I want to get started by talking about software-defined storage. Everyone seems to have a play here. How do you define it?
Ziya Aral: Each time this subject comes up, there is a slight redefinition of the term. But software-defined storage is a continuation of what used to be called storage virtualization. Storage virtualization [by] itself lacks server virtualization. A virtual machine is one of these concepts that basically argues for creating a software emulation layer. It's an approach to breaking hardware away from software. Now, there are a million good reasons to do that. Just thinking about the question for 5 minutes gets you there. Hardware and software live different lives with different expertise.
Hardware is redefined every 18 months. Software sometimes has a half-life of 10, 15, 20 years. It makes absolute sense to architect software and hardware differently and to have a software layer that is defined in a perfect world, in a perfect universe. [In that layer], storage is perfect. There's an absolutely infinite amount of space. It has all of the positive characteristics that you want for it. It has high availability, data has integrity, [and] it can move around at will. And then there's hardware. Hardware has the real limitations that physics impose. And you would like those two broken apart from each other.
Can you explain to us how software-defined storage is different from standard virtualization?
Aral: No. I can't.
Do you believe it's one and the same?
Aral: I think that it is essentially one and the same. Now, there are always detail differences, but it's like asking the question, 'How is the cloud different from storage service providers [SSPs] of 10 years ago?' Now, conceptually many of the cloud vendors have gone for direct repetition of what the SSP guys used to say. The technology has changed; commercially it's more practical. There are elements to the story that reflect the current thinking on many things. But in essence, the idea is the same. Software-defined storage is similar.
We're still doing emulation. And frankly, emulation is a bad word in our industry because emulation equals 'slow.' It's getting one thing to do what another thing is supposed to be doing. The other thing is several times more complicated. But frankly, emulation is the beginning of storage virtualization, of server virtualization. When the power of the underlying technology, the hardware technology, reaches a certain critical mass, then the software is able to really begin to abstract in a fundamental way from those hardware limitations. And that's the real basis of software-defined storage.
How do you stand out in such a competitive market right now? What's your pitch to customers?
Aral: Typically, their problem is they've got a bunch of controllers from this vendor, a bunch of controllers from this vendor, and they don't work together. So, when these fail, these don't take over. These aren't fast enough. So, we find ourselves selling after the fact in most cases.
Now, that's not always true. DataCore Software's getting to be a reasonably sized company. Today, in many cases, we sell at the architectural level. [However], the bulk of our business still continues to be making up for the damage done previously. In those cases, we're covering disaster recovery where none exists. For example, a customer must have immediate availability of data at all times. OK, terrific, that's great. It's a requirement of many businesses. It is absolutely universal.
They bought a box. The box has two power supplies, two power cords, two sets of logic and one piece of software driving all of it. The box fails. All right, let's think about this. [They say,] 'We should have another box. Better still, another box hooked up to another power grid somewhere. This would be great. It'd be terrific 30 miles across town.' [The vendor says,] 'We don't do that.' [They say,] 'You don't do that? OK. How do I do that then?' Answer: DataCore.
[They might say,] 'But, but, but EMC does that. They do that with two boxes of their own. But we have an old box and the new EMC box. Now do I do that?' [We say,] 'Well, you don't.'
That's what we do most of the time. It is a key piece of storage architecture, but eventually software defines storage, eliminates storage as such. It reduces storage devices to peripherals.
Now, EMC builds and sells more of those devices than anyone else around. Why would they talk about software-defined storage? In fact, why would any of the larger hardware vendors talk about it? Well, they're not stupid people. They're smart people, and they understand the science. They talk about it in anticipation of an evolving architecture that leads to many opportunities.
But, I am sad to say, 15 years into it, we're still the radical. This is crazy. This is so obviously true. Yet, we're still the radical.
Where do you see it 15 years from now?
Aral: Storage operating systems will, I believe, take the place of large, high-end storage controllers today. That is, you will build your application infrastructure around them. They will be one logically combined, logically integrated entity. But the entity itself will be a ball that moves. You put it on that platform. You put it over here. You put it in Phoenix. You move it to Tokyo. You outsource by taking that whole ball and putting it out on the cloud. You pull it back in because those damn cloud providers, they're charging me too much and they don't give me the response I want.
This whole thing will be essentially nothing other than the reflection of the business process or the application architecture guy's vision of how he wants this thing to work. It will be a figment of the mind. The fact that it lives on computers -- and where those computers live -- will be an afterthought. It'll be an afterthought for a different guy, like the networking guy that you have that's basically responsible for the wires.
Will that significantly change the role of the storage manager, storage admin?
Aral: Forgive me. Those are our customers, and I love those people. But their role is changing already. That was the most rapidly growing section of the IT industry for a while. It sure isn't today. Those guys are drowning. They're drowning. And they're going to up their game and lower their tolerance of this and that box.
We have a T-shirt at DataCore. It's an internal-use T-shirt. It says 'Hardware Sucks.' This is a new introduction of an old T-shirt. What it means is not that hardware sucks. Hardware is terrific. It does all this stuff. But when you're chained to it, man, it hurts. It hurts. You've got to free yourself of that.
No comments:
Post a Comment