of confidence, don’t you think?). These startups still maintain the technological leadership and sometimes it is still evident… but traditional vendors are catching up very quickly with either acquisitions (like EMC XtremIO) or experience (like NetApp, when it says that some of the knowledge it has matured on development of FlashRay has been transferred to this special version of ONTAP and AFF… without mentioning HP or HDS).
The gap has been closing very rapidly, and these startups need to react as fast as they can. It’s not just about being All-Flash or All-Media, it is more about being able to serve many more different scenarios with the same basic architecture and set of features. It means having products capable of serving blocks as well as file or objects and build integrations (or integrated data services) on top of that.
In any case, today the market landscape has changed considerably and it’s really tough now to find a major differentiator. The most interesting vendors are working on analytics to make a difference and also primary vendors are looking more at managing data (the content) instead of the storage system…
I can’t really see any of these startups growing as expected by investors and the market, unless they have a broader product line-up to compete elbow-to-elbow with traditional competitors…