Software-defined storage vendors often market their products differently, making it hard to pinpoint exactly what software-defined storage does. If it seems that storage virtualization vendors are doing the same thing as software-defined storage vendors, analysts say there is a distinction. According to Randy Kerns, a partner at The Enterprise Strategy group, both abstract the storage layer, but products from software-defined storage vendors are the ones that add provisioning and automation capabilities. In this TechTalk from TechTarget's Storage Decisions conference in New York, Kerns discusses software-defined storage options and the types of storage systems they can be used to build.
Software-defined storage seems awfully similar to software-based storage virtualization. Is it the same thing?
Randy Kerns: If you read the marketing materials, it is. I think we're going to lose any distinction shortly. I would have said up until recently, 'Yes. It's a totally different thing,' but right now, most of the information's been [used interchangeably] between the two, it's hard to distinguish between them. We talk a lot about storage virtualization, primarily at the abstraction level, and then put that automation and provisioning up to the software-defined storage level. That seems to be a very good way to talk about it, but it's getting so muddled now, I don't think [software-defined storage] is going to have an accurate portrayal anymore.
Some of the better-known software-defined storage products are kind of limited relative to the number of nodes and [the] capacities that they support. Are there also enterprise-scale products that are available?
Kerns: Yes, and certainly you've got a number of vendors now in this game. Certainly the recent announcement by EMC with ViPR is a great example of that, and there is really no limit in scale. Some of the other solutions do get very large, so I don't think that the scaling amount is accurate anymore. I think that we've moved beyond that, certainly with some of the big announcements recently.
What type of storage systems can you create with the software-defined storage product? Could you block, file, object, all of the above?
Kerns: Pretty much [all of the above], but you've got to remember there's different purposes for all the different storage technologies. Block storage is typically what we see used primarily with databases, and then everything else ultimately resolves down to block access, so everything would be blocked in the finality of accessing data, but the abstraction of a file system is a convenient state [for data to be stored].You can have the opportunity to organize data in a hierarchical file structure and access the files, and then set permissions and controls.
Object is even more flexible, but it's different. Object is about scale and dealing with billions and billions and billions of elements there in a flat address space to avoid the hierarchical structure you get in file systems so that you can access more data without the amount of time spent traversing. But what it does is it pushes the organizational structure of accessing object IDs back to the application or to another layer rather than burying that in the file system itself. The file system could be on a server, or it could be on a network-attached storage device.
Ultimately you resolve down to block access any way you look at it, but whether it's object, file or direct block access, software-defined storage, storage virtualization all have the opportunity to present any type, depending on the applications or usage. And it's more important [than that]. It's not specific to, 'Oh, I like file and block.' It's about what's best for the particular usage model.