 Okay, so hello everyone. My name is Jennifer Lynn and I run the product management effort for the contrail team We've got some friends here from juniper as well. So we're gonna leave Probably half an hour. We're gonna do the presentation and then we'll leave some good ample time at the end for for Q&A I see a lot of familiar faces part of the community. So thanks for being here. Hopefully it's a good show so far The the topic that we wanted to cover in this short period that we have today is really about sort of scaling Secure virtual networks and we wanted to share first a little bit about the activity thus far with with contrail and open contrail Recap a little bit about sort of the contrail software architecture approach and then drill into some of the specific use cases to show You know what our customers are trying to do? Leveraging contrail and open stack in their environments So just a few tidbits in terms of contrail by the numbers the first two Boxes there on the left kind of talk about the software development momentum We've now reached or surpassed 750,000 lines of code that we've contributed into the open source community And as many of you know when we made our software generally available We also released all of our software on github under the Apache v2 license Except for the kernel piece which is under the gpl v2 license So with the permissive license obviously our intent is to stimulate the community and get a lot of interaction which I think we've done forwarding throughput there was a customer session yesterday that kind of talked about The you know contrail architecture and showed and validated some of this in terms of you know We're seeing consistently about five percent within Linux bridge performance So a lot of people ask the question is there a penalty for doing software forwarding? And I think so far we've been pretty pleased with the results that we've been seeing The next two in terms of pilots and deployments as well as geographic diversity I think show some of the momentum we've had a lot of interest from various segments And I'll talk about the various segments emerging service provider and traditional enterprise But there's been a big thirst to kind of get this into a pilot and Interproduction deployments so that's been moving very quickly in the eight months since we made this generally available And then you know good geographic diversity a lot of interest in Asia And Europe and we've already moved into production deployments in all of these theaters In terms of open contrail and the download momentum It's been around 400 downloads per month and growing rapidly These are sort of unique open contrail downloads and even hopefully with with this week and the conference will see even more activity We're trying to make it very easy to You know self-serve yourself to to kind of kick the tires And that's also been very important for our technology partners as well to get going And of the website visits we're seeing that roughly 50% are of new visitors So you know once again lots of momentum We've been working really hard within the juniper team to kind of do a lot of meetups We have over 65 events in the last six months a lot of geographical coverage there We've been doing Meetups hands-on demos a lot of sort of open contrail days in in concert with other members of the open stack community The most recent one was in Tokyo and we're seeing a lot of interest there We've also been doing trials as I mentioned across various segments and I noted some of the key ones here a huge amount of interest from our You know juniper's core customers in terms of service providers But they're really looking at Infrastructure as a service as well as network functions virtualization and I'll talk more about that in the enterprise segments The you know financial services customers often want the sort of scale and resiliency of a carrier network And so they've moved you know very quickly into pilots with this type of technology and today many of them use carrier Networks for their managed VPN service, so they see a lot of this is just extending into a hybrid cloud scenario a Lot of interest also from the cable MSO's and in our cloud and hosting type customers But we've been working really hard to make sure that there's sort of this organic Community growing the the meetups have been very good and a lot of interest in just getting you know hands-on exposure From an open-source perspective as I mentioned You know this is primarily licensed under Apache v2 and even from a contrail perspective Most of the software components are based on open-source technologies so as our customers deploy a lot of these open-source technologies in their environments and Generally, we're seeing open-stack being deployed where they're bringing up new clusters of very dynamic applications We also use a lot of these components within the contrail architecture and the rate at which this is evolving is very fast It's changing very quickly the network architectures Especially in the data center where folks are quickly moving away from layer to VLAN segmentation to a flat layer 3 cloths fabric And you know making their data centers look more like the internet or an IP VPN which we'll talk more about later In terms of the dynamic applications, it's primarily the emerging applications that are demanding sort of this cloud infrastructure mobile apps real-time media apps big data analytics You know web tier applications and as we think you know about something like open-stack We're really it's about converging compute storage and network the role that we're playing from a network perspective is really to Simplify the network and open up and expose and abstract the network to higher-level applications So a lot of what we've been doing is really based around how do we create a virtual network level policy and security framework that allows the administrators to really manage by exception instead of managing every every distinct element or every every single flow and really pull this into a real-time feedback loop enabling a lot of the Diagnostics troubleshooting and analytics that the administrators need and not just the administrators from a network perspective now in a devops Scenario we're working with a lot of much smaller companies where this is a converged IT organization Essentially managing compute storage and network as one and as many of you know the networking industry has probably been the most guilty of Managing in a very siloed infrastructure with you know very vendor specific CLI's per device So as we raise the level of abstraction and standards-based rest APIs that are then exposed into higher-level applications This is really the bigger picture that we're trying to get to using a lot of real-time policies to help automate the network Infrastructure in context of the broader cloud infrastructure So how many of you are familiar with the contrail architecture at a high level? Just if you don't mind a quick show of hands, I'm just gonna cover. Okay, so maybe about a quarter of you Then that's that's good a lot of this information is Available on open contrail org getting started. We have our full technical architecture spec, but at a very high level You know this we believe is the most scalable open standards based software overlay architecture to really drive two major components One is network virtualization within a large-scale multi-tenant data center infrastructure and number two service insertion or service chaining of virtualized services a Lot of the development and intellectual property was really around How do we scale the control plane and as many of you know the early discussions around SDN? We're very focused on separating the control and forwarding plane Which if you look at a lot of the large large scale routers that are in our networks today You know from a software perspective. They're already separated, but that notion of pulling a centralized control plane You know into a controller Is is kind of where a lot of the focus has been so that we can drive Distributed forwarding in what we call here as a V router and a lot of the balance in the architecture is to say You need some level of centralization to ease Manageability and make sure you have a system level view for optimization within a very dynamic infrastructure But for scale resiliency and low latency certain components need to be distributed and so that's the balance You know that the V router is a distributed component It works in a hierarchical way with the broader controller Which then exposes a lot of the lower level? Information northbound through rest API's into an orchestration system like like open stack We also support cloud stack and other higher level systems and a lot of the you know syncing State synchronization etc happens through a real-time message bus, which is also standards based based on XMPP From a forwarding plane perspective, you know one of the major goals in order to make sure that we could drive Rapid market adoption was to make sure that we could put this overlay onto an existing physical network infrastructure And so you know in terms of the forwarding plane and the encapsulations that are used We're not religious But we started with ones that are already supported in the layer three gateways and the routers that are out there like Juniper's MX those those platforms are pervasive. They already scale very large-scale networks And once again the control plane that is used is also based on on BGP for Federation So we can capture the benefit of a very mature robust highly scalable and highly pervasive control plane and extend that paradigm into the data center I'm moving fairly quickly. I know but with only half an hour I wanted to make sure we cover just some background here what this allows us to do is essentially move from What's a fair amount of complexity in the physical layer on the bottom where you have now per host you may have you know Many different virtual machines belonging to various virtual networks and abstract that so you're just managing The higher-level entity in this case think of virtual networks as a replacement for a VLAN as Customers move away from static physical segmentation based on VLANs to something. That's much more dynamic and application-driven Allah the virtual networks we can manage virtual networks Which could consist of virtual machines across many different hosts many different clusters and in fact many different data centers So this method of segmentation which is based on sort of native layer 3 IP principles Allows us to scale into something like a hybrid cloud fairly immediately and this is with very little change to the physical Underlay as we call it so from an IP fabric perspective the only requirement that we have of the switching fabric is that is capable of Transporting any to any IP packets and then once we have any to any Transport capability from a physical perspective we can be very flexible with a pure software layer And and create and automate these policies at the virtual network level that also allows us to You know work and embrace the physical layer with the virtualization overlay But also from a service enablement perspective we can work both with physical appliances as well as virtualized appliances And they may be juniper sort of services But you know increasingly we're working a lot with our ecosystem partners To embrace sort of service chaining across third-party services And this is very important in all of the segments that we've been working in So this notion of sort of federation across Various networks creating this network of networks with a unified control plane and that's the you know key thing here is that We can have sort of a series of segmented networks and unify the control plane So essentially we can share reachability information. We can share policies We can manage things at a much higher level if you see on the left here One of the most profitable and fast-growing businesses for our carrier customers has been IP and MPLS VPNs where they can enable essentially tenant-specific networks and empower the tenants to essentially drive their own policies, but it's essentially outsourced over a shared IP fabric virtual private networks and so moving that that concept and this was driven You know out of work in the ITF that one of our control co-founders paid remarks drove with some of our large-scale carrier customers Applying some of those principles in a large-scale data center where the problem that folks were really trying to solve is how do I how do I enable a Highly virtualized multi-tenant, you know segmented environment while still respecting the principles around network access control and enabling services and How do I do that in a way that's open interoperable and standards-based and allows a smooth migration Leveraging the physical network that's already there So as you can see sort of taking a lot of the lessons learned and leveraging that so we could quickly scale environments in in in the data center and across data centers is really the the differentiation we believe we have with with Contrail From an encapsulation perspective, you know, we started Using sort of GRE and UDP. We also support VX LAN and we use VRF labels to identify virtual networks And so we sort of leverage that component of VRS to have a 20-bit label that essentially quickly identifies The virtual networks and respects policies at a higher level of abstraction So the next few slides kind of show some of the key use cases and how our customers in various segments are using the technology To scale their actual business problems and the three segments the first one is, you know, our emerging customers We're seeing phenomenal traction from web scale companies real-time media companies gaming companies and web tier companies And the segmentation that they've had to do and the policies that they need to enforce for instance across three tier web applications The second area is really traditional enterprise and for the last ten years as many of you know Many of the traditional enterprise companies have really been driving sort of TCO within the IT environment for cost reduction and It's been, you know, from their point of view at the expense of Innovation and driving and leveraging a lot of the web tier applications and the big data analytics That they now have to really heavily invest in. As a result a lot of the applications teams do that development in the public cloud Because they have an elastic architecture They can self-provision compute and storage resources and they can get things like database as a service CDN as a service load balancing as a Service without bothering an IT organization So we're seeing a lot of the enterprises building virtual private clouds and moving to this model of IT as a service so that they can On the one hand enforce the security policies and the corporate compliance that they need to do to be a good corporate citizen But at the same time empower their IT users To self-provision what they need to do on demand within minutes without filing trouble tickets, etc and Finally our service provider customers have kind of extended the notion of SDN to network functions virtualization And for them their core business is the network and their ability to roll out network services in a virtualized environment Using the economics of cloud computing and the principles of cloud computing so allowing a lot of their enterprise customers to Self-provision manage firewall manage DDoS manage IPS load balancing CDN etc is critically important to them right now They're seeing a lot of you know over-the-top players start to do this in sort of niche ways But you know we also believe that as this market matures and consolidates You'll see this notion of a service catalog for network services that end customers can self-provision So the first couple I mean we've seen a lot of our sort of emerging customers really Very anxious to within six months roll out a production open-stack environment for infrastructure as a service We've been working very closely with a actually a few European cloud providers but that notion of essentially creating a Infrastructure and platform as a service based on open-stack is definitely already in production using these types of Technologies in this case too a lot of the storage is highly distributed using seph in many cases There are no hypervisors. They're using essentially, you know increasingly Linux containers To enable a lot of the real-time applications that their cloud tenants demand in this particular example one of the Customers of this managed cloud Wanted to make sure that they could support essentially a hundred thousand flows per instance And this is for video and real-time analytics. So that's the type of scaling that folks are expecting They don't want to build their own infrastructure But they need to make sure that it is scalable and resilient in a managed infrastructure as a service scenario The second one is a major SAS provider that we're working with and the first scenario They were building their own virtual private cloud almost from an IT as a service perspective They wanted to enable their various tenant groups in this case application teams building their SAS application to self-provision compute storage network resources without Filing IT trouble tickets and they are building this on an open-stack infrastructure They wanted to make sure that they didn't have to rip and replace the existing Physical network infrastructure and they had you know a lot of sort of physical security appliances as well that they wanted to make sure were Maintain moving forward over time They'll put virtualized firewalls between each of these tenant groups And so a lot of the virtualized services are not a replacement for the physical services But in fact incremental as the granularity of these policies gets You know greater Then when they move into the production environment, they can do that very quickly and this is sort of the DevOps Value proposition as well. They are also opening up parts of their platform to their end customers and partners So that they're essentially being not only a SAS software developer and using virtual private cloud for their agile development Practices but also opening up the infrastructure for their ecosystem partners and essentially being a cloud provider at the same time So they have in this case 10 to 12 data centers You know leveraging a lot of the existing VPN capabilities that they have and over time Evolving that to be a much wider spread elastic architecture that is hybrid because there are certain bursty workloads that they're putting into the public cloud environment this all interoperates obviously using the sort of you know IP based control plane we can Carry a lot of the context from a data model perspective and and things that we've done within contrail to for instance Contribute blueprints for the AWS VPC APIs are the types of things that allow customers to for instance Develop for AWS and then without changing the scripts pull that workload into their private cloud environment Without doing a lot of reconfiguration or rewriting all the scripts that they had etc The next one kind of shows the the requirements of a gaming provider who is also building a Cloud environment for their real-time DevOps environment They had grown so quickly they started out in AWS and they're now growing so quickly They're building out their own private environment based on an open-stack cloud They do a huge amount of real-time analytics for the gamers So as a gamer on this particular case you get a lot of feedback about your statistics and how you compare to other real-time gamers Etc. So they have a highly distributed application infrastructure a lot of the open-source components that we talked about before They're doing Hive and Hadoop a lot of batch processing etc So this is a similar case where the merging customers are very sensitive to latency, especially the gamers. They don't want hypervisors They see that as sort of a performance penalty So this customer in particular is also heavily looking at Linux containers for their virtualized environment The next category is really you know We're seeing a lot of traction from the financial services and in many cases the financial services are building out their private cloud Environment for their own X as a service environment in the case of infrastructure as service Yes, it's pooled resources that that are made available to their internal IT users And they're doing a lot of work in the platform as a service layer to essentially expose that to their Application teams at the end of the day. They also will create a self-service portal to allow You know their their users to self provision and they're gonna turn IT essentially into a you know Not just a cost center. They're gonna build back each of the tenant groups based on cloud usage VPC usage The complexity in the enterprise as opposed to sort of the emerging scenario is that as you can imagine They have a traditional environment where the last ten years They've been focused a lot on server virtualization and data center consolidation And a lot of shrink-wrapped client server applications Moving forward, you know, we're seeing in juniper's broader business to a big shift to a flat Clos fabric switching architecture not sort of the three-tier hierarchical switching layers But more of a flat any-to-any IP fabric that sits within the data center Even within those new data centers. They have sort of a mixed environment They have got to support the existing virtualized workloads that were based heavily on client server and ERP type applications They also have to support a lot of the new services, whether they be physical or virtualized services and they need to have this work across multiple data centers in their private cloud environment At the same time many of the enterprises want to enable their IT users to go to the public cloud when appropriate So in this particular case one of the financial companies had five percent of their workloads in AWS And they need to essentially see that as an extension of their private infrastructure They need to be able to support, you know, network and security policies across this heterogeneous environment from a networking perspective because we can enable a unified control plane across all of this We can federate those domains and make it actually work as one sort of Environment one virtualized or physical environment This is a little bit of a busy slide But kind of shows a next level down in terms of the Heterogeneous environment that we're seeing a lot of in an existing enterprise on the top What we want is essentially the ability to federate and interconnect these sort of Domains on the bottom what you see is the physical reality of kind of what needs to happen From a switching and routing infrastructure many of the vendors are investing in supporting new encapsulation types like VXLan Already in something like, you know an MX or the layer 3 gateways that are out there We can support the control plane and support things like VRFs and BGP to manage at the virtual network level And that is not just for ease of manageability and for policy enforcement, but also for scaling In the in the top of rack switches in the gateways The ability to support both, you know bare metal VLAN termination as well as bridge from VLANs to VXLans and Later this year. We'll see a lot of the top of rack switching vendors natively support VXLan and tying this into virtual networks in the overlay So there's there's sort of several levels here We're also seeing a lot of customers as I mentioned in the emerging space Want to enable a bare metal scenario using essentially the V router without a hypervisor And so leveraging things like Linux containers and Docker as well as Linux namespaces We can sort of enable that environment as if just the same way as if we do today with a KVM or Zen hypervisor One thing to also mention here is that you know from a forwarding plane performance We support, you know the IP VPNs as well as now the new standard for pure layer to transport from a forwarding perspective with EVP ends So today we already support natively EVP N for DCI data center interconnect on the on the mx's and later this year You'll see native EVP N and VXLan support on the on the switches and and the routers as well so that all comes together sort of as a single infrastructure across physical underlay and software overlay The other major benefit of doing the L3 VPN based approach here is that we can Essentially create virtual machines across two data centers and keep them essentially Still within a virtual network so you can see here on top logically We have a number of virtual machines that sit across two data centers By using sort of the notion of VRFs in the layer three gateways We can support virtual networks that span across multiple data centers And this is something you know today and in other other overlay models once you exit a proprietary data center All bets are off in this case because essentially the model that we use within the data center is the same as across a wide area We can span across data centers with a single network single virtual network. It's just a logical segmentation The last category is really you know our service provider customers And this is based on some debt data from Info netx just a couple of months ago A lot of our carrier customers are really looking at scaling these virtualized services Across their their core business this notion of telco cloud is really exciting to a lot of our carrier customers They want to enable both enterprise services for their managed services customers as well as use concepts like NFV to You know evolve their own telco cloud infrastructure And this is just a you know a recent ranking that you can see at the top There are strong interest in sort of rethinking the access layer and the provider edge So this notion of virtualized customer premise and the notion of the virtualized provider edge Beyond contrail. I think we're seeing a lot of investment and sort of evolving to the flexibility of a software defined model Service chaining obviously continues to be fundamental and is a category in and of itself Sort of virtualizing the IMS core and investing in areas like you know virtualized evolved packet core These are the things that if if these environments don't scale You know, we'll never get off the ground in terms of these models of network virtualization And that's why it's particularly key for our carrier customers that we start with a standards-based open interoperable Control plane architecture that leverages the benefits of the last 20 years of IP networking CDN is kind of being rethought also with you know, not just outside in but inside out load balancing And how do we do this in a very real-time way? And there's this whole category of cloud brokers that are now playing With lots of data across many different clouds to optimize content delivery This is just an example and you know We have some videos out there on YouTube showing some of the contrail use cases where we've been working with carrier customers to create Sort of a service catalog of managed services as I mentioned before You know a lot of the firewall IPS DDoS SSL services have already been virtualized and we make those Sort of available in a self-provision sort of service catalog You know load balancing and CDN also tend to bubble to the top and this is really how a lot of our carrier customers are Using this type of architecture to find incremental growth and grow their businesses faster Today as you know, it takes Probably six to eight weeks to provision a new service and that's after nine months of certification testing So the goal of many of our customers is to get to an agile software development model where they can roll out new revenue Generating services much much faster. So it's the DevOps equivalent for the telco cloud and then you know For their core business too as you know carriers play across seven seven domains from you know The SMB sort of premise or consumer all the way through the access layer into essentially the metro fabric and the core And so they're really thinking how do we simplify? What's in the physical elements themselves and push a lot of this into a system-level optimization with a cloud-based controller That essentially you know enforces a lot of these policies So we've been doing a lot of work, you know in concert with our you know with broader juniper around how we can kind of Make this vision of telco cloud real a lot of it whether it's vcpe or When optimization across sort of the broader when infrastructure, you know juniper acquired a company called Wando Which does a lot of the path computation and when optimization? Similarly, how do we essentially take a broader system-level view do the optimization within a controller and then push the enforcement down to individual devices Finally a lot of the ability to scale comes with do we enable the right level of monitoring troubleshooting and diagnostics? And there's been a lot of you know emphasis in the contrail architecture around how do we build a scale out Infrastructure with the real-time bi-directional message bus so we can capture all the information that's needed either from an operational perspective or turning sort of configuration rules into actual Lower-level configurations on the elements and finally, you know a lot of effort in sort of a scale out analytics infrastructure Has a you know full-scale map reduce infrastructure so we can do Hadoop jobs on the infrastructure in real-time and Also enable that a lot of this granular flow level information is easily queryable In real-time, so I'll show you some some screenshots There the first one is really to show that at a high level We have a very large-scale infrastructure We need a dashboard that essentially summarizes the individual component so in this case You know the V routers the control nodes the analytics nodes You can look at the big picture and within that there may be a thousand or more instances at each level we're able to sort of drill down and Essentially see what's going on at a lower level, but it all bubbles up into sort of a higher-level dashboard So we can drill down very quickly into individual traffic statistics per virtual machine Aggregated at the virtual network level we can drill down into specific instances and look at Flow queries specifically, let's say for a specific port So show me just sort of port 3306 traffic in a time series fashion so that I can look at trends, etc We can also do things like you know create a packet capture between two virtual networks in this case We can say between virtual network a and virtual network be Capture all all traffic and take it into wire shark. That's what you see on the top there So that ability to do the analytics and query specific Characteristics is really part of the operational scaling that's very important to our customers And with that sort of granularity and visibility comes the ability to manage on the business side So working with partners in this case like like scalar who are doing sort of cloud management and cost per project or cost per Tenant enabling you know pulling in Purely rest API information all the data that you saw there is exposed through our API server So anything in the in the UI essentially comes directly off the API server And a lot of our partners are able then to pull it into some of their business analytics, etc So we've been working, you know very closely with our ecosystem partners Obviously no one vendor is going to do this on their own So the ability to kind of have open interoperable interfaces across the way We've been working with a lot of the you know open-stack integrators And we announced recently a partnership with with Nokia for the EPC and voice over LTE Infrastructure using their liquid core so pretty pretty diverse and this is just a very partial list of the partners A lot of the network services partners either, you know CDN or load balancing We didn't include here These are really the system level partners that we've been working with to enable the various use cases that we talked about before So this is the last slide. I just wanted to do a quick plug Today we didn't get we didn't have enough time to kind of go through, you know really hands-on But that's teed up for Thursday, so we have a session 9 to 10 30 and that'll be hands-on session with with contrail and we'll throw out a little bit of a Open contrail challenge at that event as well. So you have the opportunity to win an iPad And that will sort of take take the next step in terms of you know looking at individual functions within within contrail So I know that was a lot in a short period of time. I think we have enough time for for some questions or comments I've got my friends here from from the contrail team as well Any thoughts or feedback or questions that you have? Yeah? Hi. I don't know if this works um You initially didn't support VX LAN and it looks like you have it now When did that come online and are there any limitations to using VX LAN because it looks like MPLS over GRE? You might have some capabilities or service chaining or something And then also how does the VRF 20-bit label play into the VX LAN 24-bit? Key is that a separate layer of the Federation layer. So VX LAN and and by the way ESXi hypervisor support are relatively recent. So we didn't have it when we first launched in the September time frame We see VX LAN as another encapsulation Yeah, that that as I mentioned we started with GRE and UDP because that's pervasive But we you know had customers that wanted to support a VX LAN environment So we've essentially support that frame that framework as well in terms of encapsulation type What it allows us to do is take sort of a VX LAN environment and map it via the VRFs into virtual networks And that's part of the sort of you know L3 gateway scenario that we're showing There's a lot more detail there that we're gonna put out there But because of this sort of mixed environment We wanted to make sure that our our customers could bridge from an existing VX LAN environment and map that Segment into the virtual network policies. So it still uses VRFs as far as you know the virtual network identifiers And as I mentioned originally we supported KVM and Xen hypervisors and now with our 1.1 release Which will be out in just a few weeks We'll also support ESXi as a hypervisor primarily for those customers who are let's say in an ESXi environment today They've maybe qualified those legacy workloads on ESXi and they want to pull that workload into an open-stack cloud That tends to be the first use case that we've seen and so we've been working with some large-scale enterprise customers on that use case I'm here to sorry Yeah, go ahead. I Wanted to check with you on your role and Contribution towards open stack There is something called designate Which I heard is a new initiative from Red Hat and that relates to DCI data center interconnect And so is there any contribution you're planning are and also another question is about IPAM how you are managing in your environment? Yeah, so on designate. I don't know if one of the other folks In the control team wants to comment there. I think it's probably early days We probably don't have we are working with you know Red Hat as a partner more broadly You know to date our roadmap has been pretty full in terms of you know what we've had to deliver So we haven't committed anything to support designate But definitely something that we should we should look at if if there's a market need our customers Haven't asked to date for that tie-in, but you know We've been working very closely on a number of open-stack projects with our partners and What about the IPAM so IPAM? I think you know we we have sort of IPAM capabilities directly in the v-router today And we can also work with sort of existing IPAM You know the ability to kind of do the sort of IP address management right at the v-router is one of the key strengths And you know it's something that we've been you know differentiating on with the v-router approach Which you can't do for instance with Linux bridge or v-switch natively does it support IPv6 We actually so the release coming up. We support IPv6 in the overlay So within a quarter we'll have IPv6 overlay support There's some work to do in terms of you know IPv6 obviously with the broader underlay But that's sort of an ongoing investment that we're making so we have several customers who want to see that this year And we will have that this year in the overlay. Thank you. Sure Yeah, go ahead actually if you don't mind going to the mic Sorry, can I can I get him in yeah, thanks So can you briefly compare a country with the open daylight for example by Usability the feature list you know as well as a maturity. Thank you. Okay, so thanks I meant to mention that Juniper is a platinum member of the open daylight foundation as well as a gold member of open stack We believe in sort of where there is sort of open interoperability in the ecosystem. We want to play I also actually sit on the board representing Juniper for open daylight And we had just recently announced that we will enable the southbound plug-in to contrail from open daylight So we're seeing a lot of customers wanting to have the flexibility and not necessarily make a binding decision right now So we're doing you know our part to Federate and interoperate with those various environments and as open daylight evolves You know, there are now a number of sort of overlays that are also participating in open daylight It started initially with not as much focus on on the overlay model But I think as this evolves we're seeing a lot of traction even within the context of open daylight around Overlays so we you know 30 vendors in that organization. We're interacting with a lot of the same vendors that we are you know in the neutron Discussions in the forums of open daylight and that will I think continue to evolve. It's still early days Yes What was thinking behind using BGP for the control other than the commercial reasons as opposed to X for the control plane Yeah, okay, so well, I think one of the major things is that you know It's open standards base highly pervasive across the IP networks today proven to scale Horizontally scale and have high levels of resilience Those are you know primary reasons It's you know, it's the balance between does it fit the need and do we have to wait another five years before this? Technology is pervasive. Maybe I'll open it up to anyone else who wants to chime in in terms of why YBGP You know, okay, I think those are some fairly good reasons But in the standards bodies there was a lot of discussion about does it fit the need and that's obviously a very active Discussion, you know if it meets the need leverage an existing Technology so that we don't have to rip and replace and we can get going with sort of these new use cases Yes, how easy it is to port application between Open contrail and open daylight So that's why we've enabled and you know We've invested specifically because we had a lot of you know Our customers that were sort of kicking the tires on a number of options and we didn't want to pose open contrail You know as a direct sort of separate competitor or just a distinct Alternative however the approaches are you know somewhat different so today in terms of the southbound plug-in the interfaces that are there are still Still relatively simple in terms of L2 interfaces similar thing with neutron when neutron discussion started It was essentially VLAN based and now we're seeing you know much more of a discussion about how to support native layer 3 models and do do it at sort of a lower level of granularity and with or without Hypervisors, so I think we see a similar transition and evolution in open daylight But for what folks are doing with open daylight today. There's no porting So what you can do in open daylight via the southbound plug-ins that we're enabling you can meet that use case within you know from within open contrail So northbound a lot of the early discussions in open daylight when it kicked off several months ago is what is the northbound Abstraction for the network and now I think there's been a lot of cross pollination between open stack and daylight certainly probably more than when Open daylight started, you know the neutron API's which contrail also is you know committed to supporting are also what's being exposed as the Northbound API's from within open daylight, so there's sort of a you know a good leverage between the two of them And that's where you know it's very easy for us to actively participate there Any other comments from anyone Questions, okay, so we very much encourage you to join us in the journey moving forward We've tried to be very open and transparent with the information that we put out there But please keep sort of an open dialogue. There's various mailing lists and archives ways to get in touch with us And thank you for your support