Jan 18, 2013 (09:01 AM EST)
Software-Defined Storage: A Buzzword Worth Examining
Read the Original Article at InformationWeek
Little did Rogers and Hammerstein know they were writing about data center infrastructure. Of course, maybe they secretly wanted to ("The Sound of Modems," "My Fav'rite Backups," "IdleFlash," etc.) but the rhyming and levity might have been hard to pull off. Anyhow, the title of this piece is, I hope, not as silly as it seems. There is a serious shift happening in storage, but we haven't got the semantics down yet.
Semantics in the IT and storage business can be an absolute problem. When new things emerge we love to give them names, but pretty soon the generic name can become an impediment to understanding the specific products and values that it's meant to describe. Worse still, we all quickly reach a point where it's even deemed embarrassing to profess to not understand something, so we get the blind leading the blind. We throw around phrases that are meaningless in isolation ("high quality," "solid state"), or seem to connote one thing but are actually generic descriptors ("thin provisioning," "back up"), or that become so abused as vendors scramble to sport the latest fashion ("cloud" being the current favorite) that they are always followed by "whatever that means."
[ For more on SDS, read What Is Software-Defined Storage? ]
The latest semantic fad sweeping the industry is "software defined." It can be followed by just about any word of the IT lexicon; there's SDN (networking), SDDC (data centers), and, of course, SDS (storage). Now, because I'm not a complete cynic, I realize that the adoption and propagation of new market terminology is not some conspiracy by the vendor community to confuse everyone. There is absolute value in the various types and components of SDS. But we need some context and specificity to uncover that value, and even more of that context and specificity to enjoy and benefit from that value in any given user environment.
This is the age-old problem of, "If you don't know where you want to go then any road will do." It might sound obvious, but rushing off to get some SDS because it's the latest shiny object could be a terrible decision -- or it could be a strategic infrastructure master stroke. Which one it is will depend on two things: What you actually got, and what you actually needed.
Even if you attempt some definition of SDS, it is likely to be at a lowest-common-denominator level. Obviously, no one will be going out to buy three pounds of SDS, or asking how many kilometers you can drive on a tank of SDS. Software-defined storage is a concept, an approach, a method. And it is a generic descriptor for a host of specialist and varying implementations. And so you say, what is it? To the best of my knowledge neither the Oxford English Dictionary nor the United Nations has yet declared a definition.
Wikipedia defines it as "a marketing theme for promoting storage technologies that is currently lacking consensus. … SDS is a part of the broader software-defined data center concept wherein all the virtualized storage, server, networking and security resources required by an application can be defined by software and provisioned automatically." I can hardly disagree. Neither the wiki writer nor I are questioning the potential value of SDS -- we're both clearly intimating that there is value – but there's a lack of consensus. We're at that phase of a new thing (the Wikipedia page was only created last August) when vendors are rushing to drape themselves in the new fashion, and are naturally massaging their definitions to suit their ability to deliver.
Here's my definition of software-defined storage: SDS is where the intelligence (and we're not talking about communication tools or firmware) to run a storage device or system exists only as software outside of the physical storage hardware. SDS is not the virtualization of storage resources and functions within or applied to a storage system -- it's a further abstraction. Nor is it a box that you buy; it's software. It will inevitably lead to commodity hardware, and the general rallying cries for marketers will be flexibility, alacrity, hardware-meritocracy and reduced cost.
And where does this new software sit? It could be in many places -- literally an application on a server, or perhaps delivered as part of an operating system or hypervisor, whether storage or server specific. It could be open source or even very vendor restrictive. It could be the epitome of heterogeneous and agnostic public/private/hybrid cloud-storage, or simply a way to allow more flexibility and control to IT generalists in a one-vendor environment. Therein lies the semantic confusion, and yet therein is also a host of operational and financial value.
Although the buzzword SDS is relatively new, we've been heading down the road to SDS (and SD Other Things) for a while. Consider, for instance, that most storage systems used to have embedded code running on complex ASICS, but now the vast majority of storage systems are based on portable software that runs on an X86 box, with commodity disk attached. This isn't something that's broadly promoted by the big storage players yet, but they are all ready, or getting prepared, to play on this new field.
The point is that users need to eschew the generic trendiness and search out specific value, which is based upon achieving whatever IT, business, operational and financial goals they have set. SDS can -- and I suspect will -- be a great step forward in an IT world that increasingly values malleability and bang-for-the-buck. But all of us need to remember that it is a generic term. Sure, we now have a wheel, but whether it is deployed as a unicycle, a car or a lunar module is what will determine its value in individual environments. If you don't start from your precise, known needs, then you might just find yourself trying to hold an IT moonbeam in your hand.
Our four business scenarios show how to improve disaster recovery, boost disk utilization and speed performance. Also in the new, all-digital Storage Virtualization Gets Real issue of InformationWeek SMB: While Intel remains the biggest manufacturer of chips in the world, the next few years will prove vexing for the company. (Free registration required.)