 Ready to go? Excellent. Good afternoon, everybody. Just before we get started, I was told to remind you, this session is scheduled for 40 minutes. We're going to talk for maybe 25, 30 minutes, and then we'll open up for questions. So if you're interested in asking a question, please try to come to the microphone. There are two microphones over there because this is being recorded. So it's good for people to hear the question before they hear the answer. But anyway, let's get started. So I am Ali Kaffel. I'm from Red Hat. And I'm here with my colleague, DP, from Juniper Networks. I'll just maybe just introduce myself quickly and have DP do the same. So I've been with Red Hat for about a year. And I'm focused in the networking group for Red Hat. I've been in technical marketing and business development. And I work very closely with my colleagues from Juniper. So DP, maybe you can make an introduction. Sure, thanks, Ali. I'm DP Aidevara. I'm a director in Juniper product management, focused on contrail networking and contrail related product portfolio. Excellent. So the agenda today, as you can see, we'll go very quickly through, I suspect, a lot of people here know Red Hat and Juniper. But just in the context of this presentation, we'll just give you a quick overview. Then we're going to talk about the differences between NIV and SDN, at least from our perspective. Then we'll go into some of the challenges in, obviously, I assume, since we're all at the OpenStack Summit, people know OpenStack very well. And Neutron within OpenStack works really well in many networks. But there are some networks where it needs to be augmented. So we'll talk about some of the limitations in standard Neutron, and then how you augment Neutron with an SDN controller, like Juniper contrail, to turn on advanced services. And then we'll go into a use case. We have actually a Tier 1 telco. When we did this presentation, we hadn't gotten the OK to use the name, but this telco is actually orange in France. So we'll talk about some of the lessons we learned. That was a solution we did jointly, Juniper and Red Hat. So that's the agenda. So just very quickly, Red Hat, I think most people here probably know who Red Hat is. What you may not know is that we are very engaged in the telco space and the networking space. We actually have essentially all of the S&P 500 telcos use Red Hat, banks, and so on. And we are, as most people know, we're a software infrastructure company. And we're the leading contributor across all of the key open source community projects, obviously Linux, KVM. But also OpenStack and OPNV. OPNV, as some of you may know, it's sort of the reference architecture for building a telco-grade NFV platform. So that's sort of a quick overview of Red Hat. I'll have my colleague talk about Juniper. So yeah, to give a brief overview of Juniper, how many of you are not familiar with who Juniper is or what Juniper does? OK, that's pretty good. So I can skip through the slide very fast then. So some other key things we want to point out is Juniper is known for its innovation. And one of the key things we did is about four years back, we acquired this company called Contrail. And Contrail is the startup that came into Juniper. And Juniper has held OpenContrail, which is the product from Contrail, flourish and be successful in both telcos, enterprises, and cloud services. And a couple of key things I want to highlight is we are known for our telco presence. But what you may find interesting to see here is we are in the top 10 social media properties, four of the top five largest search engine companies without naming them explicitly. And we are in almost all the financial services, major financial services companies, as well as exchanges. So the gist is basically Juniper has been pretty successful in terms of both telco as a segment, enterprises, SaaS, cloud service providers, as well as financial services. Click the right. All right. So just how many of you here have heard of OpenContrail? OK, that's pretty good. So just to kind of give you a brief overview of what OpenContrail is, right? OpenContrail is an open source based cloud networking initiative, right? And we use Apache V2 license, so that's a standard today. And one of the key highlights of OpenContrail is that it's completely based on, it's a standards-based implementation. So we use BGP, and then we use XMPP. And everything, our entire implementation is standards-based. And what that gives our users or our customers is basically no need for a vendor lock-in, right? I mean, being open source doesn't necessarily mean that you're out of the vendor lock-in mode. But in our case, it's a completely standards-based solution, and leveraging a lot of the protocols that exist today, right? And their extensions. So that kind of gives you the interoperability and multi-vendor kind of scenarios that we can support with OpenContrail. We are not limited to, let's say, virtual machine providing connectivity for VMs, be it OpenStack and Vendor vCenter. We are also extending the same capabilities to the container world. We have Kubernetes-based integration, Mesos, OpenShift, and so on. And we recently, just yesterday, we published a blog on our upcoming release and what capabilities it provides. So I encourage you to go to OpenContrail and to jennipa.net to read the latest on that. One of the key things about OpenContrail is it was developed as a cloud-native networking solution. And what that means is a lot of emphasis on API. Everything needs to be API-driven, and that's kind of the philosophy we have today. And we can easily integrate Northbound with any orchestration system. Like I said, be it OpenStack, be it Kubernetes, be it vCenter, any of them, right? And one of the key advantages we have, something that we developed the product into, something that helped us mature as a product or as a solution through our deployments, is basically we are known for our scalability. We are known for our high performance. And we make your, in the telco environments, we make your network connectivity and networking services carry a grade, right? So those are some of the key highlights of what OpenContrail does. Probably should pick up the speed a little bit. Sure. So I think everybody that knows Red Hat knows that this upstream first from communities to the enterprise is a big philosophy and the mantra of Red Hat. So there are over a million open source projects out there. Red Hat is involved in about 1,500 different projects. And from those projects, these are just some of the projects that we're involved in. From those projects is where we actually create products. So 100% of our products are actually open source based on community project that most of them we're involved with and make a lot of contribution to. And we actually pull different open source project to create products. And the products are what we support. And one of the nice things about Juniper Contrail is that they follow the same philosophy with the OpenContrail concept. And that's one of the things I really like in addition to the different additional unique features that DP mentioned briefly. And he's going to talk more about. So let's get started here. So NFV versus SDN. Before we talk about NFV versus SDN, I just want to put it in the context of this digital economy that we're all talking about, digital transformation. How does NFV and SDN fit into this digital economy? I think we all know today that one of the biggest challenges that the telcos and the enterprises have is this data and digital disruption that's causing hyper growth of traffic. So the telcos, for example, experience in a huge amount of traffic growth, flat revenue, which means that their budget is flat. But in order to really keep up with that growth and revenue, they have to find different ways of really supporting that huge bandwidth. Just to give you some perspective, it's one thing to say there's hyper growth in data traffic. But just to put it in context, one of the telcos that has done a very good job in sharing a lot of information is AT&T. AT&T talks about from 2007 to 2015, they had 150,000% growth in traffic in their network. That's just unbelievable, right? 60% of their traffic is video, right? And when you look at that traffic, they say every day they have 114 petabyte of data traffic that goes through their network. 114 petabyte may not sound like something impressive. But when you think of it in terms of it's 130 million hours of video traffic in one day, that's just unbelievable. And video is 60% of their traffic. And just to give you context, the average video, one minute of video is about four megabyte of traffic. When we talk about virtual reality, you talk in one minute of virtual reality could be over 100 million megabyte of data. So it just shows you the growth of traffic that is going through the network right now and what's coming. So therefore, we need to really build networks differently. The other big challenge is this global competition that's coming from the new players. I have over there the fangs and the bats. How many people have heard those terms? Fangs and bats? Anybody in the room? So fang is really a Wall Street term for Facebook, Amazon, Netflix, and Google. It's a term that really describes these hyper growth companies, like those companies. And then bat is really Baidu, Alibaba, and Tencent. So it's the Chinese equivalent of the fangs in the US. And what these companies are doing, as we all know, we all experience them every day, is just the growth that they're going through the network. And what this means is there's a new reality. The new reality is that in order to really address digital transformation, we have to turn to cloud and software define everything. And that's really what we're talking about here. So when we talk about digital transformation, as techies, we tend to talk a lot about technology. Technology is a big part. In this session, we're talking about cloud and NFV. But there's also lots of other technology that's really driving the trends in the industry, as we all know, and many of us talk about this every day. But the other thing that's interesting is really the people and the processes. If you just adopt technology to really embrace digital transformation, and you don't have the processes in place, what we call embrace the open source culture. An open source is more than just about product, even though we tend to talk a lot about product. It's having an organization that thinks the way the open source community thinks and behaves the way the open source community behaves. That's really essential. And then obviously having the leadership that supports that. And if you do all those things right, you really achieve three key drivers. And the most important of all is speed, to innovate faster. And the second one is really efficiency, improved efficiency, which we tend to talk a lot about. Because when you have this big growth in traffic, you want to have a network that can really support that. And then finally, it's about simplicity. If you do those things right, then you have the results. So when you put it in that context of NFV and SDN, and it's the same thing. It's about speed, faster, deploying services faster. And just to give you an example, this is a well-documented case. Amazon is talked about on an average, every 11 seconds. They deploy an average of 10,000 to 30,000 servers at a time. Using CDCI, which people in this room know very well. But if you look at the traditional telco and enterprise, that's six to seven month. That's huge. If you look at operations, Amazon has one person for 15,000 servers. Traditional corporation has less than one person supports less than 100 servers. And then the complexity is another one. Amazon has, sorry, Google has about 10 configs. The typical telco has over 1,000. So it's a lot more complex. So really, it goes back to what I was saying before. It's not just the technology, it's also the processes and the people. So this comes back to, well, how is NFV gonna help us achieve that? So NFV, as many people in this room may know, it's really a concept of virtualizing multi-vendor, purpose-built hardware into software network functions, virtualized network functions on commodity hardware with fast elasticity, scalability, and dynamic resource allocation, which is really cloud computing. It doesn't really talk a whole lot about the bottom layers, which is really the SDN part, although they're very complementary. SDN, on the other hand, as many people know, it is about really separating or centralizing the control on commodity-based hardware and making it easily programmable and automated. So it's layer one, two, three. So SDN helps NFV, but it's not really required, although it does help NFV significantly. So they tend to complement each other. So the use case here, yeah. Okay, so a couple of things, just to lighten up that atmosphere. When Ali mentioned about fangs and bats, how many of you thought of Dracula? No? Okay, see, that's good. So if telcos are facing threat from fangs and bats, we'll help make the telcos the Dracula, anyway, on a more serious note. From a contrail perspective, right? So Ali mentioned about AT&T as well, and contrail is being used in AT&T deployments. One of the key things about contrail is, you're not just limited to, let's say, the data center, but contrail as an SDN solution is applicable in the telcopops as well, and that's where we'll see a lot of NFV use cases coming through, right? Now, the way contrail as an SDN is architected, we tend to be very infrastructure agnostic. What that means is whether you have, like we'll see in today's use case, which is the pointer? That light? To do a light? Yeah. Okay, sorry. So in today's use case, we'll talk about branch offices, right? So whether your compute environment is in branch office, whether it's a data center, telcopop, or be it public cloud, composed of different virtualization technologies or bare metal, et cetera, the objective is to provide connectivity, right? That's not the only objective, right? Contrail as an SDN solution definitely provides connectivity in this diverse environment. We also believe in being able to provide you with a multi-tenanted secure environment, and so security and policy is a very important part of what contrail tries to address. And lastly, with a diverse environment, you can't have like five different panes of glasses to monitor or to manage or to, you know, kind of, did have information out of, one for every type of, you know, your cloud environment. So the idea is to bring this all together as a single manageability platform where you can, you know, monitor, you can look at your analytics and at the same time integrate seamlessly with the orchestration systems, right? When we look at the use cases, I mean, today OpenContrail has deployed a variety of use cases which cover, I know at the keynotes, we've said private clouds are a kind of a taboo word, but you can call it whatever you want, but it's about connectivity in a legacy, VMware interconnect or be it VMs and containers, actually contrail now has the capability of, you know, being able to provide connectivity in nested environments where you're running containerized clusters on top of OpenStack, for example, you can use a single contrail control plane to be able to provide connectivity. If you look at public clouds like SaaS providers like Workday, et cetera, contrail provides a multi-tenanted environment where you can enforce enough security and policy so that your multiple tenants that are using your SaaS services are completely compartmentalized. And in the case of Telco Cloud, we have deployed in almost every tier one Telco, primarily for NFV use cases, leveraging a lot of our rich service chaining capabilities. And we are also addressing IoT and use cases like connected cars and so on. And obviously, today we'll talk a little bit about our SDVAN deployment with Orange. Okay. Okay. So what we've talked about so far is really how NFV and SDN fits in the whole need for additional transformation. Essentially, speed, efficiency, and speed, efficiency, and simplicity. Those are really the key points, the key theme of NFV and SDN. So now let's talk about in OpenStack what Neutron does and how Neutron can be augmented. I think we all know that Neutron is the networking piece of OpenStack and it does a fairly good job for most networks. It has an open V switch in it, has layer two connectivity, the whole IP address management. It has routing functions through the service plugin concept of layer three routing, load balancing VPN and firewall. And it does a fairly good job, but when you put it in a very large Telco network, there are some limitations that need to be augmented. For example, it may force most Telcos that need service chaining the ability to interconnect different services together. That's lacking. Dynamic routing is lacking. The chaining physical and virtual network functions, distributed source network address translation. It does support source network address translation, but it's not distributed. It does it in a centralized manner, which really creates some of the additional noise which adds into performance challenges. It doesn't have real time and historic analytics. And when you look at most networks, the skillability performance and availability requirements, Neutron needs to be augmented. And for many networks, it's good enough. It does a great job, but for certain networks, very large Telco networks, it needs help. And that's really where SDN comes in. With that, I'm going to have DP talk about how Neutron is augmented with SDN. Sure. Thanks, Ali. So this is a typical Etsy manual architecture if you're familiar with the framework. The place where Jennifer Contrail is working closely with Red Hat is basically providing the SDN element which spans across basically the VIM or the virtual network component in this NFV stack solution. So that's basically one of our main focus. And in the containers then when we are integrated with OpenShift as well, where didn't we bring in a lot of the scalability, high availability and performance aspects into a containerized environment? I mean, if you look at Kubernetes and Kubernetes is still evolving, it's not mature enough, but using Contrail in a Kubernetes environment, you can still leverage all the mature capabilities that are expected out of any well-deployed networking solution. So how do we augment Neutron, right? The first thing is about scale. So like Ali was mentioning AT&T, AT&T status takes about, you know, I think 140 petabytes and so on. And we have proven that we can actually be deployed in a scaled environment. So be it your SaaS cloud, be it telcos, whatever be the use case, scalability is something that's a bread and butter to Contrail. How do we do that? We use protocols like BGP which is actually running the internet, right? So we use all the right standards and we have architected it in a very, in a way that, you know, we are easily scalable. Secondly, performance. In comparison to OBS, I mean, definitely Weed Outer, which is the forwarding element of the Contrail solution is a high-performance solution and we can run Weed Outer in DPDK mode or you can actually run Weed Outer on a NIC and that's called a SmartNIC or SmartIO kind of solution. There was a press release around it in Mobile World Congress as well. So when it comes to performance, that's definitely significantly better than OBS. At the same time, there are a lot of different performance characteristics based on your requirements. You can either go with Weed Outer DPDK or you can go with SmartNIC based deployments as well. But Weed Outer by default runs in Kernel mode. So, you know, it leverages the fact that it's part of the Kernel to reduce any inefficiencies. When it comes to high availability, Contrail as a solution is designed for high availability. Control plane, we have three controllers in a cluster, a highly available cluster. And we also do in terms of, when it comes to life cycle management of Contrail, you know, we can do in-service upgrades, we can do a lot of different ways of upgrading without any disruption to your service, depending on, you know, what you're comfortable with. Lastly, this is something that's very unique or this is where a lot of, this is one of the key reasons why we have found success across the board is our networking features, right? I mean, at a high level, our objective is to provide secure multi-tenancy, right? But when it comes to feature set, it goes beyond that. Like service chaining, it's not just basic service insertion and service chaining. It's about doing end-to-end health monitoring of your service chains, being able to provide load balancing across service chains and so on. We support all services like LBAS and firewall as a service and so on. And when it comes to an overlay-based SDN solution, what customers typically look for is, okay, how do I relate what is happening in the overlay with my underlay fabric? So Contra comes with an analytics solution where we provide you this whole visible, in a visual way, an ability to relate what's how a particular flow is traversing in your underlay, though our SDN is an overlay-based solution. So pretty good underlay visibility and so on. And by partnering with, I mean, we acquired App FormX into Gennipur and a whole correlation of underlay overlay, visibility monitoring is only getting more and more feature-rich. Okay, so from a Contra perspective, we tend to talk about the capabilities in these 10 buckets. In the interest of time, I'm not going to go into each of these buckets, but you can definitely go to opencontra.org and take a look at the capabilities in each of these kind of feature areas, right? A couple of key things I would like to point out is I'm going to start backwards. When it comes to API services, like I said, we have a REST-based API in Northbound. We also have plug-in-based integration with OpenStack, CNI-based integration, Kubernetes environments, and so on. In all the leverages, the API services we provide. High availability upgrades I've already talked about, how we support high availability and kind of upgrades through ISSU. Service chaining is one of the key things that telcos look for. And as you can see, we can support physical network functions, virtual network functions, do health check and be able to do service chaining based on policies, okay? We should pick up this piece. Yeah. So Juniper and Red Hat have jointly deployed in a lot of different environments. One of the key reasons being Red Hat is the leader for open source and has been one of the top contributors to OpenStack as well. And from an SDN perspective, Juniper's OpenContrail has been rated as the number one commercially deployed SDN solution in OpenStack environments. And this is the latest survey for this particular OpenStack Summit, for the OpenStack Summit Boston, but this is not the first time we have been doing this over four years now. So it's not to brag about it, but to talk about how users also rate as well and how we are learning a lot from our users' scenarios and incorporating or how their requirements are driving our roadmap. Okay, so we called it TO and Telco case study because we wanted to check with Orange if we can use their name, but it's a public information. You can look up Orange and Red Hat and Contrail and you'll find enough information on this. One of the key things we did in this Telco case study is basically Orange wanted to, Ali mentioned about flat revenues for Telcos and while the traffic growth is exponential, one of the key things we want to enable as their partners is to help them find new revenue streams so that their revenue doesn't stay flat. So one of the key things Orange wanted to do was be able to provide connectivity to enterprises as a network, as a service at a much lower cost, but more importantly be able to enable features and so on in a very agile manner. So the idea was to reduce time to add new services by moving to virtualized infrastructure on top of their existing networks and to give enterprise customers the speed, simplicity, and agility that they look for. Yeah, so this network as a service solution is primarily about reducing truck rolls, delivering it quickly, getting rid of static provisioning and hardware being able to orchestrate these services, going from a complex kind of infrastructure to very simple and easy to manage infrastructure. These were kind of the asks that they had and they were looking for a solution that would get them there. So the solution that we ultimately came up with is basically this where there's a customer portal where if I am, let's say, an enterprise and I want to start getting, let's say, some kind of a VPN service, I get on this customer portal, I define what exactly I want, what kind of services I want, I want firewall, et cetera, other capabilities and using an orchestrator, we are able, so first step is where Orange will ship box to the enterprise branch, but then you plug it in and it'll come up automatically. It'll be provisioned with the orchestrator, provision the branch device for services, et cetera, based on what you have ordered and at the same time, to deliver on those services, they use the SDN controller, which is OpenContrail to be able to service chain. Now, OpenStack, Red Hat's OpenStack and Juniper Contrail is deployed in the Telco Pop environment where the service chaining is being done and the VNFs that are being deployed by Orange are basically, there's a VNF manager to be able to do life cycle management of the VNFs and so on, but at the end of the day, when you look at this whole picture, what we need to remember is, the service that we are enabling is, from a branch device, be able to service chain through these virtual network functions which exist in the Telco world, Telco Cloud, and send it out on the internet, so this is like the basic representation of the use case that they were driving it. TP, I wonder if we should go to Lessons Learn because we're running out of time and then we can... Okay, we can skip the timeline and just go to... Whoops, yeah, Lessons Learn. So, one of the key things, there are multiple components in an NFE architecture and when you want to deploy a service like VCPE or SDVAN, there are a lot of different components and it's more about an ecosystem-based solution. So what is important is, having open source, right? Automatically drives standards-based implementations and interworking across multiple vendors. Standard Neutron OBS is very functional but in Telco environments, to be able to support a lot more branch end points and so on, a lot more enterprise customers, you need something that can scale well and there you need some kind of augmentation using commercial SDN solutions like Contrail. And like Ali mentioned earlier, when it comes to... It's not all about the technology and the components, it's about the people and the organization and the process that's also critical. And lastly, when it comes to Telcos, typically I think the planning to deployment cycle is a little long and complex but with virtualized environment and a fully virtualized solution, you can actually reduce this time to from planning to deployment pretty significantly. Ali, you want to add anything? Yeah, no, I was gonna say and really, this goes back to what I said in the beginning. It's not, technology is critical, very important but the processes and the people in place are really critical and this is really talking about the processes. So I think with that, we're done, we'll open up for questions. I think we have about seven minutes or so for questions. Again, as a reminder, if anybody has any question, please come to the mic and ask your question and we'll try to answer your question. Any questions? First of all, thank you for the talk. Scalability is certainly a good thing but could you share us some numbers please? Can I stretch open control across a thousand of hypervisors or tens of thousands? How many virtual machines can I add that? Sure, so I can't talk specifics of this particular use case but from open control capability. We at Juniper, we have actually qualified more than let's say 2,000 compute environment. So I'm talking about the compute nodes, not the VMs, so you can have multiple VMs on each compute node. So pre-scalable, we have done more than, I can share probably some of the scalability numbers in person as well but we're talking about thousands of nodes when it comes to control scalability. The other thing to remember is you can have multiple clusters and through control you'll be able to federate across the clusters as well and BGP gives you that capability, yeah? Okay, thank you. Any other question? Either we did such a great job explaining everything or people are just tired, they've had a long day. If you have any question, please go to the mic and ask or you can just call it out and I'll repeat it. So for any specifics or any more in-depth information you would want, you can come to either Red Hat Booth or Juniper Booth, we are in the marketplace and we have a lot more teammates out there who can actually dwell deeper into how exactly did we do end-to-end monitoring for service chaining or the nature of our integration with Red Hat OpenStack and joined deployment using OSP director and so on. Yeah, and I was gonna say really the relationship between Red Hat and Juniper is being very positive. I think that's one of the key ingredients in any relationship and obviously with the customer as well. You know, we try to define a lot of processes without being too overly process heavy so that all these projects are managed properly and that's very critical and we talked about the cross-training in the last slide and the project management, those are also critical points. So, and we'll be here for a few more minutes if people don't want to ask questions publicly and have any questions, either come talk to us or come to our booths as DP said. Thank you. Thank you.