It's open season in storage right now, but the door is starting to close. If you have something that works with filer or block access protocols then develop away. Servers can use it. The product can work with storage area networks (SANs) or filers and you are free to sell it to customers with servers that use SANS or filers.
So we see Avere and Alacritech with filer accelerators, and Nirvanix, Cirtas, Nasuni and other startups focused on the cloud, basically intercepting local data centre storage operations and diverting them to the cloud. It's relatively easy to do this. You don't need to build a complete stage stack when your innovation is just a part of one layer in that stack.
Step forward converged infrastructures with integrated top-to-bottom stacks. Oops. Imagine an HP BladeSystem Matrix with integrated servers, system software, networked storage and networking. How do you slot an Avere accelerator into that? Where is the integration point for StorSimple's appliance? Basically, unless your storage innovation is software that can run as a virtual machine you are stuffed with converged IT stacks. Best of breed won't cut it, particularly in a hardware sense. Who cares if your block access accelerator cuts I/O latency by 20 per cent? If it doesn't fit into BladeSystem Matrix, Vblock or FlexPod or any other integrated stack then you are cut out of the market.
These converged stacks are pretty open right now and storage startups are still happening. But five years down the converged stack road I fear what will happen is what I'm anticipating hearing already; the grinding and squealing of open systems interface doors closing.
Once they are shut then the vendors will be in preserve-their-customer-base mode and will not tend to support disruptive innovations that eat their existing product lunch. ®