This Essential Guide is a great starting point for evaluating the market and deciding what type of network-attached storage (NAS) system is right for you. I’ll run through the key areas we cover and suggest some technology changes I think we’ll see in the coming years.
In this guide we divide the NAS market into three key levels.
Firstly, there is the high-end NAS market. This is characterized by the use of scale-out or clustered systems. These stretch to capacities in the tens of petabytes with massive throughput levels and parallel file systems that provide access to all files on all connected device nodes in the cluster. Crucially also, adding nodes adds processing power and/or disk capacity. This end of the market is driven by the need to store huge amounts of files, including virtual machine images, as well as the need for rapid access that multiplies to IOPS rates in the hundreds of thousands per second as nodes are added.
Next are the entry-level and medium-sized traditional NAS products. These can run to capacities of several hundred terabytes but do not have scale-out capabilities, so there’s the risk of running up against the limits of processing power or memory; the file systems of these arrays typically don’t span multiple units so you may end up with siloed file systems in each device. That said, there’s a healthy NAS midmarket and vendors often include advanced features such as forms of replication and block access via iSCSI, which effectively makes many of these products multiprotocol storage.
Finally, there is desktop NAS, which is really just the smallest capacity and performance end of the traditional NAS market. These devices are aimed at small businesses and larger ones that need local storage in an office or remote location. And here again vendors are providing features that were once only found in enterprise devices.
Download this free Essential Guide to NAS management
Also, strictly speaking, part of the NAS world is HPC (high-performance computing) storage, which also deals in very large amounts of files and extremely high performance but that’s outside the scope of this guide.
Those are pretty much the main contours of NAS technology at present, with two things that stand out that indicate future directions of development.
The first is rise of scale-out capability, which is something that ought to become commonplace in devices at all levels of the NAS market. There seem to me to be good reasons to do so. Even if customers aren’t demanding it yet, it should be a selling point for NAS vendors. After all, scale-out capabilities allow customers to add further units that build capacity and performance, while simultaneously providing an easier-to-manage cluster of file storage rather than a collection of siloed devices, each requiring individual administrative attention.
At present it’s only the high end that does scale-out, but it’s a feature built into controller operating systems that should be portable to devices down the range. If I was a customer for midrange NAS I’d want to know why I wasn’t being allowed to build on my investment as I added further hardware from a vendor to my estate.
The second thing is the likely development of cleverer ways of providing better performance. At present NAS performance is largely enhanced by buying bigger trad NAS boxes or adding processing power and throughput with the addition of nodes in clustered NAS configurations. But, there may well be more efficient ways to speed access times and throughput rates.
In the SAN world we’re seeing some newcomers do interesting things that combine flash, spinning disk and sometimes data deduplication. The principle here is to put the most used data on the fastest storage media and shift data between those different tiers; Tintri, NexGen and Nutanix do this, for example.
NetApp applied the idea of using a flash cache for the most used data in a NAS filer some time ago. Now Avere has taken that principle and applied flash caching to hot data across numerous NAS devices. It’s a fairly lonely furrow, but it could be a taste of things to come.
This was first published in August 2012