 This is Dave Vellante, and we're winding up here, but our guests are still streaming in, and we hate turning people away. It's such great content and really good action going on here at VMworld. We're down in the hang space, so by all means stop behind and see us. We're running out of time, but we'll try to squeeze you in. And I'm here with Jeff Eccles, who's with Commvault. And Jeff is Senior Director of Product Marketing. A lot of going on, Jeff, in the backup space. But first of all, welcome to theCUBE. Thank you very much, appreciate it. Yeah, we've been talking all week about how virtualization's been stressing these various systems, not only storage, generally, but backup specifically. And we had a segment on backup. We had Jason Buffington on. He was talking about how virtualization's been very disruptive to the whole backup business. So what are the big trends you see in data protection? Specifically as it relates to virtualized environments. No problem. No problem, yeah, virtualization definitely added, it did a lot of good things, obviously, on the server side in terms of cost savings. It did add a layer of complexity there, to your point, on the data management side. And so it's actually, from our perspective, it's been a good thing for us, because anything that's kind of a big change like that forces change across the organization. And that's when we're able to kind of come in and tell our story about kind of holistic management, single pane of glass, all that good stuff. And so virtualization did that. But you're exactly right. What we're seeing now is that people are starting to move applications rapidly into VMware. And that creates a lot more complexity around being able to tie application data down to that storage layer and then protect it. And so the big trends we're seeing, again, we're kind of focused on mid to larger organizations is leveraging hardware snapshots on storage arrays to actually protect those data storage that are running VMs. And that's a hot topic for us right now. Yeah, so Commvault, obviously a major player in the backup software business. And a lot of shifts going on there. You're seeing, obviously data de-duplication was a big disruptor and in a good way. Really drove the disk-based backup opportunity. You're also seeing snapshots come in. You guys have a very strong relationship, I know, with NetApp and are driving that vision. So how do you see the progression of the way in which we are doing backups evolving? How is that changing? Well, what's changing is we believe snapshots are going to be a fundamental piece of the equation because people, especially as server IO is more and more stressed. There's not enough time to stream a backup off of a physical server. And so people are going to have to start using hardware snapshots on arrays like EMC arrays or NetApp arrays, or we support 17 different arrays today. But being able to use that snapshot to rapidly co-asset data stores and then kind of come in and mount that snapshot off from the side and actually catalog that snapshot, be able to pull particular snapshots off for backup, but be able to do it in such a way that you're not stressing that primary ESX server. And I think the traditional methods that we kind of have used as an industry in the past are no longer going to work. You've got to start with a snapshot to get that quick capture of data, leveraging that silicon on that array, and then start pulling it off to the side. And in some cases, moving it off the array is going to leverage that hardware replication built in, such as a NetApp replication snap-all snap mirror, or it's going to be kind of deduplicating it off hosted and sending that off to secondary storage. It's going to kind of vary there, but you're definitely going to see more snap and more replication in the future. The VMware contribution to backup has evolved over the years. You remember VCB, the sort of failed experiment. But okay, it's a 1.0, and then obviously VADP has had a big impact. Basically, the VMware API for data protection. VMware put it out there and said, okay, ecosystem, you go solve this problem, and that was good. You guys know what you're doing, right? And everybody hopped on that, and that's made some good sense, particularly with change block tracking, be able to take advantage of that. And then of course you had VDR, the sort of low-end entry-level freebie with vSphere 5, and now that's replaced by VDP. Chuck Hollis came up with a blog talking about that's based on Avamar, everybody kind of knows that everybody knew it was coming. But essentially said, you guys, quaking in your boots. What's your point of view, so on the record, what's your point of view on VDP? Are you quaking in your boots? Is this, you know, you guys changing your business strategy as a result of this? Is it business as usual? No changes from our perspective. I think it's an interesting play. I think that that play is going to open up some SMB shops that are moving applications into VMware. We see that as kind of a fundamental step in having some protection built in there for smaller shops or shops that are moving applications into VM and need to have a backup story as part of it. I think it makes a lot of strategic sense for those guys to kind of go that direction. As far as the environments that we sell into, that packaging is not going to be something that those customers are going to be interested in. Having a two terabyte limit and running it inside of ESX, enterprise shops are not going to want to have backup data kind of living in ESX. And they're generally going to, they're going to have more than two terabytes, right? So a lot of those guys are still going to be in a different class of solution. They're going to be looking for something, you know, in the dozens to hundreds of terabytes, maybe even bigger than that in a lot of cases. So they're going to want to come in off of ESX. They're going to want to live it on lower cost storage media. And they're going to need a lot more than a couple of terabytes. So you're basically saying you're trying to solve a bigger problem? We're solving a bigger problem. That's exactly right. Let me, let me counter though. Are you concerned that, you know, the low end gets bigger and goes up market? Do you feel like the market that VDP targets that freemium model can only go so high? I think it only goes so high. I think that, I think that there's a classic customer out there that's okay with backups living inside of ESX. But at some point in time, you want backups coming off of a server, right? You don't want them all living in that server. And you don't want them all living on primary storage, which is, you know, can be costly, right? So at some point, when you get into, you know, dozens, hundreds, thousands of terabytes, you got to be pulling it off of that server and going off host with it. And you need to be in the hundreds or thousands of terabytes kind of scale. That's our position on it. So, you know, that'll... You don't have a freemium model. We don't have a freemium model. Yeah, not launching one anytime soon, right? No plans. You know, we had Veeam on earlier. And obviously this goes more head to head with Avain. Absolutely. At the same time, they have a freemium model. They weren't quaking in their boots either. I don't expect they would. I mean, they're a good competitor. The market's robust. Right. So, we love the competition and the disruption. I want to come back to the snapshots because we... Okay. In 2010, we put forth this notion of the time machine for the enterprise, we called it. Using, you know, the idea of Apple's time machine as a way to back up using snapshots and continuous data protection. Dial up and dial down your RPO. And it's starting to take shape. We're really pleased to see that happening and it's really attacked the sort of the backup window problem. Right. Talk about how customers are actually using that, deploying that and what the impact is on their business. Yeah, yeah. So, you know, specific to Veeam, you mentioned VADP a second ago and you're right. We all had to work off of VADP to kind of make backups work. Getting back to kind of those bigger environments, what we started to see in the last, I'd say six to 12 months, is aggressive virtualization adoption starts occurring in bigger shops, applications start moving in. Using VADP without those hardware snapshots, it can create some problems in those larger shops because what happens is you go to kind of queues that data store and you got to wait for all those VMs to stop before you can do that software snapshot with VADP. You could be talking about 20, 30, 40 minutes for all those VMs to stop running so you can do your queues and you take your VATB snapshot and then you release it, right? Then you got to let that log playback, right? There's been logs kind of building up while you've been waiting for that data store to stop. That's where you start to see some issues. At times you could see some orphan snapshots, you could see connection get lost during the log playback and stuff would kind of fail there. You kind of, now you enter the hardware snapshot game. When you go in to do something like that with Commvault with VADP, we're going to trigger that hardware snapshot. So now you're talking about queuesing and snapping that store in a matter of seconds, maybe minutes. So your transaction buildup is much, much smaller when you're talking about using a hardware snapshot model and so for an enterprise customer that can be pretty significant if they're dealing with things like orphan snapshots or they're trying to get into the hundreds or thousands of VMs, you really want to kind of be there with a hardware snapshot. And the reason you want to queues in the first place by the way is to get that application consistency in that VM. Yeah, data consistency issue. If you're moving apps in, that's going to be a critical piece of the story. Okay, so the idea is you take that snap, it's on site now. As you know, most of the recoveries are for relatively fresh data. So what do I do with that data? I leave it there for some period of time. I know best practices, day, week, 10 days, whatever it is, whatever your policy is. And then I got to get it off site, right? Because I have an RPO, whatever my RPO definition is, I got to get it off site. What are you recommending there? And where does dedupe fit in? Yeah, this is a great question. So we're seeing, you know, the bigger shops are going to create multiple snapshots during the day. I think you're going to see, you know, an average of maybe four to five snapshots a day, depending on the application, that they can revert back to. They need, they have some kind of corruption in exchange or whatever it may be. You want to fall back to a snapshot with Commvault. That's kind of a right click, fail back to that revert, you're up and running, we'll play the logs back, you're good to go. Generally, those folks will take one of those snapshots and they'll pull it off for a backup. They'll mount that snapshot from a proxy, so they're not going to hit that ESX server again. They'll mount it from a proxy, they'll run that backup job to another storage target in the environment. Now what we're starting to see from a disaster recovery standpoint that's interesting with cloud is, we're seeing people that are starting to replicate that data out into infrastructures of service. An example could be Microsoft Azure. So I replicate that data that lands in Microsoft Azure, it's de-duplicated, it's encrypted, it's sitting out there in cloud now. In the event I have a disaster at my site, I now have the ability to come in, fire up a VM over there in the cloud, mount that last backup copy, recover that thing and I'm off to the races, right? I reworked my network and now I've got a very affordable disaster recovery leveraging cloud infrastructure. So obviously not de-duping the on-site snap. Well once we back up from that snap, we would de-dupe at that point and we would keep it de-duped all the way out to the cloud. So that first snap is unde-duped, correct? That's correct, it's native files. So I don't have to re-hydrate it and then whenever I back up that snap, that's when I de-dupe. That's right. Going to either a target or the cloud. That's right, we've probably seen people do both. They'll want to de-dupe to a local target in case they need to get it back up back fast and then they'll create another de-dupe copy lands out there in cloud. And then that data out there, I can fire that up into a VM in the event I have a disaster. Okay, so I got that, and I may even take it off and off disk and get it off the tape. Right, retention, yeah, you may have compliance reasons where you got to keep certain data for. What's your tape integration strategy? We've got, that's where we come from. That's our history, right? So you name it, we got it on the tape side, right? Because people are catching up to you there. It's really good to do it. And then did you see, I'm sure you did, Amazon Glacier? I did see that. So if you were CIO, would you put your data in Amazon Glacier? Absolutely, if it's the right kind of data. I mean, it depends on what the data is. I think one cent a gigabyte is very compelling. We see a lot of CIOs right now that are looking for affordable infrastructure that's not tape, and cloud storage in many cases is it. So one cent is pretty compelling. And so I think that's going to, that's going to be a game changer. Well, it's cheap to get it in, but to get it out, we're not sure yet. Exactly. So be careful of that. That's right, that's right. All right, Jeff, we're out of time. Listen, thanks very much for stopping by. I know it was on short notice. So really great to see you again. Thank you. Appreciate you sharing your perspectives with our audience. No problem. All right, this is Dave Vellante, and we're going to wrap up shortly from here from VMworld 2012. My colleague, John Furrier, we'll be back to wrap up. So keep it right there. We'll talk to you in a minute.