DAS, or traditional Direct Attached Storage, is connected directly to a laptop or a general purpose file server and it is difficult to share DAS storage between machines. NAS, or Network Attached Storage was designed to satisfy the need to easily share data between multiple users on different file servers and laptops, by storing the files in a centralised location. However, NAS is not just disk storage, it also includes an operating system and software for configuration and file mapping. The file systems reside on the NAS device, so NAS delivers files to servers. Contrast this with a SAN or Storage Area Network. SANs work on a private, usually fibre channel network and connect storage devices to servers with switches. SANs transfer data in fixed size blocks and the file systems reside on the servers. You can usually have a basic NAS device up and running in 20 minutes, but it is also possible to tailor a NAS array to optimise performance with databases and other applications. NAS vendor's website usually contain several white papers that explain how to do this.
A NAS unit is usually not designed to be a general-purpose server. For example, NAS units usually do not have a keyboard or display, and are controlled and configured over the network, often using a browser. A full-featured operating system is not needed on a NAS device, so often a stripped-down operating system is used. The hardware that performs the NAS control functions is called a NAS head or NAS gateway. The clients always connect to the NAS head, as it is the NAS head is addressable on the network. Disks and in some cases tape drives are attached to the NAS head for capacity. NAS Heads are also sometimes called NAS appliances, based on the ideas that NAS is a commodity item like a toaster or washing machine.
NAS removes the responsibility of file serving from other servers on the network. They typically provide access to files using network file sharing protocols such as NFS, SMB/CIFS, or AFP. The most popular NAS protocols are
NAS can can also be used for load-balancing and fault-tolerant email and web server systems by providing storage services. NAS devices are availabe for the home consumer market, typically for storing large amounts of multi-media data. Such consumer market appliances are now commonly available. Unlike their rackmounted counterparts, they are generally packaged in smaller form factors.
NAS boxes have been around for a long time and are eaay to install, which means that a company might now have hundreds of different NAS file systems to manage. The effort required to keep all the devices patched, and the downtime needed for patching or adding new servers can be considerable if each NAS is managed individually. Once you fill up your first NAS box, you will buy another one, but then you need to change all your users drive mappings to see the second box and this is not a trivial exercise if you've got 2,000 users. Then, when that second box reaches capacity you add a third and have to change the mappings again, which becomes a management nightmare.
One solution to this problem is to add a logical layer of intelligent switches, or maybe a single appliance, between servers and NAs boxes, and so create a global namespace that spans multiple file systems but appears to end users as a single local drive on their computers. This can also automatically balance out I/O and capacity between servers, and you can increase volume sizes or move storage volumes without disrupting users. You can also migrate the data from one box to another with just a brief outage while you change the global namespace to point to the new box. The customers do not even know about it because it's all done behind the scene.
Like SAN virtualisaiton, NAS virtualisation has two approaches. Some products, Acopia for instance, use switches that sit in-band between end users and file servers or NAS devices. Other products such as NuView's StorageX software reside on an out of band Wintel server, and act like a postal service, directing files to the appropriate NAS device or volume within a file server. Which is best? It is possible to argue the case for either approach, but it is arguably more important that the device that you select simplifies the management of all of your NAS devices, rather than whether it is in-band or out-of-band devices is moot.
One problem with managing multiple NAS devices is that file servers based on Windows use the the Common Internet File System protocol to communicate over a network, and servers based on Unix or Linux use the Network File System. The consequence is that Unix and Windows environments need to be virtualised separately.
If you are upgrading your NAS storage, try to purchase enough capacity to last you for the length of your product replacement cycle, usually 4 years, then add a little more as you will certainly need it. Consider limiting the amount of data each user can store on their 'home' drive to try to control runaway data growth.
Tier your data, keep active, mission critical data on high performance (and high cost) drives, and inactive, or less important data on cheaper, slower drives. We could define 'active' as data that has been accessed in the last 30 days. Of course, you want the data to be moved between tiers automatically, so your NAS provider should supply software that does that for you. You could also consider permanently storing less important data, like user's home drives, on cheaper disk.
If you can, use thin provisioning. Thin provisionng means that the amount of data allocated to a drive is only the amount actually being used. So, if you define a 500GB drive, the initial amount of storage allocated to that disk will be quite small, but it will grow as more data is added. The Tiering software shoud alert you when you are getting close to the 500GB limit, so you can consider adding more disk space.
Take snapshots and backups of the data. Snapshots are perfect for recovering deleted files and directories as they are so much faster than recovering from a backup. You just look at the snapshot, then copy the needed data back to the base directory.
However snapshots do not replace backups, as if you lose the array you lose the snapshot too. The data must be backed up to tape or a different disk to cater for loss of the array.
back to top