This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
2. - Evaluating clustered NAS systems: Read more in this section
- Can clustered NAS aid NAS management?
- How to evaluate clustered network-attached storage offerings
- The difference between clustered and traditional NAS
Explore other sections in this guide:
NAS was a godsend to organisations struggling to manage disk drives in multiple servers, providing many with their first taste of shared storage. But adding NAS units can bring NAS management headaches.
In this podcast, SearchStorage.co.UK site editor Sue Troy interviews SearchStorage.co.UK bureau chief Antony Adshead about the dangers of NAS sprawl, NAS management technologies and clustered NAS.
Download for later:
- Internet Explorer: Right Click > Save Target As
- Firefox: Right Click > Save Link As
Troy: What is NAS sprawl, and why is it a bad thing?
Adshead: Let’s be clear: People buy NAS devices because they offer efficiencies. They’re simple to set up and manage, and they provide a repository for departmental and branch office file data that can sit on the LAN. You don’t need any specialised storage skills, and no Fibre Channel fabric knowledge is required.
The problem is, however, that their easily gained benefits can soon become something of a tender trap. They’re so easy to add to an environment that this facility can soon turn into quite the opposite.
Your first NAS devices are vastly more easy to manage than a load of scattered directly attached disks in servers, but keep adding to them and you end up with a disparate, siloed collection of storage just as bad as those server drives all over the place.
So, welcome to NAS sprawl. What you have now is a bunch of NAS devices each with their own file systems, their own application dependencies, data duplicated across them in many cases, data locked in in other cases, not searchable except singly, all requiring their own management and patching regimes and migrations to keep them working sweetly.
The cause of all this is that traditional NAS devices do not scale out as you add to them. They’re each discrete units with their own controller and file systems. We can contrast that with clustered, or scale-out NAS, which does offer a gridlike ability to grow as you add devices, but we’ll come onto that later.
Troy: What NAS management methods and technologies exist?
Adshead: There are two key methods of dealing with it, each with different characteristics.
First of all, there are the NAS management tools that are mostly available from NAS vendors. These offer management features, things like automated device discovery; management, alerting and reporting; storage provisioning; RAID management; replication; backup app support, thin provisioning and security features.
But, there’s a giant hole capability-wise in these NAS management products, which is that while you may be able to manage many devices from one screen, you will only be able to manage at a per-device level. These tools don’t address the fundamental inability to scale out devices and file systems and to overcome the silo effect of NAS sprawl.
So, the other key approach to dealing with the issue is file virtualisation. This technology, once boasting next-big-thing status and several vendors, is now only sold by one company, F5, or possibly two, as EMC still has its Rainfinity appliance.
So, what file virtualisation does is puts an appliance between a bunch of NAS devices and allows all the files on them to be managed as one giant file system. It would, on the surface, appear to be quite a useful technology, but given its demise in the marketplace, there’s either no demand for it or the vendors have largely decided it’s not something they want to promote.
Troy: Can NAS management really succeed?
Adshead: I guess what I’ve been leading up to in my previous answers is that by far the best way to manage NAS is to have clusters of scale-out NAS boxes. These join together to form a gridlike collection of nodes, which aggregate processing power as well as providing specialised parallel file systems that also grow as you add nodes.
The way they can be added to, with new components becoming a part of the whole, means you can add a processing node if you become CPU- or I/O-bound. Or, you can add disk if capacity is an issue. But, critically, what sets scale-out NAS apart is that a sprawl of siloed NAS devices becomes a thing of the past; you can see every file from every device and one node in the cluster can fail and you will still have access to all your data.
Currently, traditional, non-scale-out NAS is still the most prevalent form. But, given that what enables scale-out NAS’ abilities is software in the controller OS, it should only be a matter of time before clustering capability comes as standard with NAS. In fact, the tide has probably already turned with NetApp’s embrace of what appears to be true clustering in its latest OS release, although that will take some time to filter through to real installations.