Network storage traditionally resided on hard disks in individual servers. There were two problems with this approach:...
Servers only hold a limited number of disks and network traffic to the server could easily form a bottleneck.
While a SAN overcomes these problems, Fibre Channel components and management tools can be too expensive and too complex for basic network storage tasks. Network attached storage (NAS) offers accessibility and performance that falls between server storage and SAN storage. Traditional NAS devices attach to the LAN through an ordinary Ethernet connection and IP address, providing disk storage as an independent network device (although NAS devices can also be added to a SAN). Multiple NAS "boxes" can be added to the LAN as needed for additional network storage.
NAS hardware and architecture
A NAS box is a dedicated storage server that includes a NAS head and disk drives that are attached to a network. The NAS head (aka NAS front end or NAS gateway) is the control electronics that interface the network and storage. A basic NAS may use just one head or share the internal storage across multiple heads to accommodate increased network bandwidth.
In terms of storage, NAS systems are denoted by their drive support, the total number of drives and the total capacity. SATA drives are by far the most common type of drive in NAS systems. SATA drives allow for low-cost high-density NAS storage, although ATA, SCSI, SATA and SAS drives also also used. Even Fibre Channel (FC) drives can be used in high-end NAS models.
Workgroup-type NAS systems support 1 TB to 2 TB of disk storage (or more) with a small group of four to six hard disks, depending on the model and options selected. (Some models support expansion disk racks that allow extended storage up to 30 TB.) Enterprise-class NAS systems can implement many disks and offer capacities well over 100 TB. Most NAS systems include RAID support for data protection and can implement common RAID levels, including RAID 0, RAID 1, RAID 5 and RAID 6/DP. NAS systems also include some onboard memory (RAM) to cache network data to or from the disks. Small NAS devices may only provide a 128 MB to 256 MB cache, though enterprise-class NAS systems may offer cache up to 4 GB.
A NAS box must actually connect to the LAN, so the network interface is also important. Gigabit Ethernet (GigE) ports are almost universal and 10 Gigabit Ethernet interfaces are on the horizon. Some NAS products provide multiple Ethernet connections for network interface aggregation, redundancy or failover.
As an example, consider the V800 workgroup NAS system from Mpak Technologies. The V800 supports up to eight ATA disks in RAID configurations to level 5. The unit includes up to 512 MB of cache and offers two 10/100 Ethernet ports (GigE ports are an option). By comparison, the StorEdge 5310 NAS from Sun Microsystems Inc. natively supports 18 disks up to 2.6 TB and 4 GB of onboard cache, and includes two 10/100/1000 Ethernet ports for connectivity.
While NAS boxes typically work independently, they can also be aggregated into clusters. Similar in principle to clustered computing, NAS clusters appear to the LAN as a single NAS device. Each clustered element can share the data load and each box in the cluster can provide failover if another box fails, improving storage performance and achieving high storage availability. OnStor Inc. provides a clustered NAS gateway designed to interconnect Windows, Linux and Unix client systems with up to 40 PB of storage capacity from a variety of vendors.
Protocols are sets of rules that define the way that two end points communicate. NAS devices interact with the LAN using various protocols that manage activities like networking, file exchanges and applications.
Networking protocols include Novell Inc.'s IPX and Microsoft's NetBEUI. NAS devices can share and exchange files using recognized file protocols like Sun's NFS or the CIFS open standard based on Microsoft's Server Message Block (SMB) protocol. CIFS complements existing Internet application protocols like the FTP and HTTP. For example, Sun's StorEdge 5310 NAS includes support for CIFS/SMB, NetBIOS, NFS v2 and v3, and FTP.
NAS devices typically run their own proprietary operating system, and are managed and configured using integrated software utilities that run across any standard Web browser. This allows storage administrators to check NAS status, diagnose issues and make changes to the NAS configuration from any workstation on the LAN. Any management tool(s) should include support for the NAS boxes on your LAN -- including discovery -- and should present detailed information through a single console. NAS devices frequently include utilities to tackle tasks such as snapshot, replication or backup. When evaluating a NAS product, it's important to test the utilities included with the unit to ensure that the software will behave properly on your network.
Every NAS box will absolutely require some amount of routine management work, but NAS management overhead is cumulative. As more NAS boxes appear on the LAN, or even the SAN, more management time is required. Eventually, this can strain limited IT staff.
NAS virtualization is one way to deal with management overhead. The same virtualization technologies that can treat a variety of storage resources as a single pool of storage can also be applied to NAS devices -- locating, identifying and combining NAS resources so that they can be treated and managed as a single entity (ideally through a single management console). Dedicated NAS virtualization appliances are typically installed between NAS boxes and the network.
Today's tighter backup windows and shrinking recovery objectives are forcing storage administrators to rethink their tape backup strategies. Disk storage is displacing tape as a backup medium for its enhanced speed, RAID data protection features and cost-effectiveness. These features, combined with ease of installation, are making NAS devices attractive backup targets.
Although any NAS box can be used for backups, some NAS appliances are dedicated to backup tasks. The NearStore system from NetApp is designed for nearline storage from 8 TB to 96 TB using inexpensive SATA disks. When employing a NAS system as a backup target, be sure your backup software supports the network data management protocol (NDMP). Backup software should also be compatible with the version of NDMP in use. More information is available about the NDMP here.
While NAS may provide an attractive backup target for some applications, remember that NAS does not replace tape backups completely. Disk space is finite, and NAS systems will eventually fill with backup data, forcing older backups to be deleted or offloaded to tape. Disk storage also remains on-site, so organizations relying on tape for off-site disaster protection may need to continue periodic tape backups.