 Welcome to a special coast-to-coast edition of CUBE Conversations. This is Stu Miniman with wikibon.org from the Marlboro, Massachusetts office. Going to be talking about networking in hyperscale environments. At wikibon, we've been talking about how hyperscale is greatly affecting infrastructure and also how that is potentially going to bleed into enterprise environments. Joining me for this segment is Mike Jockamson. Mike is the director of product marketing over at Emulex and he joins us from our Palo Alto office. Mike, thanks for joining us on theCUBE. Hi, thanks, Stu. It's nice to be here and talk with you today. Great, so Mike, first of all, I know terminology-wise, we know what hyperscale is. When I talk about it, it's usually kind of those largest kind of 10 to 20, companies, many of them started the web, so companies like Yahoo and Facebook, the big cloud guys like Google and Microsoft, and of course, Amazon can't be left out. Why from an Emulex standpoint is hyperscale, something that caught your attention and how does that start that conversation off for us? Sure, the hyperscale companies, as you mentioned, are driving a lot of innovation in networking today. They're all building their own environments from scratch and whether it's open source software, they're building their own computers, specking them out and having them built directly by the ODMs, they're looking at open compute as a standard server platform, and all of that bleeds over into the networking as well. So there's innovation going on across the entire stack because of this hyperscale community and the unique needs that they have in applications and networking. And that's attracted to us, to the opportunity there because they present a new market opportunity. In fact, they present as if they're an OEM themselves because they are buying raw components and building their own infrastructure. Yeah, that's a great point, Mike. Boy, you had a lot of statements in there and I think we're going to spend some time unpacking that. Many people might not know what an OEM is and open compute's new to many of them too. So absolutely, I think we're seeing innovation is starting at these largest companies from a networking standpoint. It's always been some of the big companies that have helped driven forward some of the new generations of architecture. So as we could do speed bumps from 10 gig to 100 gig and beyond, there's needed to be some of those big companies that helped drive that. And it's not necessarily just speeds and feeds, but new ways of building things, new ways of operating models. So I guess I'm wondering if you can help us kind of compare and contrast conversations you have with some of the hyperscale players versus the conversation happening in the enterprise today? Absolutely. What we see in the hyperscale community, I mentioned this, build your own. It stems from the fact that these guys have probably one or two key application workloads that are driving their networking needs. And therefore they can design an infrastructure including the networking to support very specific workloads and traffic patterns. So they're looking for that minute detail, that minute level of programmability and control versus your traditional enterprise. I heard a comparison recently where one of the hyperscale guys had two applications and a financial services customer had over a thousand applications that they had to worry about on a daily basis impact in their network. Mike, if I could actually comment on that. So you know, it's a key point in the research note that I wrote on networking and hyperscale environment, which we said those hyperscale guys, they've got a big group of PhDs and they will actually spend a lot of their time and thought to build that configuration and it's very custom. So they spend that time to save money as opposed to the enterprise. The enterprise is overworked and they've got so much legacy baggage and so many applications that they need to spend money to help them save time. So I think that kind of sums it up. We do a little pyramid, which if you go to wikibond.org and look at the software led infrastructure page, you know, we've got a lot of pieces. Yeah, please continue on that note. No, absolutely. And your comments are right in line. These hyperscale young companies with lots of PhDs, it's like one big science experiment. They're trying to build the biggest fastest network they can with no constraints from their legacy. And so they have the luxury of designing from the ground up versus an enterprise that has 30, 40% of their IT budget going towards maintenance of their existing environment. And they're constrained by that. They have an inability to innovate as quickly. And there's also this natural inclination to buy commercial software and hardware that comes with support contracts because there's security in that and it lowers their risk. So it's a risk-reward ratio. It's a legacy versus build from the ground up ratio. So there's a lot of constraints that are inherent in the enterprise as well. Yeah, Mike, it's interesting. If I think about how the enterprise builds something, you know, they're very risk adverse. And while a network needs to be built so that components can fail and things can fail over and they can reroute and everything, typically they want to build it with such an architecture that nothing ever breaks as opposed to the hyperscale guys I find, you know, you take the Netflix model, they build in what they call the chaos monkey where, you know, inherent in the system, everything's going to break, everything's going to change. And it's kind of a different mindset is, is that some of the discussion you're having too? Oh, absolutely. You look at that scale out type model where I've got millions of components across the infrastructure. And when one breaks, I simply pull it out, replace it and walk off versus trying to maintain that particular server storage module. It's definitely a different mindset that you see in that hyperscale community. Yeah, you have very different mindset. You know, when you think about, you know, they want shorter refresh cycles, they need cheaper components as opposed to the enterprise wants to make fewer purchases, they want whatever they buy to squeeze as much out of it, keep it as long as possible as opposed to the hyperscale guys. They want it to be upgraded faster because then they can take advantage of new technologies and new scales and they're going to keep building on top of it. So, Mike, when we talk about networking, of course, we have to talk about software-defined networks for SDN. And, you know, I definitely have heard some of the big companies, you know, leading some of the SDN environments. You know, Google is a great example that I've seen. You know, they've been using, you know, some of the tools that make up some of the SDN toolkit, you know, for a few years now. Can you tell us, you know, what are you seeing out there? What's really real today in SDN? And how does that fit into kind of the hyperscale in the enterprise environments? That's like asking, what's real in cloud computing? You know, SDN is a concept. And within that concept, there are a lot of practical components that can be implemented. So, you see in the infrastructure that's called SDN, this extrapolation of the control layer from the data layer. And in between that, you've got a layer of APIs that will manage traffic north and south. And up on top of it, you've got capability to add in additional services. So, there's a lot of complexity there. And different companies are attacking that at different layers. So, we see through things such as OpenFlow, there's the protocol in the middle that will allow you to manage the control layer. And OpenStack, I think, is one of the big innovators that's looking at the top to bottom stack, including SDN, and that will create a framework for which all of this can fit into. And a lot of the vendors are beginning to develop towards the various components of that infrastructure. But looking at SDN holistically, it's still more of the concept of how you extrapolate those services away from the physical hardware underneath and allow it to be managed heterogeneously. And there's lots of ways you can accomplish that. Yeah, so, Mike, I got a poke, because you mentioned cloud there. So, I think we've made progress in cloud because there are numbers of solutions out there that are doing great in revenue. So, Amazon Web Services, over $3 billion last year, SaaS applications, everybody's using them. So, some of the private environments, maybe you can have a little bit of a point there. So, it's part of the problem that we don't have a clear leading solution, and I would differentiate between, there's lots of pieces of SDN. Cisco has their ACI, which has just recently come to market. VMware has NSX, which has been on the market a little bit more, but still hasn't fully transitioned from when they acquired it in a Sierra to a full VMware product. And then there's lots of other players working on pieces. So, I can't look to anything and say, there's 1,000 of those pieces deployed out there. No, it's always the definition. Not the maturity yet. Sorry, it's always the definitional issue. What is SDN and who's definition of SDN is the right definition. So, I think there's work going on in the standards community and the communities out there too, to try to settle on a single definition, much like they had to do with cloud computing in the past, which is where my comments came from. Yeah, go ahead. Now, please finish up. As the definitions settle and become agreed upon, the various components that make up SDN can progress from proprietary solutions to more open solutions that interact with each other in open ways. Yeah, that's a great point. And actually in your opening statement, Mike, you mentioned open source. And I think that's going to be key here because we can't wait for the traditional standards bodies to move these forward. It's going to take too long. I mean, I worked on Fiber Channel and iSCSI and Fiber Channel over Ethernet. And those take years to go through the standards body to get complete. You mentioned OpenStack. OpenStack Neutron is being worked on by lots of companies. There's code in place that people can take today and do things with and lots of people are coding to make it better. And OpenStack too is lots of different projects being worked on by thousands of people and try to put those pieces together. So hopefully we will go through that kind of storming, norming, and forming a little bit faster this time. And I think open source is a key component of that. Absolutely, and I agree with your comments. And I think we will move progressively faster through each of these iterations because the market's demanding it. And there seems to be a lot of market participation in things such as OpenStack. So yes, it's a vendor led community and there's a lot of vendor activity. But there's actual end users in the community who are helping to refine and define this. Yeah, absolutely. So we had the Cube at OpenStack Summit. I was really excited to be there. We did a ton of interviews. Is Emulex an active participant in the OpenStack effort? Yeah, so Emulex is active in the OpenStack community and our products are all supported in the OpenStack environment. So there's a lot more to come. Okay, great. Speaking of Emulex, I've worked with Emulex for many years and most people understand that Emulex has a strong heritage in storage networking. And if I look, storage networking in many ways is separate from the rest of networking. Can you talk to us as, is this new revolution in networking going to change that or we could finally have full convergence? What do you see as the change there and how does storage networking fit into this whole discussion? Sure, absolutely. Our heritage is from the networking side of things, long time heritage in the fiber channel area with block storage capabilities. And that's actually how we got into the Ethernet side of networking as an FCOE vendor a number of years ago. And from that, we realized that there's a larger opportunity and there's a lot more innovation going on across the board in the Ethernet networking space. In fact, Ethernet is really the most flexible networking backbone in existence today because it allows you to converge all of these different workloads, storage, virtualization, rocky RDMA, low latency type applications on top. So all of these are converging on that one backbone called Ethernet or TCPIP, more importantly. And it has become the backbone of choice for convergence of all of these enterprise workloads. And that's what led us to become a networking vendor, not just a storage vendor. Okay, when we specifically talking a little bit about the hyperscale guys, how important is networking in the architectures that they build? Networking is critical because the concept of any business is that I use my compute infrastructure to run applications that will provide business intelligence to people to make decisions. It's all about the application. And when you run applications, inherently you require the underlying compute, storage, and network infrastructure to do that. Well, we've always scaled the compute and storage infrastructure pretty quickly over time as applications became more complex. And the network has typically been that poor stepchild. So with the advent of 10 gig Ethernet and convergence of more technologies on top of 10 gig Ethernet, there seems to be a lot more concern about the ability of the network to scale and meet the needs of even more workloads on top of it. So networking is a very critical component of that hyperscale community. Yeah, Mike, can you give us any examples of how scalability and agility are being built into the networking products? Sure, absolutely. As we mentioned, 10 gig is the latest innovation in Ethernet, but 40 gig Ethernet is available. It's beginning to be deployed within a lot of the core networks for enterprises and some of these hyperscale companies and slowly creeping out to the edge. So the 100 gig standard is coming down the pike. So Ethernet is continuing to grow in its ability to provide bigger bandwidth, higher IOPS, and more transaction throughput. So the scalability is happening over time, but on top of that, within the pipe, there's also more of an ability to carve up your network and provide better quality of service and provide lower latency for applications so that they can get better end user experience performance. Okay, yeah, now obviously I understand we've always got kind of the leaps and bounds that are going on in performance. I'm curious, scale out architectures or something that are discussed a lot. Arista recently IPO'd and they're kind of known for being in not just the high performance computing environments, but bringing that scale out type of architecture to enterprise environments. So anything else you're seeing in that space that kind of makes, some of it's just the box design, but anything you want to comment on scale out? Yes, I think there's a huge move to build scale out architectures. If you look back to the typical high performance computing market, that was kind of the beginning of a scale out architecture where you could put hundreds or thousands of small nodes into a compute cluster and very rapidly calculate things by spreading the operations across multiple nodes. Well, if you take that same concept and apply it to enterprise type applications, you can do similar things in an enterprise network by putting more computer storage nodes together, clustering them and spreading the workload across multiple nodes in a compute cluster, which leads to needing the same type of network performance and scalability and agility that you used to have in this high performance environment in your typical enterprise network environment. Okay, beyond scale out, the other piece that typically is being discussed, especially if you talk to service providers or the large enterprise environments, is they want to act a little bit more like the hyperscale guys, and therefore they might want to build the specifics or their applications, be able to customize things, or just the role of developers in general has really grown in the IT space. What's the impact of that whole developer and application focus on networking? Sure, well my dirty little secret is I started in the world as a mainframe programmer, and my company's first token ring lands out on my desk, so I did a little bit of networking and developing in my early days, so understand that. Again, developers are all about building applications, and building applications is all about trying to build the best performing application that you can to run on the computer infrastructure that you have. So traditionally in a mainframe environment, we didn't have to worry too much about that. You built the application and the mainframe and the FICON environment handled any type of networking needs that you had. But as we've moved to more of these scale out networks, it's much more important to develop applications that are efficient and that can take advantage of the compute and networking capabilities to execute more efficiently, to get data where it needs to be more efficiently, and analyze data more efficiently. So there's a marriage of that development process with the networking capabilities that are underneath the infrastructure you're going to deploy it on. And that's exactly what the telcos are trying to achieve is they have to scale out these environments that will be able to serve up hundreds of millions of little iPhone applications that are all using APIs to tie into other iPhone applications to get data so they can make a decision. My client just checked in on Foursquare at the coffee shop around the corner. So I want to send them an alert that I have a sale on an item that they would typically buy. There's a whole lot of compute that needs to happen for that little transaction. And the telcos are right in the middle of providing all of the infrastructure and services for that type of a transaction to occur. Okay, wow, Mike, it says so many pieces of that. So first of all, you mentioned telcos. So we haven't touched on the other kind of hot topic in the environment was network function virtualization or NFV, from my standpoint, NFV came on a little bit after SDN. A lot of people have been trying to figure out how they compare and contrast. And we've written a couple of articles on that because they are kind of different solution sets. But the products and the revenue for the solutions that fit under that NFV bucket have actually taken off faster than some of the SDN pieces have. As you said, we're still trying to define SDN. What are you seeing from the NFV and kind of the telco carrier space? Sure, absolutely. We see that there's extreme growth in that market that the telcos are having to deal with the explosion of this new traffic from the internet of everything, if you will, any device you can imagine having networking capabilities and sending information across the provider network. And as that traffic and the demand for service grows, they have to build networks and environments that will scale to handle that. And their traditional forms of providing and monitoring and managing that network are breaking. They used to invest in specialty appliances for WAN optimization or for a deep packet inspection or content distribution networks, what have you, the old ATCA appliances that you bought as a complete appliance and plugged into your network. And as they're trying to build a more virtualized and scalable environment, they are working with the providers of those former appliances to build them as smaller virtual appliances. And we're right in the middle of that with some of the work that we're doing in that community to provide some capabilities to develop and supply these very scalable, fast packet processing applications that a telecommunication company can use to provide these services to their customers. Okay, wow, that's fascinating, Mike. The other piece we talked about, appliances are moving over to software. While so much of the discussions around software these days, one of the interesting kind of hardware related item is the open compute initiative. So, what are you hearing about OCP? And I know Emulix has a solution that fits in there, so can you tell us a little bit about that? Sure, we introduced our open compute networking adapters and conversion network adapters earlier this year at Interop. And we're working with a number of the server manufacturers, the OCP infrastructure suppliers to provide this capability. And it comes down to enterprises and web scales and telcos who are trying to build a standardized compute environment, whether it's the software with open source or whether it's your hardware with an open compute server or open compute networking now, all of the various components that you can put together for an open compute platform. And we see that there's growth in that area. Number of companies are still playing around with it to make sure that it's real and that they can run their workloads on that environment and scale the environment quickly and get servers that still bring them the reliability, security, robustness that they need. But we see that there's a real future in that and that there seems to be growth in that area as well. Wow. So, Mike, one of the things that always strikes me is there's probably more change going on in the networking space now than there has been at least in the last 15 years, probably put together. What are you hearing from CIOs today? Are they just buried in this sea of buzzwords and new projects coming on? Are they starting to wrap their heads around it? How does Emulix have that conversation? You doing anything to help retrain the workforce to bring them up to speed on some of these things? Sure. There's definitely still a lot of buzzword bingo going on out there in the ITU organization. But the reality is the CIO is trying to cut costs. They're trying to reduce the spend and increase their ability to provide these services to their own internal customers. We see, we call this the Amazonian of IT where somebody can go to Amazon and with a very few clicks they can get a complete infrastructure, a complete compute infrastructure, including storage. And in your own ITU organization to get that similar compute and storage takes a number of POs, a number of weeks to order and receive and install and provision the equipment. And so the IT organization is under pressure to implement the private cloud where they can look and act like an amazon.com to their own internal organizations and provide them very rapidly provisioned compute, storage and network where they can bring up new workloads in minutes or hours instead of days and weeks. Okay, so like since you wear a marketing hat, I need to ask you the question about Emulex. You guys obviously had a strong heritage in storage networking. You've got products and software and services around the networking. But a lot of people probably think of Emulex as a company that makes chips and cards in the world of hyper scale. And SDN and all these changes going on. How shall we be thinking of Emulex going forward? Sure, Emulex is absolutely a leader in storage and storage networking, but you probably didn't know that Emulex is the number two provider of 10 gig ethernet in the industry as well. Primarily through our OEM relationships with all of the large server and storage OEMs. So Emulex equals ethernet. We are a networking vendor. We provide core networking capabilities for enterprises, for hyper scale, for any environment. And we also see that there's tremendous growth in the need as these networks scale to now monitor and provide security in these networks. So Emulex is also active in the monitoring and security space with an acquisition we made about 18 months ago where we can do real time packet capture and analysis. So you should see Emulex as a holistic networking vendor providing endpoint access and security for enterprise and hyper scale networks. All right, well, hey, Mike, I really appreciate you helping us sort through the buzzword bingo and on your bingo card you actually get a bonus chip because you worked in a mention of token ring with always the old school networking people all excited. So Mike Jay from Emulex, I really appreciate you taking the time sharing all this. This is Stu Miniman with Wikibon when this has been a CUBE conversation. Go to wikibon.org slash S-L-I which is software led infrastructure. Check out much more research on everything going on in SDN, NFV and everything else there. And emulex.com is where Emulex is we appreciate them joining us for this segment. Hope you come back and check us out for more CUBE conversations in the future. Thanks Stu. All right.