Friday 12 March 2010

Intel's NAS-specific Atom platform hastens PCification

Intel's announcement last week that the company is planning two versions of its Atom processor specifically for the NAS market was easy to overlook. After all, there are a few Atom-based NAS options on the market already, and the new single-core D410 and dual-core D510 aren't really different from their netbook counterparts in anything other than their target platform. But the roster of vendors that has already signed on to produce Atom-based NAS devices—QNAP, Synology, and LaCie, among others—gives a glimpse at the fact that the home/SOHO NAS market is one place where Intel is definitely poised to take significant marketshare from ARM, and in the near-term. This trend toward x86-based NAS will be great for consumers, because it will hasten NAS's integration into the home network.

First, though, a quick note about the Intel hardware. The main thing that makes the new platform specialized for NAS is the amount of I/O hardware on the southbridge: six PCIe lanes, 12 USB 2.0 ports, a port multiplier function, and eSATA ports. This would be overkill for a netbook (compare Pine Trail's two PCIe lanes), but for a NAS that may host a number of peripherals, it's perfect.

If you're thinking that an x86-based machine with this much I/O and storage is essentially just a headless "PC," you're right. But the line between NAS and PC got blurry a long time ago, at least in the home/SOHO market. We're already well into an era where NAS boxes are competing primarily by multiplying software features, and many vendors have moved into a new phase of differentiation via add-on services like cloud backup. Despite the fact that many of them use ARM-based chips and have very limited CPU and RAM resources, a NAS is now a Linux PC on your LAN in all but name.

Going x86 and joining the network

If there's a downside to the NAS becoming more PC-like, it's that NAS vendors typically load their products with a boutique, in-house, Linux-based OS/application stack that's... well, "workmanlike" and "adequate" come to mind as descriptions. A trip through the support forums of leading NAS vendors will show that the platforms can be quite buggy in real-world use—showstoppers are rare, but there are plenty of niggling bugs (especially where add-ons are involved) that require hacks and workarounds. Web interfaces are often klunky, and not everything "just works."

ReadyNAS is one example of everything mentioned above, and the company is already moving in the direction of x86 for its NAS line. But the x86-based ReadyNAS Pro commands an enormous premium over the rest of the ARM-based ReadyNAS line. Though they haven't announced an Atom-based product, Atom will provide a much less expensive route for the company to make the transition to x86 across the rest of its product line.

The first and most obvious advantage that the NAS market's shift to x86 confers on both vendors and end-users is that vendors can opt out of the software race entirely and just use Windows Home Server. WHS has been very well received, but so far its reach has been restricted because you have to shell out for a real PC. The new Atom-based NASes will change that, and will let users get into WHS without paying much of a premium over the current cost of a NAS.

If users or vendors don't want to go the Microsoft route, they can also adopt and easily tailor any one of a number of popular Linux distros.

DisARMed

Regardless of what ends up being the market's ultimate preference, it seems likely that the current proliferation of boutique, vendor-specific, ARM- and Linux-based NAS OSes is probably not long for this world. The enormous legacy code base of the world's most popular ISA may not give x86 an edge in mobile phones, smartbooks/tablets, or GPUs, but the PC is its home turf, and insofar as that's what the NAS is rapidly becoming, Atom seems ideal for it. In the future, then, NAS vendors will make their value-added software contributions at higher levels of the stack, focusing on drivers, service and support, UI, add-on applications, and networked services (e.g., cloud backup).

My ultimate hope for x86 NAS is that OS vendors like Apple, Microsoft, and even Google will embrace it and integrate it seamlessly into the user experience. The NAS should be a local, largely transparent cache for my and my family's cloud data, and not something that I have to manage as a storage volume in its own right. Hopefully, the fact that client and server hardware are now on the same architecture will hasten this development by lowering the development cost for such integration.