the market for servers would be very-fast persistent disk cache and/or just really fast disk. "enterprise" types are willing to pay rather a lot for that sort of thing. Rather more, I think, than most people will be willing to pay for an instant-on laptop.
Combine a few of those things with appropriate marketing and something like bcache[1] and you have a product you can mark up 10x over parts cost. Look at FusionIO for a current-gen competitive (though slower) product.
I totally agree with you on the "10x" price increase for "enterprise" gear, along with much higher profit margins per sale. For notes, I've bought and tested the hellishly expensive FusionIO PCIe devices along with a ton of related flash based memory/storage hardware, including some stuff that is not commercially available (i.e. without contracts/NDA's/...). A good example of the latter --well at least a publicly known example-- would be the
CurtisWright battery backed NVRAM cards [1].
The trouble with this particular flash backed RAM device is manufacturing costs, particularly your BOM (Bill of Materials - components). It's not a terribly difficult or expensive device to make, so there will be competition if it proves useful, and now the important thing is to quickly achieve volume sales to reduce mfg/parts costs as well as competition. Initially targeting "enterprise" for its high profit margin is most likely just a business strategy to get started, with cheaper products planned for the future.
For example, FusionIO started off with extremely expensive "enterprise" products, and now that OCZ and others are entering the market with roughly similar "PCIE SSD" products for the high-end but "mass" market, the company with the ability to ramp up production and reduce costs will be the eventual winner.
And lastly, no, spending $15,000 for a single "hard disk" from FusionIO is certainly not bragging rights. It was mostly just stupidity, albeit forced by a need to test the most recent hardware. When I was a kid in the 70's, all hard disks cost that much. ;-)
>And lastly, no, spending $15,000 for a single "hard disk" from FusionIO is certainly not bragging rights.
I'm pretty sure that middle management feels differently. I also evaluated a FusionIO, and told my client he'd be /much/ better off with a few intel X25-E devices instead. Much cheaper, and it's standard, so it's much easier to upgrade once newer/better stuff is out, or to get a replacement if it breaks after the support contract runs out.
The FusionIO was faster, but both of them were so much faster than the spinning disk we were moving away from that I don't think it would have been noticeable in production.
