 Good afternoon everyone. So this is the network virtualization panel We're gonna talk to with a number of the vendors and users of network virtualization and talk about its current state What we think it's going to in the future and a little bit about how that impacts open stack So before we get too much started here. I wanted to have people introduce themselves. I'll obviously go first My name is Ken Peppel. I'll be moderating today. Not with any of the vendors here But I'll let each one of them actually introduce themselves and a little bit about their solution and their company Dan Good afternoon everybody. Yeah, good afternoon. My name is Dan Demitriou. I'm the CTO of Mido Kura And we have a virtual network overlay solution for open stack handling layer 2 to layer 4 Please stop by our booth later I can tell you more in detail about exactly what we do but basically we cover the gamut of the open stack networking functions from Floating IP and that security groups virtual layer 2 layer 3 isolation. Thank you I'm Mike Cohen. I'm director of strategic alliances for big switch networks. We are a platform independent software-defined networking vendor You know our solution is distinct in that it actually is a Platform for a number of SDN Networking applications and network that virtualization turns out to be one of the one of the key solutions built on top of that You know if you look inside our solution, you know, there's probably Two distinguishing elements of it one is that it's actually designed around open standards and open API's And this is actually very very key to Offering programmable interfaces and being able to interoperate with things like open stack The other one is actually in addition to an overlay solution. We actually offer a physical and You know and virtual and virtual or hypervisor based solution where we can actually see the entire network and rather than offer You know in addition to offering overlays, we can actually give you a consolidated view of your network And and actually allow people to control the entire network So my name is Gavin Pratt. I'm the product management lead for cloud storage and networking at HP cloud services so HP is a big company we're a Both a vendor and a user of software to find networking and we've had some announcements past month or so and happy and Q&A to get into that in more detail, but You know from the cloud service provider perspective You know we're operating one of the largest open-stack based public clouds and You know we plan to move me quantum wasn't originally part of open staff But now that they're you know bringing it in as game more stable. We're planning to support quantum with sdn controller and happy to chat more about that and again obviously on the Hardware side in the private cloud side. We're planning to be involved there as well Again, there were some announcements last week or two on that and happy to elaborate in Q&A Excellent, so I've got a number of questions that I'll be asking them here over the next 10 or 20 minutes And then we'll open up for audience questions when we do open up for audience questions I'd ask you to come up to the front microphone Since we are recording this if you could speak up so that people who actually see this later can hear it But I'll open it up for them then but let's start with what do you believe the current best use cases for network Virtualization or software to find networking are within the context of the cloud and since I started with Dan last time I'll start with Gavin this time Yeah, I don't think so Yeah, I think one thing is worth mentioning is You know, I think software to find networking is really important. I think that's why the quantum sessions seem to be packed every one of them and HP is a very big believer in SDN But it's worth noting the SDN is not right for every customer in every use case I think some of the first adopters we're seeing of SDN are the cloud service providers And we're starting some large enterprises as well. That really revolves around two Q-value propositions of SDN that I think SDN is somewhat uniquely positioned to address. So one is enabling Network isolation at the tenant layer For large multi-tenant clouds, it's very hard to scale VLANs in a especially in a multi-tenant way At the scale that HP and Rackspace and Amazon web services are operating it So that's one important way from the cloud service provider perspective another important use case is we look at large enterprises I mean so be very in the HP keynote earlier today was talking about the whole notion of the production workload at the enterprise level and how Enterprises need to be able to span different data centers in a true hybrid cloud way and a lot of people talk about hybrid cloud But it's something that currently just a lot people are talking about But to really enable hybrid cloud you need to be able to have a single network topology that spans multiple data centers And do things like handle IP address overlapping Stuff like that. And so that's the second large use case we see in the near term So I would probably describe, you know, three key drivers. We're seeing around network virtualization You know the and they're really speed scale and efficiency So speed becomes important in sort of the rate of change that happens in In cloud, you know in cloud environments in the network, you know, it's no longer feasible for someone to actually You know Modify a particular switch or modify how a VLAN trunking, you know behaves in a network in a cloud environment That ends up too slow for the way clouds are actually supposed to behave They need to be you know, they need to allow, you know, second base provisioning not, you know, file a ticket and wait a week So that's driver number one scale tends to be the you know, another big driver, which is you know, actually, you know Using something like a VLAN, you know, you know, there are you know 4,000 86 VLANs available and that turns out not to You know not to scale to really large multi-tenant environment So as people are building clouds today, you know, they don't want to build them with that inherent cliff You're basically designed in from the get-go and SDN can offer an essentially unlimited scaling ability And the last one is efficiency and it's you know, the idea that you you actually end up You know prior to SDN you may end up with stranded capacity in your cloud data center Where you know a workload can't actually take advantage of another part of your cloud because you know the network won't actually stretch to allow it to Do so, you know SDN and programmable networks can actually allow the network to morph just as flexibly as the hypervisor virtualization layer to allow you to really use every bit of computing and you know computing and other resources in your cloud Yeah, so I think these guys pretty much covered it But if I was to add my my two cents to the to the point here We've been focused on infrastructure as a service Network virtualization for a while So the so the problems that come up there are of course the way you do isolation in an environment where you really Want to decouple the location and the identification of VMs and virtual resources and that's why we chose the overlay approach to that so We're talking about SDN in general SDN is such a broad thing It's about programmability and agility of hardware pieces as well as software for solving the network virtualization use case for infrastructure as a service like OpenStack The overlay approach seems to make the most sense because it allows us to completely decouple the location of things from their Identification right including enabling the migration without relying on the traditional isolation mechanisms of VLANs or even MPLS VPNs or VRFs Which may or may not be available in lower-end hardware, let's say and It also allows us to do all this in software at the edge of the network and I think it was also mentioned before but by doing this at the edge of the network We can also We don't need to actually update any sort of state in the physical network in the core the core can stay rather You know solid only if there's a failure of a link or an actual hardware device Do we need to actually update? Network state there all the virtual network state it gets updated at the edge So if we were gonna have the same conversation, I was gonna ask you about these same use cases But we were in Barcelona perhaps next next fall. What do you think that future use case that best future use case that you're looking towards? I'll start with you this time Dan Well, that's a tough one actually. I don't know if I could think that far ahead, but basically we're looking at multi Multi-site use cases. I think that's gonna come up more and more right now We've basically effectively been focused on a single site network virtualization So I think that's gonna come up people talk about cloud bridging a lot more and the hybrid cloud use cases Of course of multiple kinds, you know interconnecting Amazon with rack space, but you know that's been tried The providers actually have to play nice in order to make that kind of seamless But a lot of it is a hybrid cloud on-premise connecting with the public clouds. That's definitely one of the biggest ones So I would say you know at a high level there's some of the use cases I talked about previously You're gonna continue to be true, you know very much so but you know if I kind of think ahead about other things that you know SDN and network virtualization will enable You know, we're beginning to see you know a lot of requests for layer 4 through 7 services And and how we can help those vendors basically integrate you know integrate into a virtual networking domain and expose their services Efficiently, so I think that that's one you know really key driver I think we're gonna start seeing things like you know an intelligent network where you can actually you know The network can actually start working with you know things like a nova scheduler and actually give it information About what's going on in the network and help it intelligently place VMs for particular workloads So I think you're gonna you're gonna start seeing a network that's very API driven And can actually start exposing information and now tightly integrating with other pieces of open stack And I think that'll be where things start getting very exciting So I mean certainly when you're thinking you know six months out, I don't know that the Use cases will necessarily be dramatically different than At least the two that I was talking about earlier, but I think we know they'll get better And so you know first of all when you think about it from a cloud service provider perspective of offering You know multi-tenant Network isolation at scale. I think you'll see the better ability to do that at scale So one challenge with SDN is on the one hand There's this campus as you know, let's pull out the control plane have a central control plane And then you can handle all the devices, but the problem is that doesn't scale super well So how do you on the one hand? Centralize a control plane, but at the same time distribute it so that actually doesn't have Performance bottlenecks such as there's single points of failure or some general databases that don't operate well as scale So I think it was you know as you start seeing people handle this, you know centralized yet distributed model That'll be one thing and also kind of like you know, we're saying earlier We're bridging multiple sites. I think you know initially you're gonna see that as a somewhat you know Not automated process, but I think you know down the road I think you'll see much more automation and integration across clouds I like they're saying you know we have to all play nice with each other and you know I think certainly one you know kind of requests to the community is we look at quantum currently a lot of functionalities in the core Apis there's also a lot of functionality that's currently being delivered by different vendors via different extensions and That that's not really sustainable if we want to see true Inter-cloud operability and certainly you know long-term like to see you know automation and management Security tools and kind of go on top of that but I'm to facilitate it more but a long term we need to see more Functionally moving to the core apis versus project writer extensions and once we see that then I think we're gonna see an explosion of different use cases and value added software and services that will You know enhance those two use cases I described Excellent, so we've talked a little bit about why people might be moving towards network virtualization What do you think the biggest obstacles though that are really facing people that are looking to embrace that today? And I'll start with Gavin down at the end again Yeah, um at the risk of being like a broken record so on the cloud service provider perspective Many STN controllers don't work. Well, I'll call it at scale, but when I say at scale, I mean, you know the You know the HP Amazon Scales certainly you can operate a small cloud with you know, even the free open source STN controllers out there But they actually work at scale you need You know a much more complicated and yet elegant solution and so that's something that I think one thing people are struggling with And then again on that second use case I think we already talked earlier for just having the technology and the open standards To facilitate the cross-cloud error connection. I think is another key issue and I guess I'll mention a third thing which is you know, so eBay for example was one of the first adopters of STN and they're using nice era But if you look carefully they're doing it just for a small depth test workload right now And so one thing that people are struggling with is even if they decide STN is right for my company And by the way, it's not right for certain companies, but for them it might be right. How do you rip out the entire? Network control plane and replace with an STN controller without destabilizing your business And so we're already seeing some vendors starting to think about that and start to develop solutions But basically having a seamless migration path that enterprises are going to require at least for non-green field deployments I think it's something that needs a lot more development before we see widespread adoption and existing Networking infrastructure You know, I would say, you know, one of the bigger barriers You know we're seeing right now is a bit of kind of a mind-shift a my you know You know a mind-shift that needs to happen in terms of how people think about networks and they what are the inherent constraints of network? So STNs actually You know start eliminating a lot of the things you thought you could not do in networking You can actually now do because your network is very is open and programmable But you actually have you know the people that have been running networks for many years are used to the way things operate And they're used to a set of primitives So it's you know, there's kind of a bit of reeducation and you know The awfulness that kind of has to happen in a community to really understand these new capabilities and how to deploy them And probably the other area I would highlight is you know Deploying software defined networking in a cloud is very much an integration place So, you know your network touches, you know touches, you know all pieces of your cloud, right? It's very important that it integrates with your servers It's very important that integrates, you know, whatever kind of WAN solution you have in place and whatever orchestration stack you're using You know, so it's actually really important that you know, STNs You know basically build these integrations and can actually deploy seamlessly across them And I think that's you know, that's happening very quickly right now But that's going to be one of the things that will drive adoption very aggressively Those are all great points and I would say one thing that I think is relevant to all our software based overlay solutions Is a question of trust, you know Basically the networking gear has been assumed to be something that's like this hardened piece of hardware That's delivered by a vendor and it just basically works seems I mean I know some people are gonna sit laugh because they're bugs all the time But surely it is not like a piece of open stack or even some piece of Linux software It's something that's assumed to be much more hardened and with the SDN and overlay solutions It's all brand new software. It's very new So it's gonna take a while to establish that trust and confidence that this stuff actually works, you know You know really mission-critical environment. I think that's one issue another issue I think that might have been touched on is about troubleshooting if you're using Very new ways of routing traffic around or encapsulated, you know a lot of it the traditional tools Maybe don't work anymore. We have to build new processes for that and that surely will take time And I think the vendors we have a responsibility to make sure that we offer either reasonable alternatives or integrate with the existing solutions if that makes sense So one of the other things that have come up quite a bit and especially Appropriate in the quantum community has been the issue of openness and the openness of solutions Especially within regard to vendor lock-in so kind of a two-part question on your particular solution Which parts are open which parts are closed and how important do you feel that is in the short term? Dan That's a great question actually probably the toughest for us here because actually right now none of our solution is open Except our integration with with with quantum and open stack So I think you know we could answer this in several ways In terms of vendor lock-in if the API is standardized as quantum aims to be as more and more of the API's move in there That's certainly should alleviate some of the vendor lock-in concerns Ultimately, I think we will have open-source solutions for the entire stack But that'll probably take some time, you know So in the meantime Something like quantum is great because it facilities facilitates the common API that somebody could use But I think given the heterogeneity of the solutions It's going to be a little bit tough to find like true commonality all the way Even if the software opens source that wouldn't necessarily mean that it's the pain of switching is actually Not there and that after all I think that's the that's the idea of vendor lock-in is you can always switch to another vendor Just how painful is it right? It's going to be painful no matter what at some point And I would say at big switch we've taken a much more aggressive stance towards open source and open standards You know, then I guess we've seen in other parts of the market So we have the core the the core of our controller Do you know that the kernel that actually speaks down to different switches in the network is open source? It's a patchy license. It's a project called floodlight and this is something we actually embedded in the commercial controller that we ship So that's actually you know one big piece. We're doing and this is available to the community And in fact floodlight has its own community building around it You know build it you're building different SDN applications The other thing we've been doing is trying to stay very very close to to open standards So, you know our controller speaks a number a number of protocols down to different switches open flow is one of them And you know open flow has been a friendly to us because it is a standard It's being picked up by a number of vendors on the physical switching side And it's also showing up in hypervisor switches and then we're obviously gonna you know We'll extend that to support VX LAN and VGRE and you know in different protocols as they become You'll really you'll really tightly supported standards as well But you know, I think this is where you're what's gonna really drive SDN is really solidifying these standards and and you know People using open source and open standards as the foundation for these platforms and that's what we've been doing You know, I think from the very foundation of the company really Those are HPs pretty similar You know certainly we think about any of the interconnection points that a customer might ever interact with We have a firm commitment to making sure those are on If not open source at least open standards so that they're you know, there will not be vendor lock-in the one thing we did talk about earlier though is You know currently in quantum as a small set of core APIs And it requires you by definition to use extensions to those APIs For a lot of the advanced layer three and other functionality So that's something that HP and I think Cisco and others are pushing aggressively to get You know more of those into the core API and what we generally want that because it's better for us as a community If more of these clouds are Can interconnect So I guess that that's the one limitation and it's limitation I think for all of us But again, we're firm backers of that but certainly when you think about things like you know If I want to you know create a you know a tunnel between the HP public cloud in my on-premise enterprise Data center, you know that's gonna be entirely via open source open standards You know such as IPsec and you know, we're certainly be supporting although not requiring Open flow on You know switches and routers so certainly if you have that I'll probably work best if you don't have open flow It'll our controller will still work, but it you know, maybe it won't work quite as fast But any open flows and open standard and HP is an active contributor to that community Great, so we've talked a lot about use cases and it seems a Very clear use case in the service provider markets But a lot of people here are probably looking at using open stack as a private cloud solution And so do you see big drivers today? That will push private clouds to actually use network virtualization, or do you think that's gonna come further down? Probably down the pipe later Kevin Yeah, I mean, so I think that there definitely is a reason to use that so You know, I think long-term we'll never see Most companies using a full public cloud model and part of it is Just based on the economics it often makes sense to kind of have your base low capacity Be something that you kind of if pre-bought and you've kind of locked in and then pay per use for you know Flexed room and in special projects and things like that So the implication then is you need to be able to have your private cloud Interconnect in with public clouds to get that full promise of the cloud being able to kind of flex up resources as you need to So the implication is then that you have to be using Some form of SDN so you can you know flex into the cloud But you know without having to you know break into a different network topology Which often makes it hard for legacy applications to function that way I Can say definitively we're seeing demand today in the in the enterprise for private cloud deployments you know with with open stack and the use cases really go back to what I said before you know speed scale and Efficiency there there's you know very similar to what you know very there's strong overlap at least what service providers are demanding Now the the one caveat I guess I'll put on it if you're deploying an extremely small private cloud You know a single rack potentially You know network virtualization you know a lot of its value is about you operating at scale So if you have a very very small environment, you know to some degree any you know in a single switch For example any environment you know any amount of networking would probably work for you You don't need something as complex But if you're deploying at any scale you often run into the same problem service providers do and that's what we're seeing in our customers Right, so I just wanted to say that you know a private cloud doesn't necessarily mean that it's a small one Right, so in that sense it could be a large cloud even inside the enterprise Also the IT the internal cloud inside an enterprise could operate very much like a service provider You know for different business units so in that sense I think there is a lot of overlap and I would say also that even at the smallest of scales, maybe even at the one rack level Network virtualization can add fault tolerance So in that sense it does add value there at for really not actual not much cost in terms of complexity So I think it actually does add value at at all layers Excellent, so a lot of people Today when they talk about security and security in the cloud is really being enforced or sometimes hindered Really by networking. What do you feel like? SDN or network virtualization? What effect do you feel like that has on security today in the cloud? Not my area of expertise, but I'll give it a go You know, I think the in the cloud the security is very much driven by the hypervisor Right, so you have to have the hypervisor across as part of your trusty computing domain and in that sense You know a solution like ours like an overlay based solution the Software that runs on the edge also has to be part of the trust of computing base So it has to be hardened so that it is not vulnerable to internal or external threats That said there are a lot of different kinds of security here right like there's threats from inside the cloud There's threats from people snooping on the actual physical network. Those have to be hardened by encryption and you know could basically Authenticating encrypting the control pads as well. There's also threats from outside and of course network virtualization can help there by providing much more visibility into The traffic that's going on at the edges possibly getting that into some sort of ideas analysis and really quickly responding to threats by blocking relevant traffic again at The edges without involving changes to the the physical network which might have to go through various types of Vendor specific scripts or God knows what tools I mean until every single device out there is Programmable by an open standard like open floor or what have you Doing things in software is still gonna be a lot more flexible. So Yeah, lots of things Yeah, I think you know honestly from a security perspective those both challenges and opportunities and you know in the challenge department It's very much that you know, this is a new this is a new networking paradigm You know, you're changing, you know, you're changing the deployment structure you've used before and you know people need to understand what those differences are and and and the corner cases around them to make sure that you know Mistakes are not made You know, but I there's nothing inherently that makes things less secure It's just you know, the world is changing you need to learn you need to learn how to adapt The big opportunity which do I think is significant is actually created in the NSDN controller actually has this global view of The you know of every piece every element in the network and all the different hosts and how and actually at the point At which they're communicating with each other So it actually gives you you know, it actually gives you a central policy location where you can set exactly who can communicate with Who and how? And this actually creates a really interesting opportunity to enforce tighter security than you did before and you know This can be you know, this can be at the edge Using edge-based devices You know to the degree your virtual network spans into the physical domain it can also encompass Physical physical firewalls as well and making sure those policies are synced up with your virtual networks So, you know at the end of the day you may actually end up with a better automated solution that can deliver better security than you have today You know, but obviously the challenge around that is you need to understand how to put all these pieces together properly So I mean I definitely agree with what's been said so far is just to elaborate I mean from a theoretical perspective it's easier to hack software than hack hardware. And so, you know, certainly You know where we're taking security very seriously at HP and certainly You know, we'd argue it's important to go with a trusted vendor You know could be HP could be somebody else, but you know has a history of You know taking security things very seriously at the enterprise level and You know, we're certainly leveraging some of our technologies like to being pointed in our site and others In the SDN space to make sure we are you know, we are doing things like intrusion detection and and others But like we said, you know as long as you do it right, I think there are opportunities and Certainly even in your term, I mean just offering tenant isolation, which isn't possible at scale without SDN At least in an elegant way Is, you know exciting so I just let you know just I'm happy. There's lots of opportunity, but You know, you want to make sure you have somebody that you can trust and has a track record in the space and You know security is not something to take lightly Great, so I'm gonna ask one last question But while I'm asking in there answering feel free to queue at the microphone so we can have audience questions here So on my last question will be what are your hopes or what specific features? Would you like to see in quantum to come out at the end of grizzly? Yeah So it's obviously security groups are one of the key features that I know we were trying to push into Folsom 3 and I think it's I think currently the plan is for grizzly one if not grizzly 2 so I mean The key is you know to fully move over to quantum first. We just need to have Future parity with the existing of a networking model and so that's one of the big gaps So obviously it's on people's radar is it is happening, but that's probably feature number one I think feature number two is I think we like we're all talking about is having more and more of that functionality into the core API's You know it because again the full promise of SDN Certainly from a user perspective, you know, you know going beyond just cloud service providers really depends upon having clouds Be able to interconnect via open standards, and that's not really possible until you have more functionality at the core API level So a lot of progress is being made there, but it is something that we as a community need to push I'm personally very excited by the work being done around layer 4 through 7 services and how they integrate into quantum APIs as well You know quantum, you know Quantum's made a huge amount of progress to finding things at layer 2 and layer 3 and you know But we're hearing now from people and you know we're also seeing other networking vendors Interested in saying well, how do we integrate higher level networking services into the same framework and you know Quantum's just now tackling that problem And I think that'll be I think that'll be worked out in the grizzly time frame And I think that'll you know it'll offer a great solution to people where they can actually do kind of an end-to-end Kind of full network stack provisioning in a net virtual networking sense And I think you know that's probably the next big challenge for for quantum Okay, well I see that I see the quantum API situation divided into two I mean not just quantum, but basically the the open stack network API Situation divided into two areas the tenant facing APIs, which basically I think need to be figured out and be very stable Right, so in that area. I think we should take our cue from AWS I mean not because we can't be more creative than that But because there's so much there to do before we start getting more creative like you know IP side VPN gateways all kinds of stuff and on the other side Right like ELB low balancing services, you know, that's been a hot topic. There's like four sessions on that these days And then on the other side the provider side APIs of quantum. I mean, they're extremely immature There's almost nothing there so each vendor presumably is doing their own thing You know for the for the provider side and I'm sure that's going to that adds to the vendor lock in the pane of Switching out right so the provider stuff needs to be standardized just as well as the tenant of course not as much But it's also important so Good are there any questions So picking up on that the topic of layer four through seven services What layer four through seven services? Do you think will be most popular other than load bouncing? That's an obvious one And where do you believe that the best way to integrate that in will be? So there's been a few different ways out there that you've seen in the market today One was really around automating hardware There were also a lot of talk of just doing software appliances other people have actually looked for real services People like mBrain and such where do you feel that that's going for layer four through seven services? Definitely load balancing top priority right because it's a fundamental requirement Maybe deep back an inspection firewalls things like that I have to say I really don't believe in integrating the hardware appliances into the into the model because we're an overlay solution So I think probably you all know why we don't believe in that And I think it also complicates things quite a bit right so I like the mBrain model You know fundamentally the scalable software appliances We'll see how it goes, you know So, um, you know, I I agree in terms of services people interested in it's going to be a little bouncing would be number one You know firewalls will probably be number two and there'll be a number of kinds of other security Intrusion detection for example, maybe you know, maybe also pretty high on the list You know as things you know people interested in in terms of the model that that ends up waiting here You know, again, I probably have the view that it's going to be probably a mix of you know edge-based and hardware appliances You know, so we don't actually build solutions at this level. We work with partners and the partners I've talked to You know the the hardware the hardware boxes tend to have a you know kind of you know some different use cases and some you know Fundamental capabilities that you know some of which can be matched and some of which can't be matched in the virtual appliances So I ultimately think that you know the solution will probably end up being you know needs to support a mix of these you know for a while you would and And I you know honestly with the APIs we developed should not you know should should not be closed off to working with hardware appliances So, I mean, I think Dan I are pretty aligned in a Most part I guess the one I mean, I think like Dan was saying I'm probably more of the mindset That'll probably be more at the software layer Just because of that enables more distribution of both low balancer and firewall functions And if it's not distributed as hard to really take full of the energy of STN the promise of STN Okay, so switching gears a bit just about adoption and where you see the market going today How fast do you think that customers will actually adopt some of the network virtualization that you see today? Obviously, there's been a handful of use cases and people announcing like eBay But how far do you think or how long do you think it will take to actually get a broader penetration? So just from your guesstimate when we're talking next fall 2013 what percentage of open-stack solutions do you think will have some form of network virtualization in it? Well, so if you read the latest garden report, they say that STN is five to ten years out from mass adoption So but that said I mean I think that kind of misses a lot of nuances. So as service providers are obviously adopting it today You know enterprises like eBay are starting to play with it, but they are there. I mean, I think even the more You know sophisticated companies like eBay are still probably three to five years out from mass to STN adoption But certainly for specific use cases, I mean we have large enterprises today that are engaged with us I mean today on leveraging STN principles to be able to you know connect two different data centers and burst in and But again, that's kind of on a per-use case basis and it is with more of the sophisticated IT for looking companies. So, you know, I probably say three to five years out from mass adoption a little sooner than Gartner, but And you know, I probably don't know if I could be as specific in giving you a specific kind of year guidelines But obviously, you know, this is a new technology and it represents a paradigm shift and a huge opportunity So you're gonna see kind of you know an early adopter to you know an early adopter to mainstream model that follows most other technologies So I think of you know, I was probably in a minority of folks here. You know, I was at the first VM world conference You know, you're rolling out hypervisors for people that were not used to running them and that technology, you know Took a number of years to kind of mature and now it now it essentially dominates the industry And I think we're gonna end up seeing that with software-defined networking as well It really will redefine what's going on in networking and you know, there'll be a number of use cases where it has immediate value I'll start getting deployed immediately. It's being deployed today. In fact, and I think next year We're gonna we're gonna see a large number of customers talking about how they deployed SDN But obviously, you know for for kind of old-school enterprises that you don't want to change anything It's gonna take them, you know a long time to get comfortable with changing. I'm an engineer So if you ask me, I think all of them should be adopting SDN by the end of next year And if they're not that's probably because the solutions from vendors like us and others aren't quite ready enough You know to meet their exact needs, right? And of course, there's a conservative aspect that kicks in so I'm just gonna focus on figuring it out what they need and implementing that Okay, excellent I mean we're kind of talking about this earlier, but there's a whole issue of you know greenfield versus non-greenfield And so certainly, you know, so eBay and their latest deployment was for a greenfield depth tests, you know solution in-house and so I think we will see more adoption for those greenfield deployments, but You know, I think it the challenges of you know vendors such as perhaps us or others need to think about is how do we give You know existing customers with large Installed network infrastructures and migration path. And so, you know, my prediction is over the next five years We'll see more and more companies come out with migration past solutions And if that were to happen, I think it could rapidly accelerate the adoption that you were saying you hope think make sense It's I guess again, I mean people are doing it from a greenfield perspective because there's no switching costs So with that if there's no any audience questions, I'm gonna go ahead and wrap up Sorry, could you come to the mic so that we can ever record it? So a lot of us are interested in building like large-scale multi-tenant With separation in the network, but with open-source software and it sounds like you know There's there's not much we can do without some sort of proprietary solution, what's what what's actually missing from From the open-source Implementations that are limiting us from doing this I'd actually I respectfully disagree so I would say that you know There are a number of vendors out there that have open source They are seeing controllers and they work well at a limited scale. So You know, I mean, I guess if you're operating a very large implementation then you probably would need to Talk to the vendors that have open source those things to say it could I get access to your you know commercial solution that might have a Database of scales better or things like that But certainly if you just want to play a play around with me today, there's a bunch of open source stuff The worst great. Yeah, and I would I guess I would add to that So you I mentioned floodlight are you know our open source controller? You know as a good starting point is not going to be an out-of-the-box solution It's not designed to be that an open v-switch is also a great a great building block And you know with those two you actually have a lot of the building blocks You'd need to create and an open source SDN solution But you know, you know like all open source This is it this is something there would be work on your part to do it You know there's commercial vendors that are offering solutions But the pieces a lot of the pieces are there in open source today That could potentially work at like a very large scale with Open v-switch and floodlight Potentially right. This is you know, you know, so these are the building blocks. We're using in our solution, too So, you know, there's no reason they can't but you said, you know, there's you know, it's not a package solution, right? It would be work on your part to get it there Okay, if not, I'd like to thank the audience for staying with us the whole time And especially all of our panelists for all their knowledge