Mike Matchett: Hi, I'm Mike Matchett with Small World, Big Data. I'm here today with Craig Nunes from Datrium. Craig, welcome to the show. Craig Nunes: Thanks for having me today. Mike Matchett: Yeah. We're going to talk a little bit today about Dat...
datrium, storage, backup, hyperconverged, scale, out
A brief intro into Datrium's converged storage and data protection architecture that can help simplify the most demanding and complex environments.
Mike Matchett: Hi, I'm Mike Matchett with Small World, Big Data. I'm here today with Craig Nunes from Datrium. Craig, welcome to the show.
Craig Nunes: Thanks for having me today.
Mike Matchett: Yeah. We're going to talk a little bit today about Datrium's solution, particularly what's new, and what angle on the market they're taking with their storage solution, and I almost hate to call it storage because it's actually a convergence, a data platform. We've got a lot of different names for it, but it's very cool. Before we dive into a little bit, though, Craig, maybe you could just tell us a little bit about where hyperconvergence went, and then where Datrium came from ...
Craig Nunes: Sure, yeah.
Mike Matchett: ... and where that came from.
Craig Nunes: Yeah. Yeah, first of all, I'd say look, for a lot of folks who are watching this who are probably using arrays throughout their infrastructure, there's been a big move off of arrays as folks are modernizing, and hyperconvergence was a great answer for scaling out performance of those flash drives and bringing flash on the host, which is where we believe it ought to be. However, what folks have found over the last five or six years of using it is, a hyperconvergence is not scaling to the degree folks need in the heart of the data center for low latency, mission critical apps, and so, a big part of what Datrium is focused on is kind of what we would call sort of tier one HCI hyperconvergence, but at the same time bringing in scale out backup, and in fact, cloud DR into a single converged system, so we were effectively taking converged infrastructure to kind of a new level, really aimed at folks modernizing for hybrid clouds.
Mike Matchett: Yeah, I know we worked hard on trying to come up with the right name for it because it really takes hyperconvergence to another step, so it's not the same thing as a hyperconverged appliance, but it's what we ended up calling, I think, open convergence ...
Craig Nunes: Open convergence, yeah.
Mike Matchett: ... a way to do that. And then I want to take the conversation a little bit towards this idea of, in the same solution for storage you've got a primary, you've got a way to deliver primary storage, highly performant by putting flash on the server ...
Craig Nunes: Right.
Mike Matchett: ... and using your storage forward on the server, but you've also married that to secondary storage effectively by having a capacity tier that's very cost effective.
Craig Nunes: Yeah, yeah.
Mike Matchett: Could you just tell me a little bit about how you guys got to the idea of converging ...
Craig Nunes: Sure.
Mike Matchett: ... primary and secondary?
Craig Nunes: Yeah, so first of all, super tough problem, right? So, you've got a system that you need to scale to a petabyte. You've got to deliver very low latency, hundreds of thousands or millions of IOPS, right? At the same time, you want to provide a supremely cost effective infrastructure for long-term backup retention, and those two things sound like polar opposites, and that was really the nut that our founders cracked, was at the heart of the Datrium architecture. We have a way to do just that, and if you think about it, we have all of our VMs, containers, we support VMware, Red Hat, CentOS, running on the host side with our data services software built in, so all of your high performance activity going on there.
Then we've got an object interface out to capacity on the network, so it's not much different from EC2 out to S3, and that split really allows us to scale performance massively, while at the same time having this very cost effective repository for primary and all your secondary copies.
Mike Matchett: That's right.
Craig Nunes: It's kind of a breakthrough for data centers.
Mike Matchett: That's right. You protect the primary storage as well as the capacity storage on that tier.
Craig Nunes: Absolutely.
Mike Matchett: Now, you mentioned a little bit about cloud tiering, and I know that's where we wanted to get to today, to talk about the Cloud DVX, if I got that right ...
Craig Nunes: Mm-hmm (affirmative).
Mike Matchett: ... and how you then take that convergence on primary, secondary and tier it to a cloud, and then run it, replicate it in a cloud. Tell me about what's going on there.
Craig Nunes: Yeah, so the simple story is we took the software that today will run on PRIM, on pretty much industry standard gear, we ported a cloud native version of that to Amazon AWS, and remember, I mentioned we had kind of the high performance side? Basically we dropped that on EC2, and that durable capacity side, we dropped that on S3, and remember I said it's an object interface. Frankly, it was a very clean port for us. We didn't touch our file system at all, and this cloud native instance of DVX, the first service we're offering is a backup and recovery as a service, kind of one click setup.
We do the rest as a service, so you don't have to know anything about EC2 and S3 and Lambda and DynamoDB and all the giblets there. The bottom line is it looks like a Datrium PAAS, an extension of Datrium into the public cloud, and the killer app for any cloud, and look, if you're not thinking about Datrium today, totally cool, when you're looking at cloud, whatever you're looking at, global cloud deduplication has got to be a part of the infrastructure because you've got to skinny up the data footprint to get the economics to line up the way you need, and that's exactly what we've done in Cloud DVX, and you're probably going to store 20 percent of the actual data size in the cloud for tremendous economics.
And by the way, that's also a key to super fast RTO. You're bringing back a very small amount of data. Whatever does not exist on your primary DVX system, we will ship back deduplicated, over the wire, so very fast RTO even from your S3 repository, so it's, I think, the first backup to cloud that, at least customer feedback has been it's economically viable, it's super simple, and hey, I can actually get my backups back when I need them.
Mike Matchett: I know. It sounds so interesting, and I think we could talk for half an hour about that, but there's one last thing. I remember we talked a little bit about end-to-end encryption ...
Craig Nunes: Right.
Mike Matchett: ... and how that now is an important part of being able to do this securely, yeah.
Craig Nunes: Absolutely, yeah. So, we have an end-to-end encryption we call blanket encryption, and we actually encrypt data in use on the host, so in memory and in flash, in flight across the network and at rest ...
Mike Matchett: Okay, so both in flight and at rest, and into the cloud and so on.
Craig Nunes: Yeah, and one of the milestones ... So, by the way, with Cloud DVX, super important, so you don't have to pay AWS for your VPN. Just turn this on. It's part of your DVX. The other thing, though, is we've also seen great demand for better data security in healthcare, and we just recently completed a FIPS 140-2 validation with NIST, so we're certified, hardware and software for your data security, and it meets all of the regs like HIPAA, Hitech, PIPEDA, everything you're going to need to do in the healthcare industry to protect patient data, and that security these days is at the foundation of a lot of these infrastructures.
People talk a lot about encryption at rest, and to be honest, a lot of folks aren't giving enough thought to protection in flight. We do that, and we do that totally deduplicated, compressed data, so it's going to, again, consume a very small footprint on your flash devices.
Mike Matchett: I think you're really throwing a gauntlet down here against some of those other guys who claim some of those capabilities, but there's no way they turn them all on at once all the time and run them up into their cloud instances and stuff, so a lot for other people to step up and try to compete with there. I think we ran out of time here tonight.
Craig Nunes: That's right.
Mike Matchett: Thank you, Craig for coming on here and giving me a little bit of a brief recap of that conversation.
Craig Nunes: My pleasure.
Mike Matchett: I look forward to talking some more, and just finally, where can people find out more information about this?
Craig Nunes: Sure. Just head for our website, www.datrium.com, all kinds of good info there, or drop me a line, firstname.lastname@example.org. Love to hear from you.
Mike Matchett: Well, that's great. Well, thanks for being on tonight.
Craig Nunes: Alright. Great.