Over the last several years, storage hardware has become increasingly sophisticated. With so many new features and capabilities introduced each year, data storage administrators may have a tough time determining which storage offering will best fit their organization's needs: NAS, DAS or SAN.
What is NAS?
NAS is a mid-level storage offering. At its simplest, a network-attached storage device is an appliance that connects directly to the network. Like other devices on the network, companies assign a NAS appliance with an IP address. Communication between servers and network-attached storage is usually based on TCP/IP. NAS uses a traditional file system such as New Technology File System (NTFS) or NFS, and the storage is accessible at the file level -- usually through a file share.
New storage users sometimes confuse SAN and NAS. The acronyms are similar, and they both refer to types of networked storage. The main difference between SAN and NAS is that a SAN is a higher-end storage offering. A SAN refers to a network dedicated solely to providing block-level access to storage hardware. In contrast, a NAS device is network-connected, but a dedicated network is not required.
Another key difference is that storage area networks usually use Fibre Channel (FC), though there are exceptions. IT arranges FC switches in such a way as to provide redundant access to storage hardware. In addition, server hardware in a SAN environment uses the SCSI protocol to communicate directly with storage hardware, whereas NAS devices rely on IP-based communications.
DAS is storage that's local to a server. The term DAS can refer either to a server's internal disks or to an external storage array that directly attaches to a server via a storage controller or USB connection -- as opposed to being attached through a network connection.
Do I need NAS?
Before implementing NAS, organizations must consider how they will use the appliance. NAS is general-purpose storage: It's a great choice for some cases, but it may not be the best option in other instances.
NAS is generally accessible at the file level, which makes it a good choice for storing unstructured data. If a company is looking for a storage appliance to replace an aging file server, NAS is a good option, but not all NAS appliances support the use of NTFS permissions.
Depending on the device, NAS may not be a good option for storing structured data. While there might not be anything physically stopping companies from storing a database on NAS, a NAS appliance might not be able to deliver the performance the database requires to operate efficiently. Even if the appliance is equipped with fast disks, the network connection will usually be a limiting factor. Because of performance limitations -- and other factors -- many application vendors don't support NAS for use with their database-driven applications.
NAS and SAN can also be used for shared storage, which attaches multiple servers to a common storage device. For example, failover clusters commonly use a cluster-shared volume, which resides on a shared storage device. This enables all the cluster nodes to access the same data.
When deciding whether using a NAS appliance for shared storage is a viable option, companies should look at the software vendor's requirements and the capabilities of the NAS device. Some failover cluster architectures require iSCSI connectivity to the cluster-shared volume. If the network-attached storage device supports iSCSI and the software product the company is using doesn't specifically prohibit the use of NAS, then network-attached storage can be used as a shared-storage medium. Keep in mind that NAS doesn't provide the same level of performance as a SAN.
Some organizations consider NAS as an option for storing virtual machines (VMs). On the surface, this seems like a good idea because NAS devices are generally high capacity and low cost -- at least comparatively speaking -- and IT can configure them to provide fault tolerance. However, companies should consider two factors before placing VMs on NAS:
- Network connectivity is often a performance bottleneck for NAS appliances and may cause VMs to perform poorly. Performance can also be a problem if the NAS appliance’s disks all attach to a common controller, because the disk controller can become a bottleneck. Remember that storage I/O is usually the decisive factor that determines VM performance.
- Vendor support. Microsoft, for example, allows companies to store a Hyper-V VM on NAS, but only if the NAS device supports SMB 3.0.
One of the most common uses for NAS is as a backup target. However, NAS appliances can't usually provide the block-level storage access required by some backup applications.
Another consideration when deciding whether to implement network-attached storage is some NAS devices don't support IP version 6 (IPv6), which could be a problem for organizations transitioning -- or planning to transition -- from IPv4. IPv6 support among NAS appliances has become far more common in recent years, but isn't yet universal.
Scalability is another concern. While SAN environments can be pricey, companies can scale it to ridiculous levels. NAS devices have a finite storage capacity that is limited by the number of drive bays and, in some cases, by firmware. Although products exist to turn collections of NAS appliances into scale-out file servers, it can be a complicated process, and not every NAS appliance supports scale-out architectures.
Organizations also need to consider their options for protecting against NAS failure. As a general rule, they should protect their NAS deployment against both disk and appliance failures. Nearly all NAS appliances support RAID or erasure coding, either of which IT can use to mitigate the effects of a disk failure.
Because a disk controller failure can cause the entire appliance to fail, companies should consider replicating the NAS appliance's contents to a secondary NAS device. This will double the costs since two appliances are required, but doing so is necessary as a way of ensuring that the company's data remains online and available in a failure situation. Not all NAS appliances include native replication capabilities, so companies should be sure to check that the appliance they're considering supports replication.
One more consideration is data deduplication. NAS devices have a finite capacity, so it's important to make efficient use of the available storage space. A deduplication engine can often shrink an organization's physical storage footprint, thereby reducing storage costs.
What class of NAS device should I buy?
Network-attached storage devices vary widely in terms of cost and capability. For those who might be trying to decide between a network-attached storage device geared toward the SMB market and an enterprise NAS device, there are two things to consider:
- If capacity needs are relatively modest -- up to approximately 24 TB spread across four disks -- the company might be able to use an SMB NAS. Higher capacities typically require a higher-end product.
- SMB-oriented NAS devices often consist of a network interface card (NIC), a storage controller, a power supply and not much more. Although developers can usually configure such devices to protect against a disk failure, the failure of any other component will usually take the appliance offline. In contrast, enterprise NAS devices may provide redundant power supplies, redundant -- and possibly bonded -- NICs, and often include a replication feature that allows the appliance's contents to be replicated to a standby appliance.
As shown, a NAS device may not be a good fit for every organization. The key to using NAS effectively is to purchase a network-attached storage device with a feature set that's well-matched to the organization's size and functionality requirements.
NAS, SAN or hyper-converged: Comparing the costs
Scale-out NAS systems can combat performance problems