Network attached storage (NAS) is one of the fastest-growing sectors of the data storage market. For enterprises and other large organisations, NAS provides an efficient way to store ever-larger volumes of unstructured data, from media content and documents to AI training materials.
For smaller outfits, NAS offers a step up from internal storage in servers, providing familiar files and folders arrangement and largely similar management tools. They also run over the same standard Ethernet networks as existing PCs and servers. At the micro-business end of the market, a NAS box might even take the place of a server altogether, due to their simplicity and low cost.
NAS systems, however, have their limitations: access times are slower than direct-attached – or server-based – storage, and they will often be slower than a storage area network (SAN), not least because they use Ethernet rather than faster transports, such as Fibre Channel.
But the main drawback of traditional NAS is that it’s harder to scale. Once a NAS box is full, IT departments had to upgrade disks, or buy an entirely new NAS. This leads to high costs, or – if they keep the old NAS – silos of storage.
This prompted vendors to create “scale-out” NAS systems, where users can build a mesh of NAS nodes to form a single pool of storage, and it prompted vendors to create NAS functionality in the cloud.
What is NAS?
NAS is file access storage that can be made available to any device on a network. At its most basic, a NAS box is a standalone, integrated device with a CPU, networking, memory and a number of bays for hard drives, SSDs or M.2 drives. It has its own operating system, and does not need a connection to a server or storage controller to function. Most admin tasks can be carried out via a web browser.
The heart of NAS is its file system. The file system allows the NAS to appear as a shared volume to any PC user or server on the network. Typically, NAS systems support Network File System (NFS) and Server Message Block (SMB) protocols, for maximum application and OS compatibility.
A NAS is optimised for unstructured data – files and documents – whereas direct attached storage and SANs are optimised for block access, such as for databases.
It’s the integration of the file system that defines NAS, in fact. When using a SAN, the file system resides elsewhere – such as on servers or PCs – with data held on the SAN in blocks accessed from those externally-held file systems.
The how and why of NAS
NAS devices start small, but can be very large indeed. Small NAS boxes are popular for home offices and micro businesses. At the other end of the scale, enterprise vendors provide NAS units with eye-watering capacities. Chinese vendor Huawei, for example, recently launched its OceanStor Dorado V6, which runs to 32 controllers, 6,400 drives and almost 300PB of capacity.
The first enterprise NAS systems were “filers”, designed to deal with the growing volume of unstructured data coming into organisations as they moved to digital processes.
Filers from vendors such as NetApp and EMC (now owned by Dell) mopped up vast numbers of files in areas from medical imaging to media and document storage. Without filers, these documents would have overwhelmed server-based storage.
But the current landscape is far more complex. More applications use unstructured data and the performance of NAS hardware has increased with faster versions of Ethernet and the move to flash storage.
Scale-out NAS systems, with multiple nodes operating as a single, very large (virtual) volume, or single namespace, have also overcome some of the limitations of traditional standalone NAS. It is now possible to add storage quickly to a NAS system without applications or users being aware. IT teams can add nodes, and storage, as needed. A global namespace also allows enterprises to distribute storage across locations, including the public cloud.
Two other versions of NAS technology are also worth considering. First, for enterprises, hyper-converged infrastructure (HCI) brings together compute, storage and networking. In some use cases, HCI can replace NAS and SAN systems. At the other end of the spectrum, open-source utilities such as FreeNAS can turn redundant servers, or even desktop PCs, into cost-effective NAS boxes for smaller offices or home use.
The focus of filers and early NAS systems was capacity, not performance. Today, there is still a case for NAS for very large capacities. Few other systems can offer the same low cost per gigabyte of data. But conventional, capacity-driven use cases such as backups, archiving – including for compliance reasons – and handling relatively low-value documents have been joined by more demanding applications like storage for AI/ML training data and advanced analytics datasets.
This trend has also pushed NAS towards convergence with object storage capability, and HCI.
What is cloud NAS, and how to access it?
A cloud NAS uses public cloud storage to provide file access, usually via SMB or NFS. NetApp’s ONTAP, which includes NFS, is also widely supported.
Connecting a user device, or more likely a server, to a cloud NAS should be as simple as connecting it to local NAS hardware. The differences will largely be around capacity, cost, performance and the level of intelligence offered by cloud systems.
Cloud NAS has no real capacity limits although vendors may set an upper limit for practical administrative purposes. Google Cloud Platform, for example, limits a single namespace to 50PB.
The main performance limit of cloud NAS is the WAN connection to the service provider. If applications need high speeds and low latency, firms will need to invest in high speed fibre links. Alternatively, they could relocate applications to the cloud, if the provider supports compute. But on-premises systems are – for now – faster and much less costly.
When it comes to cost comparisons, customers need to consider that on-prem NAS hardware requires up-front capital investment. Also, firms often also have to buy more than they need and over-provision to minimise the cost and disruption of physical upgrades.
This is not an issue with the cloud, as it is charged by usage, but CIOs will want to control this. Long-term cloud storage can be expensive if users “file and forget” documents.
Cloud NAS use cases
For this reason, the large cloud vendors and vendors such NetApp and IBM sell products aimed specifically at the backup and archive markets. These have more favourable cost structures than regular cloud NAS volumes.
Vendors also offer different tiers of cloud NAS, based on performance. Less important workloads or less frequently used data can move down to cheaper volumes. But, in all cases, buyers need to be aware of bandwidth and egress costs to access or download data.
However, there are use cases where cloud NAS is the best solution. These include where a business needs to scale up storage quickly without rewriting applications or porting them to the cloud; or where cost is less of an issue, such as for a short-term project; or to provide remote storage, collaboration and backup to small offices or remote workers.
Read more about NAS and unstructured data
- Unified file and object storage: Three suppliers and their approaches. We look at unified file and object storage products from NetApp, Pure Storage and Scality, the differences, the workloads aimed at, and how unified they actually are.
- Unstructured data storage – on-prem vs cloud vs hybrid. We look at storage for unstructured data on-premise, in the cloud and across multiple locations. There are advantages to a hybrid approach, but there can be hidden costs too.