With so many vendors contributing varying definitions and products, people are still asking themselves: What is software-defined storage? When that question was posed to Randy Kerns, a partner at Boulder, Colo.-based Evaluator Group, during this TechTalk from TechTarget's Storage Decisions conference, he said it comes down to whether a product is software that runs within a system or software that runs alongside other applications and that provides storage functions. The latter, according to Kerns, is considered software-defined. If it seems that software-defined storage is just a new label for technology that's already being used, an important difference that Kerns pointed out is the inclusion of management and provisioning capabilities. To find out all he had to say about software-defined storage, how it differs from other technologies, and some of its benefits, take a look at the video or read the full transcript below.
Randy, how would you define software-defined storage?
Randy Kerns: Well, that's obviously a great question, because I've become very critical. There's so many different definitions of it going around today, and there are many companies that slightly alter the definition to suit whatever particular product they have. In our environment we look at it and say, 'Well, obviously embedded software is in everything when it comes to storage.' Inside very traditional architectures, dual controller storage systems, you're running code there. Over time, that's become more generic in that a lot of the storage control functions are done on standard servers now.
From our perspective, we talk about software-defined storage primarily as [a] delineation between storage software that runs in a system versus software that runs coresident with other applications and is doing some of the storage function. We talk about software-defined storage as an application running in a server that does storage functionality.
Hasn't software always defined storage? What's different with what we call software-defined storage?
Kerns: Several things are different, and one is that there's an availability of software now from multiple vendors, whereas [there used to be] relatively few suppliers that could do the storage function. The second one is [that] we have the capability now with processor power and multicore processors and hypervisors inside server systems to run multiple functionalities. Now you can run storage functions while you're running applications on the same server, so we've got an enablement primarily due to processor hardware and with the hypervisors that exist today. It gives us more flexibility, more capabilities than we ever had in the past.
Is a storage hypervisor the same thing as software-defined storage? If it's not, what's the difference?
Kerns: I would say it's not, although you could construe it to be that and certainly you'll get marketing messages that make it appear to be the same. A storage hypervisor, I view it as more of an enabler of abstraction of a storage device. Think of it as storage virtualization software, specifically where that's the function to federate different storage elements and present the logical entities to the applications or the systems.
Software-defined storage includes not only that -- that would be part of it -- but it also includes the data services functions that you may have. It includes the provisioning mechanisms, a whole lot more than just a hypervisor, which is more of the federation type of software.
One of the most appealing aspects of software-defined storage is its relative low cost. What are some of the other benefits?
Kerns: First, low cost may be an illusion, so let's keep that in mind. The other benefits, though, are things to think about, in the fact that you can get to the point where you can manage many different heterogeneous storage elements the same way. You abstract and now you can automate the management much easier.
The biggest issue with storage is obviously growing demand for information storage. That capacity demand, the number of elements, how can you manage it in the same way you've done in the past? The answer to that has always been to try and automate that. If I can abstract that to where my management scope is around those abstracted elements, then it's much easier to scale. If I have to deal with individual devices and those characteristics and those element managers, those can hit at my scaling capabilities, and so the software-defined storage and the abstraction helps move through the automation level. That's one of the bigger advantages, and the cost, like I say, that may be an illusion.
Software-defined storage often sounds a little like a science project or some kind of do-it-yourself kit. Is it only for geeks who like to get under the hood?
Kerns: Well, it's certainly that element, and some of those people are incredibly bright and clever and they do some very unique things and some enter into the mainstream. The problem with storing information is that's the lifeblood of most companies. That information and management of that information, retention, security, integrity is so very important that most companies, any type of business or entity says, 'I can't afford to have that put in jeopardy.' It's really not an opportunity for a science project for many of those companies.
While great things are done, and some unique things [are done], those are typically in isolated environments for the mainstream business. Now we see these types of major vendor companies entering into solutions to get that abstraction, to get that automation, to get that portability, if you will, while maintaining all those other elements necessary to protect information. I don't think it's really a science project, but I look to what's coming from the major vendors with long-term support.