Exclusive Interview With the CEO of DataCore

                          Key Facts and Figures of DataCore

 George S. Teixeira, 53, president and CEO of DataCore Software, co-founded the company after ten years of senior management experience at Encore Computer Corporation, where he was a team leader of OEM marketing and sales for Amdahl, IBM and DEC. His work culminated in the $185 million sale of Encore’s storage control business to Sun Microsystems. He also held a number of senior management positions at Gould Electronics’ Computer Systems Division. Teixeira is a native of Madeira, a Portuguese island off the coast of Africa. He earned an MBA at Florida Atlantic University, after joint B.S. degrees in Chemistry and Biology.

StorageNewsletter: When do you think DataCore, founded twelve years ago, will be considered a major independent storage software company, along the lines of CommVault, not to mention CA or Symantec?
George Teixeira: At that time, that was the expectation, the problem was that the market took a lot longer to recognize that virtualization for storage, to be honest, I don’t think until about 5 years ago when VMware and all of the virtual server market, when we started the company, there was no storage virtualization, just the mainframe. In our space there was none. When we started in 2000 with the product, the first customers were OEMs, in fact Fujitsu and Hitachi invested in DataCore because they wanted to know how to do thin provisioning. So that was how they helped us get into the world, and even IBM resold our product in order to first learn about storage virtualization themselves. The problem is we were fighting a very hard battle. Most people would say that we and FalconStor were the pioneers. To be honest, when we started our investment round, we met with the bank that wound up investing in FalconStor, so I don’t know if they took our ideas, or if it was just convenience, but that two us came at the same time almost. The group that met with us then met with Cheyenne and CA, and I guess they talked to the Cheyenne people, who became FalconStor. I’m not sure of the whole history. The big difference was that FalconStor were tape people that came from Cheyenne, so they more logically focused on the virtual tape libraries and the agents, while we were focused on a different version, which used caching, so for us to do virtualization, we saw caching as more important, because we needed to overcome the performance issues.

So the question still is, when will you be considered big enough to be known among the big storage software companies?
I don’t know, because that’s a market size. What I do know is that we’ve been growing 30% per year in revenues and bookings. More importantly, last year, to give you an idea, in the last quarter, I think we make 1,000 deals. I’m calling them deals because it involves existing customers doing upgrades as well as new, but to give you an idea, last October, we did 225 deals. Now that’s substantial, because for all the talk about how many people are doing storage virtualization, you start doing 1,000 per quarter, and you’re seeing a difference. Now we saw a big uplift in the second half of 2009, and I think it’s because of three trends, probably. One, the Microsoft people are starting to get interested in the SAN world. And the truth is that most of them don’t have the expertise, but they know that they’ve got the capability. So in fact the first announcement where we talk about a Virtual SAN Appliance, part of it is that we recognize that unlike the VMware kind of customers who already have expertise, these are network administrators, they’re more used to Exchange or SharePoint, so we wanted to create a SAN appliance where everything is pre-configured, all they do is download, click and it’s there.

Is this the first time you’re entering hardware with your Virtual SAN Appliance?
No, we’re not doing hardware, it’s a Virtual SAN Appliance. It runs in a VM, so for Microsoft, here was the problem, they have all these new customers that are with the Windows Server R2 and Hyper-V, for them to try the new advanced features, the motion, as an example, they have to buy an EMC, they need the big shared storage. So they’re finding it hard to teach their customers about shared storage. So we came up with a Virtual SAN Appliance – they download, they put it on, they can try it. In reality, most customers will deploy something different, but to learn, and to at least see it, they get exposed to DataCore. So one trend is I anticipate Microsoft will have a massive impact in the market, and the reason is that they’re bringing in all kinds of new customers who weren’t thinking about virtualization, but because it’s basically free, with Hyper-V, they’re getting it. A second issue is that because the economy was so bad, people are what I’m calling, “Return on Yesterday,” like Return on Investment, because what they’re finding is that they don’t have the money to buy a new storage array, they don’t have the money to buy new servers, so we say if you’re consolidating, you have extra servers, if you already have storage, it’s not being used well. Do a return on yesterday with DataCore, which is use what you already have.

You’re not the only one…
No, of course, but with DataCore it’s pretty easy to do, I guess that’s the key. The third big trend is one that appears a year ago. If you bought, if you were going to do a virtualization project, you were going to spend €100,000 to do virtualization, and then you would find you were going to buy an EMC and spend another 100K. What’s happening? That 100K of VMware today, you can get a XenServer, you can get a Hyper-V, the price went down to 20K. The problem is this didn’t go down, so we have this mismatch in the market. So we’re finding that a lot of customers now are saying "I want to do virtualization because I not only want to consolidate, but I’m stuck, this price is too much." So we give them a way to use their existing materials, and thus get the price down. The other thing I saw was that the first wave to virtualization, I would call consolidation. Very simple.

Aggregation.
Aggregation. The second wave, I’m not sure what to call it, but it’s more around productivity and automation, but it’s also business continuity and VR and all that. What happened here is that a lot of the customers started this way, but now they’re saying "I can do the motion, I can move it to another place, I can deploy many more units," so here the requirement for a SAN has even increased more. So now we have all of these forces. What I’m seeing is that Microsoft is causing downward price pressure, with the cost of existing storage is too high, and I see the advanced features of virtualization appealing to more and more of the customers, who want to do that.

Although it was founded in 1998, do you still consider DataCore a start-up?
Not really, not any more.

You had a financial round in 2008…
$35 million.

To a certain extent, I’d say that a start-up is a company looking for financial funding…
But here’s what happened. The original funding, most of those investors, after we had a hard time in 2005. So that money dried up, well it was invested in the product. The employees own the company. That $35 million still leaves the employees and management at 51% ownership. So we’re in a good position. I wanted some extra money because of the recession.

Last time I saw you, in December 2007, you told me you were profitable. Why then did you need this last financial round?
Because we’re a small company still, and when we deal with, for instance IKEA, one of the first questions they ask is "Do you have money." So we have an answer. And I haven’t touched the money, I’m still making cash from my business, and I can use that other money in case I want to acquire somebody, or do some other project that requires cash.

How do you see the long-term future of the company?
There are multiple possibilities. We clearly think that for the next few years, we focus on what we’re doing, we’re on the growth rate, seeing 30% growth. Next year, you’ll see a complete next generation of software that will simplify that people do storage. Basically what we’re learning is that because the concept of a storage administrator doesn’t exist anymore…

The iPhone OS of storage software – it’s pretty complicated, no?
It’s too complicated. For instance, on our systems today you have to know is it iSCSI, is it FC, is it FCoE, who cares? So what we’re doing, to be very simplistic, is just like Microsoft Works, you have the Wizards… It will say: "What’s the task you want to do?" You say: "Create a Network." It’s going to detect that you have FC or iSCSI, it’s going to take care of the connections, set it up…

It’s a little more complicated than that because of the different connections…
You can always make changes, but the difference is the CPU has gotten faster, the channels have gotten faster, so that even if you’re wrong, you’re not that wrong anymore. Two years ago, if you were just doing this, it wouldn’t have worked, because the differences were too great. Today, and by this year, it’s not a big difference.

Do you believe in the future of a Converged Network Ethernet, FCoE?
Yes I do.

Replacing FC?
Eventually, but much longer than all of these people think. If you remember, it took a long time for iSCSI, it seemed like 10 years.

But it wasn’t pushed as much by larger companies…
The problem I think is that the iSCSI is moving faster. As it gets faster, I’m not sure how many people need to, unless you have the big cabling, then obviously FCoE makes a big difference. And we announced it. And to be honest, I have almost no customers using it. But from an architecture, at the moment we can support it, we need to support it, but we’re agnostic, I don’t really care. iSCSI, FCoE, FC, it’s all the same.

When will you support VMware View 4?
Which is View 4 – you may have caught me there… We keep up to date with VMware, so if it’s part of the normal package, unless it’s something special? Maybe it’s something we do and I just don’t know it.

What’s new in DataCore’s Advanced Site Recovery Software?
That’s one of the lastest announcements, so two things. We announced it with SANsymphony, what we announced in January is that it will also work with SANmelody, so you’ll be able to go to some of the smaller sites as well. The big difference with the site recovery is what we announced, Microsoft decided to put it in the Tech-Ed in Berlin. And it’s part of their Pavilion. The reason for that is obvious: VMware has site recovery. So when you look at the two platforms, you say, "Do they have VMotion," –Yes we have Motion, "Do they have site recovery?" Microsoft didn’t have a solution, so they worked with us and we put it on the site. So the big difference, VMware assumes I have a mirrored site, and everything is mirrored, and by the way, it’s only VMware to VMware, so it’s recovery. Microsoft, when we talked to them said: "That’s not really the real world. The real world for us if that you have a bigger center, and you have many different branches," so they want what we call the hub and spoke. So what we did, the way our Advanced Site Recovery works is that these elements do not have to be the same, this can be a SANsymphony at the data center, this could be a SANmelody at your hotel, this could be a very small system on another spoke. The second thing is that we allow you to distribute the workload to the different sites, so at one point, this could be Exchange, another could be SQL, another could be VMware, whatever, you can take it and put it by group. So literally, you press a button, and it takes care of suspending, moving all the stuff and shipping it to the remote sites. And it not only works with VMware, it works with Hyper-V, with XenServer and with physical storage. So if you have a Mac with its disk in the pool, it can move that disk to this site as well. So it is generic, would be the best way to put it, and it’s interoperable. So the disks here could be in HP, with HP servers, this other spoke could be Dell, this could be IBM, it doesn’t matter. So the announcement we made first was that Microsoft is using us to help fill the whole of site recovery management. The second just announced is that we do it with SANmelody and, we’re supporting a petabyte block now. How many people are using a petabyte disk size? It’s coming, though. The reason for this is that up to know we were like most people, 2TB limits, there are some that can do 16, but it’s still tricks to do it. When Microsoft announced R2, they support 256TB, and there are now a version of Solaris and a version of Linux that support a terabyte. So for us, we decided that if we’re going to do more than two, may as well go to a full petabyte.

Do you see virtualization ultimately being included in the operating system?
Eventually over time, there will be more of it in there, because, well the first example is Microsoft. You can argue that Microsoft has already taken server virtualization and included it…

And storage virtualization?
It’s still hard to do, but in the long, long term, I think yes. And so will desktop virtualization, but those will take a much longer time.

Two years ago, you launched virtualization software for iSCSI SANs under $1,000. What was the market’s response?
It’s good. The problem was we’ve found that it was not the right place for us to focus. The problem is that people don’t understand that at $1,000 for an iSCSI, you don’t see the benefit of what we do, which is really fault tolerance. So it confused people, who then asked "How is this different from some other iSCSI appliance?" And what we found, more importantly, with DataCore, what we differently from everybody else, I don’t care if its NetApp or EMC or whatever. What they’re telling you is that we have this one box and we have two controllers, and this is really high availability. And they do everything they can to make this robust. With DataCore, you’re taking two servers, first of all, you put them in two different locations, even if it’s just the next room. And the moment you go to two in different locations, it’s about ten times more reliable. And the reason for that is in this one, if water drips on the one unit, you’ve got a problem. The second part, usually the most common thing I find, someone is in changing a wire, their elbow hits something, so no matter how high availability it is, they’ve got just a single point of failure. By separating the two, when we say DataCore fault tolerant storage, it is two separate points, up to 40km remote, iSCSI or FC or FCoE, connection, and the beauty is, if I do VMware, if I want to go from this box to another box over here, this disk over here has a volume ID. So that one is 100, the other is 200. When you want to put it somewhere else, you are not doing a synchronous, you are doing a replication, and it’s disruptive. With DataCore, you have a complete lookalike ID, so when I move the VMware with VMotion, it’s seeing the exact same disk in either location. That little difference is what we focus on. To answer your question, then, yes, we sell a lot of the $1,000 solution, but it’s the wrong place to show our true differentiation. What we want to sell, and what our customers are looking for, is a high availability over multiple sites.

On which hardware storage platforms do you sell the most?
It’s interesting, the most hardware is probably HP. I think HP is about 50% of our systems. Now it’s interesting because I have to answer you in two ways. A lot of times, that HP comes in two approaches. One is somebody has an HP server, and then behind it, they have HP MSA. So that’s one example. The other is different. Somebody just buys an HP, so IKEA for instance, when they first started, they buy an HP server, they fill it up with all the disks, and then that becomes a storage server. But in IKEA’s case, they started with HP, and then they switched to Dell. But because they had our software, it’s not a big deal. So HP’s the dominant one, hard to say who’s next, Dell or IBM. First are the ones sold as servers, HP first, Dell and IBM second, then a bunch of no-name distributors. In terms of storage, HP is first, you would be surprised, I bet you Hitachi is very high up there, and Hitachi is for a different reason, because it’s very expensive to do the replication. Unlike FalconStor, or anybody else, our caching can absorb the very high I/O. So we can sit on top of these very big arrays. So I don’t know if it’s EMC or Hitachi that’s second on that.

On which virtualization platforms (VMware, Microsoft, Citrix)?
VMware clearly. We work with all of them, but VMware is probably 65% of the deals. Microsoft and Citrix you’ll find in a number of them, and don’t forget we still have people, 10% of the business, who don’t do virtualization at the server level.

How successful was SANharmony, which shipped in 2Q08, and integrated SAN and NAS virtualization?
Not very. What it was for us was more the marketing concept. And I think the world changes to be honest. What we were saying was that you could put the NAS in front of DataCore. If you remember when Microsoft had their Windows Storage Server? In a funny way, we were taking advantage. Here’s the difference now. Windows 2008 has everything Storage Server used to have in it. So remember, we run on top of Windows 2008, so the truth now is that I don’t really need the SANharmony, because if you want to do NAS, use Windows 2008, do the NAS. While with SANharmony, if you have the disks formatted under the Windows file system, I can still see them, I can move them. Harmony was a way to bridge until the Windows world had enough power to do the NAS.

What’s the percentage of your sales associated with virtual servers?
It’s almost 90% now. It’s very rare that we don’t sell with either VMware, Citrix XenServer or Hyper-V.

Do you continue to sell more in Europe? Why?
It’s still more. And for the same reasons The food is better. Seriously, it’s closer to 70% in Europe, 10% in Asia. Germany is almost 50% of our business alone. It’s grown. And when I say Germany, I should be careful, we call it the central region, so it includes Denmark, Austria, Switzerland and even Italy.

Percentage of sales to OEMs?
Very little now. In fact, almost nothing.

You have no OEMs left?
I think we just completed one, the last was NEC. And we’ve finished with them at this point. But they still resell our product, but we weren’t doing the business.

Will you enter tape virtualization as FalconStor did?
No expectation of that. I think what will happen in our future, we could do it, but I think there’s enough choices coming on the market that we will simply integrate with others that are out there.

Will you offer a product especially for cloud storage?
Can you tell me what that is?

Yes, but few people have the same definition. Basically we’ll use all the disk drives inside the companies or even in the world to hold all the raw data on clouds. Because only 25% or 30% of drives are being used.
I don’t have a problem with that kind of definition, the problem I have is the marketing definition. The recent announcement by EMC, VMware and Cisco, to me makes no sense, because there are saying you have this one cloud, but IBM is going to have a cloud, HP is going to have a cloud. But these are not clouds. The real problem with clouds is not the storage, the servers, the applications, the problem is connectivity. If you can’t move to run an operation, your oracle database effectively today? It’s hard. To be honest, I’m against marketing in terms of the cloud, seriously, we’re more of a cloud than anyone else, because we’re software. How can a hardware guy say he’s a cloud. Our partners today call it private cloud. To me that’s a data center. They take VMware, they take DataCore, they put the backup, and what they do is, for their department, they have machines that you can provision, the storage gets provisioned, and then if you want to archive, they back it up. That’s a private cloud, and it works by the way, and we have a number of those installed. I could call it a cloud, but to me it’s no difference for all the things we’ve been doing. And from a marketing side, I know our guys will make some slides with the word cloud, but I don’t know how it’s different.

You already have CDP.
Different from most companies, we’ve sold it in really high-end jobs. For instance we have it in Maimonides Hospital in New York and in that application, they have 150TB going to 200TB, it is four or five different databases, it is tens of applications. So their view, they’re using CDP as a way to buffer the contents before they even archive or backup. And what they’ve found is interesting. If you have an error, or a virus attack, something like 90% of that occurs within 48 hours. So by keeping all of the storage for all of this, as much as 30 days back, it relieves them of the problem of doing the backup immediately, they can leave it in the disk, and then do backup as they need it. The ones we’ve done are literally 50TB or bigger, really big. Where we’re going in the future is to take that same technology and bring it to the low end. And instead of 30 days or 100 days, maybe just a 48-hour rewind. Because for a lot of companies, it’s the 90% problem that I want to solve. The problem is that it goes to far back, you have too much disk storage. But if you go 24 hours or 48, you catch most of the problem. Because you usually know that "Joe deleted the file." And you know that in a day or two.

When you’re a storage software company without de-dupe, that’s a big problem, no?
Well who has storage de-duplication? I only know of one, and that’s Data Domain. Unless you’re doing backup. You tell me who else.

Quantum?
Sort of. But I’m saying, everybody says they’ve got it, but are they doing it at the file level, or is it production data. If you’re doing backup, we have people using CommVault with DataCore all over, they’re doing de-dupe with that, what do I care? We are working on compression technology for moving stuff. We are looking at de-dupe, I won’t give you all the details, but we are. And not the way lots of people are doing it now. Most of the de-dupe nowadays is done at the file level, not at the block level. The other thing about de-dupe that’s funny, I was with a customer, and I won’t mention the name, but the customer explained to me how he had this major advantage with de-dupe. He went from 8TB to 4TB. And I said: "That’s great. How much did you pay for the system." $90,000. Let me ask you how much does it cost you to buy 4TB that aren’t for production data. Less than $1,000. So what problem are you trying to solve? If your objective is to reduce costs? If that petabyte isn’t production. How many people do you know that are using de-dupe on production? If it’s for backup, there are a hundred products out there, and do you want to pay $90,000? Why solve a problem that’s already solved? I mean you’re right, I could go out and do de-dupe, and I will, but I don’t think that’s what most people are looking for. What they think when they say de-dupe is that they’re getting benefit on their production data.

There’s de-dupe for primary and secondary storage?
But if it’s secondary storage, if you’re doing for instance what we do, which is thin-provisioning…

You can do both…
Yes, you can, but how much do you get for doing the extra work for de-dupe? It’s not that much of an advantage. It’s the one exception out there. I’m not arguing it’s not valuable, but I’m just saying – I’m not convinced by what everybody is doing.

Are you looking at de-dupe for primary or secondary storage, then?
I’m looking at primary.

Which is much more difficult…
Absolutely.

Some personal questions – tell me about your hobbies…
I like to bike ride. I play racquetball, and I work too much.

I seem to remember from last time there was something musical?
My son is an opera singer, both of them are opera singers.

How many vacation days did you manage this past year?
It was better this year. Last time I told you 10 days. This year was much better, I went to Mexico, and had a wonderful time, and I’m planning to go the Bahamas, and I was in Barcelona. So I’m looking at 20 days.

And do you ever go to Portugal?
I do, but they’re quick trips. I get a weekend here and there.