Back office
What's going on in networking, operating systems, servers, and data centres?
Tuesday 27 October 2009, 9:03 AM
What does storage virtualisation really mean?
It means you can use the term how you like. Perhaps the best-known use of this technology is tape virtualisation, where your automated library of tapes takes care of exactly where your backups are stored, and just returns the data you want. Eventually. In theory.
On the other hand, virtualising rotating media is less well-known, mainly because it means different things to different people. Here at Storage network World in Frankfurt, the most – er – innovative, but perhaps instructive, use of the term so far was by analyst Tony Lock, who pointed out that every file system is a form of storage virtualisation.
How so? At heart, virtualisation is about hiding the underlying hardware from the software and management layers above, allowing the technology to take care of the intricate details of where an application is run, or data is stored. Given that definition, a file system qualifies.
So even when I was running a BBC Micro back in 1982, it included storage virtualisation. Who knew?
So when the vendor sidles up to you and whispers seductively about storage virtualisation, make sure you're both talking about the same thing. I'd be interested to learn what storage virtualisation means to you. No, really...
PreviousComments on this post
Nice post Manek,
I'd be interested to know what these slippery sliding vendor's first line is when they try and extol the virtues of storage virtualisation.
Is it "turn off and onable" storage in a sort of an On Demand as and when you need it with only the power you require yet with more management capability - from your post it seems that it might be.
The door is open for more sales of storage virtualisation management software then - or has HP already stepped in there an owned that space?
Adrian
Interesting article
To your question Adrian, the turn on and offable storage sounds a bit more like thin provisioning - lots of guys are doing this but as you say, its the turn it on only when you need it approach.
Like Manek says - i kind of see storage virtualisation like server virtualisation in that you dont really necessarily know where your data / VM physically is, you just know its there. In the same way that server virtualisation bods are talking about creating a 'software mainframe', storage virtualisation software can potentially do the same on the storage side by just pooling the storage resources and striping data across multiple arrays.
To me that doesnt sound like rocket science but once the marketeers (and damned PR people) get involved it can be a different story. Its been around for ages - just like some folks forget about virtualisation having been done by IBM donkeys years ago,
You could go way, way back to the AS/400 or its predecessor, the System/38, where there was no concept of separate memory or disks, just a "Single Level Store".
As a developer you never needed to manage where your data was, either in main memory or on disk, since the operating system did it for you.
As a user, you knew to find stuff you looked in libraries, but a library is just an index of objects, and doesn't represent any physical or logical structure on disk.
A single level store blurs the line between virtual memory and virtual storage. With greater use of virtual machine architectures like the JVM and .NET CLR, I think we're going to see more of these kinds of techniques making their way back into use.
I hope so.
I note there's a nod in that direction with Vista/Win7's concept of libraries: logical folders with no real existence other than as pointers to some other areas on the disk. Pity MS screwed up by being amazingly restrictive about what you can put in the libraries. Try adding a network location, for example...
