VMware's (VMW) 42nd Annual J.P. Morgan Global Technology, Media and Telecom Conference (Transcript)

May.20.14 | About: VMware, Inc. (VMW)

VMware, Inc. (NYSE:VMW)

42nd Annual J.P. Morgan Global Technology, Media and Telecom Conference Call

May 20, 2014, 10:00 AM ET

Executives

Martin Casado - Chief Technology Officer, Networking and Security

JR Rivers - Co-Founder and Chief Executive Officer, Cumulus Networks

Analysts

Rod Hall - JPMorgan

Rod Hall - JPMorgan

Welcome everybody. Really go ahead and get started here in just a sec. So my name is Rod Hall, I am J.P. Morgan's telecom and networking equipment analyst. And I've got the great pleasure of having Martin Casado from VMware. He is the CTO of Networking, but Founder of Nicira Networks. And I've also got JR Rivers here, the CEO and Founder of Cumulus Networks. So welcome guys. Thanks for coming.

JR Rivers

Thank you.

Martin Casado

Thank you.

Rod Hall - JPMorgan

Two esteemed panel members here to talk about SDN. And no panel or presentation would be complete without a forward-looking statement readout, which I had some, will be very unhappy, if I don't read. So let me do that.

So statements made in these discussions, which are not statements of historical fact are forward-looking statement based upon current expectations. Actual results could differ materially from those projected due to a number of factors, including those referenced in VMware's most recent SEC Filings on Forms 10-Q, 10-K and 8-K.

JR Rivers

Does everybody have to pause to read that prior to listening to the discussion?

Rod Hall - JPMorgan

If anybody wants to see this, please come up, I'll keep it right here for you. So I wanted to maybe kick off just for a couple of minutes, get you guys to weigh a little bit of framework for us, describe how your companies fit into SDN? And then once we've done that, I'll start into some questions. So maybe Martin you want to start us, and then JR you could follow.

Martin Casado

So let me just paint a little bit of a mental framework. So something people don't realize about SDN is it's not some existential threat that's going to happen some time in the future, it's like it's already happened. And we already have a glimpse into the future by looking at how mega data centers and new data centers build outs happen.

So over the last, say, decade, the most technically advance, most sophisticated customers on the planet picked an architectural model for the data center, which were software defined. And I'm talking about, Google, Amazon, Facebook, Zynga, Tencent, Baidu, et cetera. These are by far the most scalable data centers. They are the most cost efficient, both CapEx and OpEx, and they are the most successful by most metrics.

And if you look in the makeup of these, the majority of functionality is implemented in software and the networking hardware provides capacity or bandwidth. So nobody sold them this architecture. This was in-part of the sales cycle. This is like how you build a good architecture, if you're a technically advanced organization.

So the dirty little secret is, however, is in order to build the data center like that you have to own the application. You have to have to be able to build your entire application. Google, they own their application; Amazon owns the application; eBay owns the application, right. If you don't own the application, you don't get those benefits.

So if I go to financial enterprise X and, Mike, listen, you run 7,000 apps, many of those you didn't write yourself, so while it would be great for you to get these benefits, you can't unless you rewrite those. So what VMware does for in the SDN space, in the network virtualization space, is we provide the CapEx/OpEx operational model, future velocity of the mega data centers to everybody else.

Rod Hall - JPMorgan

Great, thank you. JR?

JR Rivers

I'm going to kind of tag on to what Martin mentioned from a slightly different angle, which is, in very early days of networking, pretty much all of the kind of higher level functionality was relegated to software. That'd be bridging and routing and firewalling, all those sorts of things.

On the hardware front, you mainly were looking at that from transmission systems perspective, like how do I send data over a Tier-1 link or even an ISDN link. Over time what happened is, especially in the early days, we would recognize that there were functions that were well understood, broadly applicable, and we start putting those pieces in hardware, kind of slowly generationally and then keeping that most of the functionality in software.

There is this interesting period of time over the last, say, 15 years or so, where there was a transition and companies started kind of making big bets on hardware going forward. They'd add some brand new functionality to an ASIC before it ever been tried in software, and a customer would buy it.

It sounded good on paper. It might solve the business problem. I'm going to go through and try this brand new technology and they try it. And some times it would actually workout pretty well and other times it wouldn't. But then they start to become locked into specific systems vendors that provide this kind of advanced forward-looking technology with no real proof points.

What we're seeing is, as driven by the mega-scale data centers is, these people that have both capital as well as operational efficiency requirements and speed of innovation have recognized; you know, what maybe we should go back to the older model. We'll take pieces that we know that are tried and true, and we'll buy hardware that supports those tried and true things. Things like bridging and routing, and most recently, VXLAN tunnel terminations, which is an SDN construct, and we'll put those things into hardware, because they're well understood.

But you know we're going to input everything else and keep it in software on endpoints. Some of the endpoints might be where we host applications or hypervisors, like where ESX kind of steps in. Others might be a standalone server that access a firewall or a load balance.

The key note today from FireEye, they were talking a lot of about these very flexible ways of deploying intrusion detection systems throughout a network. And that's all based on dropping these things down on bare-metal servers. There is no purpose build hardware in their data path, because it would limit their agility and their ability to effect the security concerns of the enterprise.

So that's really this big push in SDN is going back to the old days. We're going to go back through and we have industry-standard hardware around networking from companies like Broadcom, Intel, and Mellanox. We have industries standard silicon around compute from companies like Intel and AMD. And customers are figuring out -- customers and commercial enterprises like ourselves, figuring how to package those up and make them consumable by both enterprises and the mega-scales at the appropriate price points and the appropriate deployment speeds.

Cumulus networks, my company is very focused on the physical networking infrastructure, how to help customers build super-high capacity, very cost effective data center networking, so that they can deploy technologies like VMware ESX to their business advantages.

Rod Hall - JPMorgan

Thanks, JR and Martin. So let me kick-off with one other thing. I think what we're going to try to do on this panel is maybe address with fact, some of the FUD that's going around the market right now. And I want to start off with the scaling question.

So there is a lot of chatter around about NSX scaling and the idea that it doesn't scale well, and that certain people have had problems with scaling. So I wonder, Martin, if you could just talk a little bit about maybe what's happened with regards to scaling, how you guys have progress, and where we are today?

Martin Casado

So I'm going to make two points. The first point is, if there is any scaling issues that have happened have been kind of early product in market, largest customers on the face of the planet trying to push it well past kind of what we designed it for, and there has been no architectural issues. And then I actually want to talk about how the fact that we've been able to scale it over a 100 times from times the initial value shows they are the part of software.

So on the first one, I think we're the only production network virtualization solution in the market. We're certainly the only one that's been around for three years. When we launched, we launched with five of the largest customers in the world, eBay, Rackspace, NTT, AT&T and Fidelity. And this is an immature product, develops by a startup. And at the time, I clearly, it was being pushed well beyond what we've been able to test or validate or so forth, as happens with many immature products.

This is primarily a legacy issue. The scaling issues are implementation like we didn't tested under this configuration and so forth. But there has never been an architectural scaling problem. And in fact, one thing I'd like to highlight, which goes to my second point is, since the international release, we've actually increased the scale 100 times, which I mean is enormous for scaling of any networking equipment. And if you compare that to a hardware-based approach where basically you're taking and you're chiseling scalability limits into hardware, for example table sizes.

Normally, in hardware like looks like back in the days of VLANs, right, so there are existing hardware products today that have VLAN scalability limits, because of the hardware table space, that the only way to actually improve the scale is you have to buy an entire new product line, and this is actually driving a lot of hardware sales today. Where with software, for us, we work on the implementation, we fix the implementation and then we improve scale.

As of today, with many, many deployments some of the largest deployments in the world, we don't hear issues about scalability anymore. And I think that we've got a lot of breathing room to go forward to go ahead and continuing to innovate here, and we'll be able to do it much, much quicker than if you put this stuff in hardware, so sure, early products have scalability issues. There is nothing architectural here and we're in the best position to address scalability challenges in the future.

Rod Hall - JPMorgan

And I think what's interesting you're saying that hardware-based systems may actually find it more difficult to scale over time?

Martin Casado

Well, that's the thing. I mean we've got very specific examples of this, right. I mean there are switches today that have VLAN scalability limits and the only thing a customer can do is buy a new piece of hardware. And when they were having a scalability limits early on, they had to wait for the new ASIC to be spun, which is minimum, what four years?

JR Rivers

Two years.

Martin Casado

Two years, so in order to get like the new product to get it out, right. So yes, I do think that if the scalability limit is in hardware, you're really in trouble because now you're tied to an ASIC design cycle in order to have it addressed, and again, for us in three years, a 100 times scalability growth.

JR Rivers

Kind of along those lines, there is some interesting anecdotes that are industry-wide around building up technology, making sure that it's massly applicable and then accelerating it. I'll give something little bit outside of the networking space. If you look at the latest Intel CPU chipsets, for the three or four years ago, I don't know how many of you know this, but they took AES, which is a cryptographic security mechanism that we use for transferring sensitive data across the internet story and file systems.

It became clear that AES is where everybody was heading for on the encryption side, and Intel reduced that to a couple by super-simple, super-fast instructions to make cryptography very easy to both write-in applications and high-performance on general purpose CPUs.

And you look at that, it was really awesome in the grand scheme of things, because we used to have these hardware appliances that would do cryptography. They were super expensive, purpose-built, and now every CPU can do that. And that's where you want to start to take scaling issues. You want to do it in software first. Make sure it's broadly applicable and then put it down into hardware once you understand it really well.

Rod Hall - JPMorgan

Let's talk a little bit about the switching market, obviously, dominated by Cisco today at their bare-metal options making in-roads now, especially at some of these mega-scale guys, as Martin said at the beginning. How do you think the switching market is going to look in a few years, four or five year's time? Is it dominated by bare-metal? Is Cisco still a big player by selling more commoditized switches? I mean how do the things play out over time. Maybe, JR, you could start us off.

JR Rivers

We look at the networking hardware market as being relatively mature at this point in time. And as markets, when you look at the history of the market where Cisco grew up, and I was personally a Cisco for quite some time helping it grow, we had to do a lot of the technology from scratch. We were the innovators. We built things downstairs. We wrote the operating systems from scratch. We designed all the ASICs. Now, almost all of that is done by somebody else.

And in a mature market, you start to see kind of a horizontal stratification of the supply chain, which in the end gets to the point where people can buy roughly the same hardware from a lot of different avenues. And when you get -- and that's what we call bare-metal networking, is where you can go back through and buy an optic from Finisar, instead of buying a relabeled optic from Cisco, you buy it from Finisar. Currently, the price difference is about 6x to 7x. So you can buy kind of more optics for less money -- sign me up, why not.

And so we see that progression continuing. In fact, we know it's going to continue because we look again at the leading customers, the mega-scales, the financial services, the pharmaceuticals, and they are driving that transition inside of their own organizations right now. So I don't see it ever stopping.

So I see the networking, hardware capacity market devolving into a hardware supply chains and software supply chains. You might get them kind of combined at resellers. For instance, we have a relationship with Dell. So an enterprise can go to Dell by their networking, system IT hardware by our software from one source as a reseller, so the route to market is one way, but it's still a desegregated offering.

Rod Hall - JPMorgan

Do you think that means that the hardware vendors overtime just end up combining self-storage compute networking, all kind of belongs in one place, one single supplier or do you think they remain specialized?

JR Rivers

I think you're going to see in a slightly, how to say it -- right, I think you're going to see a market where you have a hardware aggregators, so use Dell as an example, where as a customer I can buy storage compute, optics, cables, discs and networking, everything from Dell in the hardware side.

They're aggregating that. I can also re-buy or buy a bunch of software through Dell, that's resold and build out a full data center. The Dell is one of VMware's largest resellers as well. So you can put together very complex systems going directly to Dell to acquire the technology.

At the very, very larger scales, let's say, you're Google or an Amazon or something like that, you're likely to go straight to a manufacturer to get access to individual hardware. And I think what you're going to start to see is that same manufacturer that Amazon reaches out to, to build some networking hardware, computer hardware, is going to be the same manufacturer that Dell gets their hardware from. So customers get kind of one degree of separation from the original technology.

Rod Hall - JPMorgan

I guess no SDN panel would be complete this week without referencing John Chambers' comments last week. And for those of you who that didn't hear, I mean he said that he has won back a lot of business from VMware, which I actually wasn't aware he'd lost, but never mind that. So I thought I'd hand it over to you guys to maybe discuss, just what your reaction to that. I mean what might be causing these kinds of comments?

Martin Casado

It's actually been very interesting to us that Cisco has kind of highlighted us as the number two player in networking, which is very interesting, because any opportunity that we get is a 100% upside for us. And in fact, it has totally changed the nature of our sales motion, which is, two years ago, if you're like, okay, let's say, some random enterprise is going to be making a decision on their networking architecture. Who do you think they will talk to, probably Cisco, probably HP, maybe Alcatel.

Today, anybody, anybody that's making a decision on their next edition of architecture, I guarantee it, we're in the conversation. And it's a 100%, because John Chambers has made us the number two player. We are in every single conversation, conversations that we probably shouldn't be in, ones that aren't virtualized, ones that don't have -- aren't trying to compete with the cloud, but we're in every single conversation. So for us it's been a huge, huge windfall, right.

Now, as far as the actual comments themselves are going, it's actually kind of conflating a number of things. I'm going to let JR talk more about it. But Cisco is selling their 9k, which is a great cheap switch. I mean it's a cheap switch, but it's not a competitive or something like NSX.

NSX is a network virtualization layer that offers L4 through 7 services implementing software at the edge. So these things are purely compatible, right. And so I expect that Cisco sell a lot of cheap switches, and that's great, because the flatter the network, the dumber the network, the better it is for us. And Cisco can fastback as well as anybody and better than most, which is good for us, right.

But this is not an either/or thing, right. There are many features within NSX that we can deliver to customers that you cannot do with even the ACI product that doesn't exist today. And even when it does exist, when it does land, we will still have differentiation within NSX. And so I expect Cisco to continue to sell hardware. We think that's great. We'll continue to be compatible and overlay on top of it. And I think there is a lot of room for both of us.

JR Rivers

I'm going to put my VMware hat on here, which I am not, I am not a representative of VMware, but one of the things that I always find interesting in some of those comments is, Cisco will talk about ACI, which is their next-generation kind of networking hardware, whack-a-mole systems project. And they compare it to VMware and it's not the same thing.

VMware is a workflow orchestration tool. It allows customers to take a virtual machine, put it anywhere in the world, have storage disaster recovery backup, stitch it together with network virtualization, put a Layer 4 through 7 services between that VM and anything else in the world, whether its VMware developed, L4 through 7 services or third-party. So it's a phenomenally complex, extremely matured tool.

Outside of anything that people have build purpose in the mega-scales, VMware is the only commercial application that's available to customers today. And it's been doing this for longer than most of you in the mega-scales have. So I'll take my VMware hat off. I just think the argument and the relationship is ludicrous, the way that the fact that they're putting these two things is still compatible and equivalent.

But back to some of the other of Chambers' comments, from my side, again, focused on kind of the bare-metal, high-capacities to deploy switching hardware perspective, it's hard for me to watch as an ex-Cisco employee, John stand up and say things that are factually untrue.

I don't want to kind of get into the tit for tat of that, but there is a bunch of things that he listed as kind of gospel truth out there that are not true, things like who their customers are, who their customers aren't, where their customers are going, cost differentials, all of these things are patently false.

And I'm not sure he knows that, I'm not trying to say that he is lying to you, I think he is being told this by other people, and then he is saying, okay, great, somebody told me this, and so I must believe it. If you want to research it a little bit, there was a statement in his address about the fact that he argued that Cisco is less expensive to deploy than bare-metal switching. And the numbers that he got from that were based on an outlier deployment. This is a mega-scale data center that offers super sweetheart deal to keep us out of the deal.

So we know exactly who the customer was. We know what they're paying for Cisco. We also know what our final quote price was. There is a blog on our website that kind of references this. And even worse the Deutsche Bank analyst who wrote the report that Cisco leveraged in their collateral, once he got -- he was told all the right information, he kind of updated his analyst report to indicate the new facts.

So the analyst they took that the data from, came back to, oh, my dear, it was little bit wrong, it's actually more like this. And they didn't follow the new data. They took the data they wanted to see, not the data that was on the table. So that's disheartening to me. I'm a private company. I can say things like that.

Rod Hall - JPMorgan

I agree with you. So I wanted to move on and maybe talk a little bit about the complication story. That's another one that's out there. You know, gosh, if you don't use a vertical provider, it's too complicated to stitch all those together, who is going to do that. Maybe could you guys address that a little bit. I mean how complicated is this really? How is the level of complexity and difficulty going to progress over the next few years, because it maybe it is something more like a J.P. Morgan sort of thing? Our technical capabilities would be needed to do it, whereas a few years from now maybe not.

JR Rivers

And moving into just kind of in a layered approach, I'll go first, because we provide a foundation and then Martin can go second, since you kind of lives on top of that foundation. On our front, again in this high capacity bar e-metal networking, it's super easy.

Even right now, we have a ton of transactional customers. They call up, they login. They can try the product kind of remotely without paying any money. They try it out. They love it. They buy it. Shows up in their network or on their loading docks, they stick at a network and they start using it almost immediately.

One of the thing, you asked yourself, if you are a private company or a company that's building up around the ecosystem of bare-metal networking, you're a software provider, what's you number one goal in life? Make your software super-easy to acquire, super-easy to install and super-easy to use. I mean that's all we focus on. Those three things, we're very super crisp on that. And so we see it all the time.

We have a couple of customers right now that are building clouds, again reasonably large customers that they get the technology, they've been buying networking, deploying networking from a broad host of providers, and they get a hold, and then they go, damn, this is easy to use. I completely understand. This is really easy to deploy, really easy to use, really easy to license. So we get that feedback consistently from our customer. So I think it's really just more of a matter of time than it is a matter of any real complexity and FUD as you pointed out.

Martin Casado

So I actually just want to echo these comments, maybe a little more specifically, which is like the history of computer science is horizontalization and that actually leads to better products. So I actually come from the computational background. I used to work on mainframes, and like every manner of high performance computer, where the vendor would do the hardware, they would do the software, they would do the debugging tools and all of them were junk, right, because like listen, these are not domain experts in compiling, these are not domain experts in debug, these are not domain experts in operating systems.

And this is what happens when you have vertical integration, in which basically one company has to do everything and it evolves on the speed of the slowest piece, which in this case is going to be hardware. So every time you see horizontalization, you see things like Intel and Microsoft and Linux, and all of the third-party operations in debugging and management tools in a rich ecosystem, right. I mean this is like how the last 50 years of computer science has played out, and the best products have always down in a horizontal market.

And so the exact same thing is going to happen here, where you've got experts in the operating systems, you had got experts in the network operating systems, you had experts in the virtual networking layer, you've got experts in building ASICs for hardware, and when this happens, the best technology is going to win, each layer can evolve at its own speed, which means that he doesn't have to wait for an ASIC chip in order to kind of evolve his software. And I think you're going to get much more manageable and much more feature-rich systems as a result.

JR Rivers

And also interestingly enough in a strange way, it forces all of the players to be customer-friendly. When you have a lock-in, because you have the end-to-end system, you're allowed to show up and have things not work really. I mean, if you look at some deployments around current technology, you'll see large consulting companies that popup or companies that say, hey, my next growth is in my services business.

That basically means my stuff is hard to use, so I'm going to make a bunch of money after paying people to use it. When you have horizontal stratification of industries, what you tend to see is almost the opposite where most of the people and the application stacker fixated on, trying to make their technology easy to acquire and easy to deploy, because if you're not, you're going to get swapped out.

Rod Hall - JPMorgan

Let's move on to some of the technology change. I mean there are some technology differences with SDN that are talked about a lot and I think there are others that aren't talked about as much. And I want to focus on maybe one of those here with this question, which is, there is one significant shift with SDN, which is that the switching mechanism is much closer to the hypervisor, in the hypervisor, and I think there are probably some unique advantages of being in that position. And I thought maybe, Martin, you could talk a little bit about what that enables from a networking point of view?

Martin Casado

So one trend that you guys should be very interested in is as functionality moves from the network infrastructure to the endpoint, it enables things that we've never been able to do. And this isn't just a VMware thing, like imagine if Google or Facebook or Yahoo or whoever, if they're reimplementing security controls as part of the application, it's a very powerful, because you've got tremendous context, like you now the users, you know what data's transact, you know what needs to happen. So you have like all of the switch context when you do that.

And so the same thing goes as if you remove networking functionality into the hypervisor, it's actually in a very, very unique position that it has visibility into what's going on in the end-host. But it still has the isolation of infrastructure. And this cuts across all sorts of verticals, right, this can be used to help with performance in the data center in ways that networking has never been able to do.

It really helps with security, right. Imagine next-generation firewalling where you actually know what's going on, our vulnerability assessment, our network access control. And so I think, step one of network virtualization is reimplementing L2 and L3 and providing a lot of the operational benefits and the economic benefits that we talk about. But going forward, we can use this very unique position in the edge to kind of extract context and value in ways that we've never been able to in an infrastructure.

JR Rivers

I'll do kind of elaborate on it a little bit more, in that Martin, in his opening statement talked about some of the mega-scales where they've gone through and implemented an application stack top to bottom. And they were able to leverage the underlying hardware whether it was some characteristic or call it even network or some characteristic or quality of storage. And they were able to do that because they own the application stack.

For most enterprises, you pointed 7,000 applications you're running as fast as you possible can. You can't coordinate that level of complexity. You need to be able to put it into an environment that is, grossly vanilla, we call it kind of generic infrastructure.

The place, that position of a hypervisor is slightly beautiful, and that you can take a general application and stick it there. A hypervisor has visibility in a lot of the same context that you would leverage, if you're running the application stack from scratch and making enact, many or most of the same functions that you might have gotten, had you done that.

Clearly, it takes time to evolve and figure out what matters and doesn't matter and move forward, but it's in that unique position. Some of both the advancements in the hypervisors as well as the advancements in the underlying silicon like the Intel's and AMD's, have allowed the hypervisor to also be a very thin impact on the application operation.

Back in the early days the hypervisors, everything was done in software, page table updates and all the frame transmissions and everything like that was all done in software, which slowed things down. But this is kind of back to the other piece, we recognize what hypervisors needed, the CPU companies added some reasonably straightforward hooks to make the hypervisors much more efficient, and you've got a situation where you can run a single application on a server with a hypervisor, so that you get the performance of the whole server, but the leverage of the hypervisor as an infrastructure element.

Question-and-Answer Session

Rod Hall - JPMorgan

Let me just ask the audience if there are any questions out there. I want to give you guys a chance to ask a couple questions if you have them. Yes, over here on the right.

Unidentified Analyst

I'll speak louder.

JR Rivers

I'll repeat it as well.

Unidentified Analyst

I hope you could just clarify your, just to expand on your statement that ACI is very different than VMware generally, with Nicira specifically. Can you just explain that?

JR Rivers

So the question was can I expand upon the statement that ACI is different than VMware generally and NSX or Nicira specifically. The Cisco ACI, at least the pieces that they've kind of leaked out and shown the pieces, all the publicly available information, it provides effectively two big elements, one is the network virtualization function and then the other one is a policy underlie that they can push out things like access control policies and hardware.

And I guess, the last third piece is, they have a few little hooks that they talk about around maybe data path optimizations and certain constructs. Grossly speaking, the feedback we've gotten from customers that have tried this out is that some of these things are kind of nebulous at best, even now.

If you look at VMware NSX, Martin is probably the better person to talk about this, but you asked me. NSX provides that virtualization function, but on top of it as a product set it provides Layer 4 through 7 network services as well, so firewalls, load balance, intrusion protections. I think I would say in general, I think you're seeing that ecosystem evolve, where right now some of them might be wholly-owned by VMware and some there is also third-party options that VMware enables through the NSX network virtualization pieces.

I think you're start to seeing merging of some of those technologies where you have industry experts, kind of figuring how to put pieces together and share the wealth. And then obviously in the VMware, as a whole, when you're a customer and you're deploying VMware, you look at it as your workflow orchestration tool.

You go to a Web UI and you click on, I need 30 VMs and a 3-Tier web app and that uses VMware to deploy all the VMs, stitch together all the storage, boot these things up, get their networking on their I/O, deploy their routers, deploy their firewalls, all of those things, that's what VMware does. ACI does none of that.

Rod Hall - JPMorgan

Here is another.

Unidentified Analyst

[indiscernible].

Martin Casado

So I would say that where it's the least applicable is for those environments where the customer owns everything including the application stack. So otherwise NSX is applicable. There is two primary used cases for NSX that drive sales -- well, there is three.

One of it is provisioning time, right. It takes time to provision the network. This has nothing to do with the scale it has to do with agility. So whether you've got 10 servers or 2,000 servers, if networking provisioning is in the way of you onboarding a customer or a new application or whatever NSX adds value.

The second one is security, which is if you put a network virtualization you reduce the attack surface significantly. It probably drives 40% to 50% of our sales. It's just reducing the attack surface in the datacenter. And again, this is not an issue of scale, this is just an issue of, I mean, if security is important to you within the datacenter.

So I would say it's generally applicable along these two used cases, independent of size, but if it's a bare-metal only, the only application, they've reimplemented these control somewhere else within the application stack and then the value proposition reduces.

Rod Hall - JPMorgan

We've just got a few more seconds to go here. Let me maybe get you guys to finish up by talking about what we should look for as key milestones over the next 18 months or so as SDN and bare-metal progress. What do you guys think will be the key kind of milestones that investors at least should be watching for?

Martin Casado

So I mean for me right now this is the point of the customers, right, which is, I mean look at customer adoption and then look at innovation on top of the platform, because these are the two most important things. So like who is consuming this, like for example, in the last six months, we've announced kind of major customers in every vertical. So look at the customer traction, because that shows adoption.

And then second, once these are being adopted, look at the feature of velocity that comes on top of it, because a major differentiation of a software-based approach to a hardware-based approach is the ability to bring new features to market quickly. And along these two vectors, I think it will show you how this is progressing.

Rod Hall - JPMorgan

So you think by the end of this year we should see some major customer adoptions, be able to start observing that.

Martin Casado

Customer adoption and features.

JR Rivers

And so from our side in the bare-metal networking side it's already happening. If you look at -- I can't tell you the name, because every thing is under NDA, but if you look at the customers that we're working with that are transitioning to work with us or in some super big cases are deciding they're going to go off and build their own technology, it just continues to driver the market.

So I think you'll see that in customer adoptions. The big milestone is start looking at Cisco switching revenue and Cisco switching gross margins, you're going to continue to see that under massive amounts of pressure, because they have been and will continue to compete at an unnatural level to try and keep some of those customers.

But what we keep seeing over and over again is the customers they say, okay, great, I got a fantastic price from Cisco, I got a slightly better price from you guys, actually like your software better, so I'm going to make that go. And we're not seeing this from a 2% or 3% company, we're seeing this from pretty large providers.

Rod Hall - JPMorgan

Okay, guys, thanks a lot. We'll finish up there. Thanks everybody for coming. Thanks Martin and JR.

JR Rivers

Thank you.

Copyright policy: All transcripts on this site are the copyright of Seeking Alpha. However, we view them as an important resource for bloggers and journalists, and are excited to contribute to the democratization of financial information on the Internet. (Until now investors have had to pay thousands of dollars in subscription fees for transcripts.) So our reproduction policy is as follows: You may quote up to 400 words of any transcript on the condition that you attribute the transcript to Seeking Alpha and either link to the original transcript or to www.SeekingAlpha.com. All other use is prohibited.

THE INFORMATION CONTAINED HERE IS A TEXTUAL REPRESENTATION OF THE APPLICABLE COMPANY'S CONFERENCE CALL, CONFERENCE PRESENTATION OR OTHER AUDIO PRESENTATION, AND WHILE EFFORTS ARE MADE TO PROVIDE AN ACCURATE TRANSCRIPTION, THERE MAY BE MATERIAL ERRORS, OMISSIONS, OR INACCURACIES IN THE REPORTING OF THE SUBSTANCE OF THE AUDIO PRESENTATIONS. IN NO WAY DOES SEEKING ALPHA ASSUME ANY RESPONSIBILITY FOR ANY INVESTMENT OR OTHER DECISIONS MADE BASED UPON THE INFORMATION PROVIDED ON THIS WEB SITE OR IN ANY TRANSCRIPT. USERS ARE ADVISED TO REVIEW THE APPLICABLE COMPANY'S AUDIO PRESENTATION ITSELF AND THE APPLICABLE COMPANY'S SEC FILINGS BEFORE MAKING ANY INVESTMENT OR OTHER DECISIONS.

If you have any additional questions about our online transcripts, please contact us at: transcripts@seekingalpha.com. Thank you!