In common with many fields of computing, storing data on a network-connected server has gone full circle. In the 1980s, Novell led the way in network file serving with early versions of NetWare, whose only purpose in life was to store files in a central location so that they could be shared between users connected to a LAN. Throughout the 1990s, though, server operating systems became much more general purpose, and the emphasis moved away from simply storing files safely and efficiently.
With the advent of Network Attached Storage (NAS) the focus has now moved back to the task of file serving. The reasons are plain to see: although we need all these client-server applications (email servers, gateways to mainframes, computerised fax, and so on) we also need to be able to share files efficiently. This means having a file server that spends all of its CPU time doing file serving.
What is a NAS box?
NAS devices are, to all intents and purposes, dedicated fileservers. Now we’re not talking about Storage Area Networks here – they’re a much more complex concern, are big and expensive, and link to the world via Fibre Channel connections and super-efficient information exchange protocols. Instead, NAS units generally connect to the company network via a normal Ethernet connection (either 100Mbit/sec or, preferably, Gigabit Ethernet) and they talk ‘normal’ file sharing protocols.
By ‘normal’ we mean ‘the same as most networks’. So these units are able to pretend to be any or all of Windows NT/2000, NetWare, Unix or AppleShare servers, and thus client computers can access files stored on the units via whatever protocols are supported. Of course, there’s nothing to stop you simply running up Linux on an old PC and turning on Samba (for SMB – Windows – file sharing), NFS (Unix) and Netatalk (for AppleShare), but the point is that in a NAS device you have a box that is specifically tuned for serving files efficiently.
The other side to dedicated NAS units is hardware configuration: they usually give some kind of resilience in the form of RAID support (either pairs of mirrored disks or clusters of disks in a RAID5 arrangement), normally with the ability to hot-swap failed disks. Again, this is nothing you couldn’t do with (say) a NetWare 6.0 or Windows 2000 server, but such beasts are usually a compromise (a server is basically a PC, which means most of the space is taken by floppy drives, CD-ROM units, sound hardware, video cards and so on) – with a dedicated NAS box the manufacturer has the chance to optimise the hardware layout for the disks themselves rather than for PC-oriented operation.
As we’ve mentioned, a NAS unit is really just another server, albeit one that’s optimised for file serving. So once connected to the network, most of these units are able to use existing access control mechanisms for their security and user identification requirements. So just as the actual file sharing is achieved through standard means such as SMB, NFS or AppleShare, so these devices can also integrate with directory services such as the NetWare Bindery, Microsoft Active Directory, Unix NIS or Novell NDS. This means, of course, that they can generally be managed through whatever tool you already use to look after your user database rather than needing their own (possibly proprietary) configuration manager tools.
Now that we’ve mentioned the key points of NAS, there’s one final aspect that you should be aware of: some companies like to jump on bandwagons and flog you a label, not a product. So beware that some of the ‘NAS’ devices you’ll see on the market are little more than pre-configured Windows 2000 servers that look, and operate, just like a home-built box running Windows 2000 would. We’re not saying, of course, that these boxes won’t do the job – just that if it looks like a server and works like a server, you should ask yourself what’s special about it, whether the hardware is as robust as something that’s clearly been built as a dedicated file serving unit, and whether you could build your own identical server for less money.