Network Attached Storage (NAS)

Network Attached Storage (NAS) provides large amounts of file space as a network resource. In some respects, NAS is little more than a self-contained and highly optimized file server. NAS units support the same network storage protocols as network operating systems do. Many high-performance NAS disk arrays use proprietary operating systems; others use embedded versions of the Windows or Linux network operating system.

NAS has several advantages over a traditional file server. Some characteristics that make NAS attractive to system architects are the use of open protocols for shared file services, ease of installation and management, scalability, and high performance. The ability to install and manage NAS devices easily also translates into lower operating costs. A file server may take hours, or even days, to get up and running in its most basic form. A NAS system might do the same in minutes.

Another important advantage is that a NAS device does only file services. At first this seems like a small thing, but it makes a difference when it comes to maintenance, performance, and uptime. NAS devices are unencumbered by the extra capabilities that a general-purpose operating system has to have. Everything that is not related to efficient file serving can be stripped out. There is also a lower chance that a nonstorage program will crash the system, because none can be installed.

In the early days of networked storage, there was a tendency for storage vendors to argue over which was better: Network Attached Storage or Storage Area Networks. The debate was silly and unproductive. Both are useful for different applications and must be viewed by themselves, not in opposition to each other.

Storage Appliances Have a Distinct Advantage

NAS arrays are appliances: special-purpose systems that perform one function very well. It is easy to think of storage appliances as being similar to a network router. In the early days of networking, routing was accomplished with software running on a general-purpose computer. Most server operating systems, such as UNIX and Linux, can still perform this function. It was just inefficient. Too many things could happen to the general-purpose server that affected availability or performance. The dedicated router or a router embedded in a network switch overcame these limitations, and the server acting as a router all but disappeared.

The same holds true for NAS and other storage appliances. By focusing on one or two key functions, storage appliances are more efficient.

Many high-end NAS devices have the ability to scale to very large sizes, into the multiterabyte range. This is more than a typical file server can often handle in a single file space. Recent NAS software has the ability to create a single file space across many NAS devices. This feature lets the system administrator connect multiple arrays and have them appear as one file space, expanding the file space by adding new arrays when needed.

For some time, the NAS market has been divided into three spaces: low end, middle tier, and high end. The low-end NAS market has targeted products to the small office home office (SOHO) and as desktop backup devices. These are little more than a large hard drive with an Ethernet network connection and an embedded network operating system. They are small, simple to install, and easy to maintain but have practically no features suited for a large data center. The Achilles heel of products in this category is cost. Although total costs are lower, owing to quick and easy installation and negligible maintenance needs, the cost of acquisition has traditionally been higher than for a general file server with similar capabilities. This has made these products unattractive to small businesses.

Most NAS falls into the mid-tier category and is designed for web servers and departmental file servers. Network Appliance is a leader in this arena, along with storage giant EMC. Arrays in this space often have high availability features, a faster network connection (usually 10/100/1000 BaseT Ethernet), software protocol support for enterprise-level management, and high performance. The storage capacity of these arrays ranges from 250 gigabytes to roughly 1 terabyte.

The high end of the NAS market is defined mostly by capacity and management features. High-end NAS disk arrays can scale up to multiple terabytes and include features such as global file management and virtual volume capabilities.

The File Head

What turns a bunch of disks into a NAS array? It is the file head. Often called the NAS head, it is a specialized server that contains at a minimum a processor, memory, network connections, storage connections (usually SCSI or Fibre Channel), and a network operating system. From the network's perspective, the file head is a file server, communicating via one or more common or proprietary protocols.

The file head may be external to the disk array or embedded in it. In many cases, the file head is physically integrated into the NAS device. The software it supports may be nothing more than the TCP/IP stack, simple web-based management software, and a stripped-down network operating system. On the other hand, it may contain specialized volume management, advanced device management, support for NAS backup, clustering, and other high-end storage software. It all depends on what type of applications the NAS device is targeted to.

NFS and CIFS

In beginning, a lot of NAS manufacturers used their own protocols for communicating between a host and the NAS array. There were certain drawbacks to this, chief among them being locked into one vendor. It also required special drivers, which limited the number of supported platforms.

Soon, most vendors of NAS products began to use two popular protocols for file sharing: the Network File System (NFS) from Sun Microsystems and the Common Internet File System (CIFS). NFS provides for file access and integrity when accessing files across a network. It was created by Sun Microsystems in 1985 and since then has been widely deployed on UNIX and Linux systems. Client-side NFS drivers are available for a variety of platforms, including all major versions of Microsoft Windows.

CIFS is an updated version of the Server Message Block (SMB) protocol, which also dates back to 1985. Originally developed by IBM, it was adopted by IBM and Microsoft as LAN Manager, and forms the underpinnings of file and print services common to all Windows platforms. An open-source implementation of the client and server side of SMB/CIFS called SAMBA is available on Linux and is widely used to create Linux file servers for Windows environments.

Some manufacturers still ship products using their own protocols, often for specialized applications or to gain some performance advantage. Most, however, use NFS, CIFS, or both.

Storage Architectures Using NAS

NAS can be used anywhere that a file server would be deployed. It can be supported on the same network infrastructure as servers or other network devices.

The most common uses for NAS are in web server farms; as departmental file servers; and for storing large files, such as GIS data or engineering drawings, that are typically shared among many users. For these applications, the high performance, efficient design, and ease of use of NAS devices makes them cost effective. NAS disk arrays are often called filers. The term originated with Network Appliance but has become a generic term.

File Server Replacement

In environments where there are limited or no IT resources, the ease of use and simple maintenance required of NAS makes it a worthwhile investment. Very small organizations tend not to have file servers at all because of the complexity of the software and skills needed to maintain them. NAS is an excellent choice for these organizations.

For larger companies, the need to change or add new file servers on a regular basis makes NAS a viable option. The ease of installation means that NAS arrays can be deployed quickly without draining IT resources.

Web Server Farms

One of the most popular applications for NAS is web server farms. ISPs, commercial sites, and large company web sites usually deploy many web servers for performance and availability reasons. By spreading the load over many computers, these web server farms are capable of handling many users simultaneously without performance degradation due to processor loads. Web server farms usually have a very large number of files that must be shared among all the servers. Copying files to many individual arrays is both costly and time-consuming when changes are made. NAS centralizes the files into a single space.

Because NAS is file-based, it matches the structure of the main components of web sites: HTML files. HTML files can now be housed on only a small number of high-performance NAS arrays and remain accessible to all web servers. This is clearly more cost effective and easier to maintain than many individual arrays. Web servers provide a processor and memory; the NAS devices provide file services.

Hybrid NAS-SAN Devices

For a long time, NAS and SAN were independent, often hostile, technologies. Unfortunately, that's not what system architects and storage administrators wanted. The IT community understood one thing that vendors were slow to catch on to it's just a bunch of disks. The network interfaces were no longer much of a concern, especially with iSCSI using an Ethernet network to create a SAN architecture. All that differentiated SAN and NAS disk arrays was the file head and other supporting software. A file head could be placed in front of a SAN to produce a NAS system and iSCSI used to get block I/O over Ethernet.

The data storage industry responded with a series of hybrid devices. First, SAN disk array manufacturers and resellers began to place a file head on the Fiber Channel SAN to produce NAS and SAN together. NAS vendors responded by allowing iSCSI commands to pass through the file head and perform block I/O. These hybrid systems are flexible and make the best use of expensive resources. Most IT managers will buy separate systems when they can, for the sake of efficiency. They deploy hybrid systems when cost control is more important.

Most hybrid arrays require that different volumes be used for the SAN and NAS portions of the disk space. It would not be very safe to allow changes to the data underlying the files without the knowledge of the file system. Reading blocks of data won't hurt anything, but writing could corrupt files and directories. Different protections can be set on volumes to accommodate this.