Clustered NAS is a growing trend. Where traditional NAS has often fulfilled the role of office or department file storage, clustered NAS is a true enterprise technology, offering advantages of scalability with a single view ranging up to petabytes, high performance and ease of management.
In this interview, SearchStorage.co.UK Bureau Chief Antony Adshead speaks with Martin Glassborow, a storage manager with a major UK-based media company who also blogs under the name Storagebod, about the differences between traditional NAS and clustered NAS as well as the key difference between NetApp cluster functionality in OnTap 8 and its competitors’ clustered file systems.
You can read the transcript or download the NAS clustering podcast.
SearchStorage.co.UK: What’s the difference between traditional NAS and clustered NAS?
Glassborow: The difference between traditional NAS and clustered NAS is basically the traditional NAS solutions such as NetApp boxes and the boxes from EMC Celerra are based around a server with CPU, RAM and storage but these work as independent entities, independent servers. We do cluster them, but only to provide failover capabilities.
With clustered NAS we have a number of servers that have disk, RAM, CPU, but the difference is they can see all the data across the whole storage environment. So, you can go to a single IP address or single server name and you’ll be able to see all the data, all the file shares behind it. It doesn’t matter which one you talk to. This has some major advantages when we start talking about scalability and performance and also ease of management.
For performance it means you’ve got many servers serving the same data. For capacity, we have the ability to put a lot more capacity behind these, but also from an ease of management [point of view], all this capacity can be seen in a single pool. That means you’re not subdividing things up into separate file systems. It means we can handle much larger files and many more [of them].
Obviously, this has come from the worlds of media world and seismic processing, [where they’ve needed very large files]; but also in virtualisation, where we now have many thousands of virtual machines, and having them all in a single pool could have some advantages with data mobility. It means a number of servers can all look at the same pool, and from a data mobility point of view we’re not going to have to move virtual machines between storage arrays.
SearchStorage.co.UK: What is it about NetApp’s clustered NAS systems that makes them different from others, and what are the implications of those differences?
Glassborow: NetApp’s solution is built into OnTap 8, [which has] two modes. It has what’s called “cluster” mode and “7” mode. [The latter is] the traditional NetApp mode: a two-filer setup with mutual failover. [NetApp] cluster mode basically allows you to cluster multiple [devices] up to 24 nodes.
But, what it doesn’t have, which all the other competitors have, is a clustered file system. That means all the nodes cannot necessarily see all the same data, but also all the data on the NetApp cluster mode doesn’t reside in the same file system. So, you have lots of little file systems; this really constrains you in [terms] of growth, the size of the file system, the size of files. [Media files, for example,] can be multiple terabytes. I’ve seen media files that were 1 TB or 2 TB in size; if your file system is limited to a size of 100 TB, that’s 100 or 50 files. That’s not particularly useful.
I would expect this to change. I would expect to see NetApp eventually offer much larger file systems, but for the moment it’s a very segmented, fragmented way of doing things. It gives you the advantage of being able to point to a single server name or a single service name, but it doesn’t allow all your files to live in the same file space, whereas [with] the others -- Isilon, [IBM] SONAS, [HP] Ibrix -- their core special sauce is their distributed clustered file system so everything lives in the same space. You can move things round and have a lot less planning and management to do.
This was first published in April 2011