 Hello? Wow, I'm so happy to see so many of you here. So first of all, who knows who this person is? Who can tell me? Oh, this doesn't work, I guess. No! Clint Eastwood! So, so for those of you who are not from US, so in the 60s, there was a movie called The Good and Bad and the Ugly starring Clint Eastwood. It was a very popular movie and then after we're after that people start using the phrase The Good and Bad and Ugly to describe the truth of a specific subject. So that's what we're gonna do today. We're gonna talk about the truth of how telcos experience is using OpenStack. And my name is Annie Lai. I'm with Huawei IT product line. And I've been with Huawei for almost four years. For the last three, four years I've traveled to over 30 countries talking to telcos about their cloud initiatives, especially OpenStack. And today I'd like to share you the learnings and the findings I got from working with them on their cloud initiatives and their OpenStack initiatives. And I just want to let you know it's an overview kind of presentations. So it's more about, you know, how telcos use OpenStack. If you're looking for more technical details we definitely have a lot of, you know, experts. We have architects, we have senior engineers here. We can talk to you more about technical details later. But due to the limited amount of time we can only talk about very high level. So first of all, for those of you who are not very familiar with telco, just very quickly want to talk about telco's current situation. Telcos are going through a very interesting time. And because their competitive landscape is changing. They are no longer competing against other telcos. They are also competing against OTT guys. And even though there's more and more traffic, more subscribers, more users, but their margin is shrinking. And they have to find a way to change that. And they also found out that their OPEX, their OPEX is increasing, which means not that very efficient. So they need to improve their OPEX. And also they need to improve the way they innovate, they develop, they deploy services. And as you heard this morning, you know, from Morning's Keynotes, OTTs can deploy their services pretty quickly. And telcos takes months, sometimes close to a year to deploy, to develop and deploy a new service. So with that kind of development processes, they cannot compete. So they have to innovate not just their IT networks, but also the way they develop and deploy services. So cloud computing presents great opportunities to them. Because with cloud computing, virtualization, abstraction, they can transform their IT, their networks. And with automation and orchestration, they can simplify their operations and cut down their cost. And also with the software-defined paradigm, they can innovate a lot faster and deploy a lot faster. And so I'm going to use three telco stories. And unfortunately, I can't give you their names, because I need to protect the confidentiality of their projects. But as a telcos partner, and it's our duty, and also we're an OpenStack Gold member, it is our duty to over time welcome all these telco customers to come and join us in the OpenStack community. And hopefully next year we're going to hear a lot of super users from our telco friends, and they can talk about their stories themselves. But today I'm just going to be very general, and I'll give you enough description, you probably can guess who they are. So first telco is actually a hybrid cloud case. And so this is Amir-based multinational telco with over 400 million subscribers, and they have a lot of a huge partner network running over 65 countries. And as you know, they are tier one telco, they need to innovate, they need to compete. So a lot of their line of business, they can't wait for their ITs to get their act together. So this is the shadow IT, right? They use Amazon. And ITs will try to learn about cloud computing by implementing VMware, but this is the time that they realize that, okay, the shadow IT is happening, and they are actually spending at least $2 million on Amazon. And they figure this is the time they have to come up with an open cloud platform that incorporates what they have existing, which is VMware, as well as Amazon, and create this unified open cloud platform for their own internal innovation. And also this cloud platform needs to work across multiple data centers. As you know, telcos, they have a lot of subnets, a lot of partners they work with, and they want to standardize this open cloud platform so all the subnets and partners can leverage this open platform to innovate together. And so this is their vision. This is a project. So from basically, they need this open cloud platform that's multi-tenant with one unified portal, one service catalog, and one service management, but it has to run across multiple data centers. And the cloud platform has to be highly secure, highly reliable, and also they need to plan for future growth, especially for storage. You all know that we are growing data in an exponential way, and they need to have a hybrid storage strategy to grow their storage for this hybrid cloud. And also, from a user standpoint, they want to be able to use any kind of public cloud or private cloud, but they do want to have a consistent way, a single pan of glass of administration and usage of their cloud services. And this is the architecture that Huawei proposed to them. So basically, we are leveraging OpenStack cloud platform, and this is going to be a multi-region, multi-availability zone-based architecture. And we need to plan for future growth. So we need to plan for a large growth, and we're talking about 100 availability zones and more than 100,000 physical servers. And then on top, we want to have a unified pass that all the developers can leverage. But however, if they still want to go to Amazon and stuff, they can still do so, but the goal is internally, they want to have a very mature pass. And the portal has to be fully distributed. In other words, each data center needs to have their own portal, but they do have this single pan of glass view. And so it needs to be Amazon, AWS-like kind of experience, and underneath it has to be a heterogeneous environment. In other words, the CTOs want to have vendor choices. They don't want to have vendor login. And this is Huawei's solution. So Huawei Fusion Sphere is an OpenStack distro. In the middle of this year, we just passed the interoperability certification. So we got the OpenStack powered logo license. And so we're very proud of that. And so with this solution, the Telco A can connect with VMware as well as Amazon. And in the future, they like to connect with Azure as well. And so the existing OpenStack that we got from the community does not fully meet the requirements. So we actually had to add an extension to incorporate more availability, more performance, more reliability. And on top of that, because Telco A's requirements to run this cloud in a distributed data center environment, so we need to add this cascading OpenStack capability. And there's currently a project within the OpenStack community called Tricircle. And today I have our lead engineer who is also a core member of Tricircle, Huang Zhipong. He will come out and talk to you more details about what this cascading OpenStack is about and the current status of Tricircle. Zhipong, can you hear me? Maybe I should just stand next to you. Is this turned on? It should be on now. Okay, hello. Okay. This is a nice crowd. I'm Howard Huang. So I'm currently the core contributor of the Tricircle project and also OPFV Multicite DPACC and PTL of the OPFV parser project. So today I will share with you guys how we open sourcing the cascading solution to the open source community. So first of all, just show our hands how many of you guys went to yesterday's Multicite OpenStack deep dive session. Okay, I guess many of you just can't get in the room. You know, this is so crowded. So first of all, a overview of cascading OpenStack solution. So basically we want to enable users to use one OpenStack instances to manage multiple OpenStack instances that is distributed across multiple data centers. Okay. So in this way, it enables you to have a unified management plan. The benefit of this method is that you could, for example, you could automate, you could do automation of the network creation just across multiple sites using cascading. Okay, you don't have to manually connect side by side, side by side. I think another benefit is that when you use cascading solution combined with, for example, OpenStack Magnum, and if you are familiar with Kubernetes, they are now proposing also a Federation solution called UberNatives. I think with OpenStack Magnum, you could actually just combine cascading and with UberNatives to give you a really great Multicite Federation solution. Okay, so let me introduce some of the history. I think we introduced the whole concept of cascading OpenStack in Paris Summit. That's October last year, I think. And then in earlier this year, we did the POC to see how many VMs, what's the scalability of cascading solution could support. And the POC shows that we could support the scalability up to, like you see in the slide, millions of VMs. That's really amazing. And in, I think around June or July this year, we put our cascading core architecture and contributed to the OpenStack community, which is called a tricycle project. And we also, from our cascading practice, we also contributed some ideas to the requirement project, you know, PMV Multicite project. So I think you could say that the open source cascading solution consists of the OpenStack tricycle, which is the backbone, the architecture, and the OPMV Multicite, which is the feature enhancement on the existing OpenStack components, like NOVA, Cinder, Neutron. Okay, I actually update this slide. Okay, so I think you could search on Google. We have a wiki page about the tricycle architecture. Actually, I think this is one Ayo and Pino showed yesterday. This is an experimental architecture. So the basic architecture we use in the POC would be to have a single standalone tricycle service that serves as proxies that talks with the top OpenStack instance, the one you manage multiple OpenStack instances through. So the tricycle service talks via RPC with the top OpenStack instance, and then the top OpenStack instance use tricycle services to manage all the, what we call the logically bottom OpenStack distributed instances. Now, the figure here is still experimental, you could call it. We are still trying to modify and have more discussion about how the architecture would evolve, but the POC concept would remain the same, that with tricycle we provide an extension as a service. We provide a standalone service that when you use tricycle you could deploy your multiple OpenStack instances across multiple sites in a cascading way, okay? Okay, that's... Sorry before I misunderstand. John Wayne is back. No, it's not. It's Clint Eastwood. I have to keep saying it's Clint Eastwood. It's not John Wayne. So in summary, we know that OpenStack actually offers a very good open cloud platform for telcos to deploy their IT and innovate services, and however, the nature of telco is multi-data centers, so the project tricycle is very important to meet telcos requirements, and we hope that project will move faster and then upstream to the community so telcos can fully take advantage of that. However, in order for telco to fully deploy IT as a service there needs to be a lot more automation that needs to be done and also in the hybrid cloud area that the data and applications need to be able to move between public cloud and private cloud smoothly back and forth, and those are the areas that still need a lot more development and so hopefully we can see more maturity in those areas. So the next story is the... is a public cloud use case, and so a lot of people in U.S. think that telcos run public cloud is very challenging, right? In U.S., I totally agree because you have the Amazon, you have the Microsoft, and you have Google. It's really hard to compete, so the U.S. telcos have a very tough time, but outside of U.S., actually telcos have a great opportunity running as a public cloud service provider, and it's because the regions that I have been to, like Africa, EMEA, and Latin America, Asia PAC, a lot of telcos, they're considered to be technology authority, they have the government backing, and also because of data sovereignty issues, a lot of U.S. UTT guys can get into their market as easily, so they actually have a great opportunity, and it is actually a very hot topic, telco enable public cloud when I travel to those regions, and telcos like to differentiate themselves from the UTT guys, so try to call themselves telco cloud instead of public cloud because they want to say we are more, we can offer more SLA, that kind of stuff, and so later you'll see this is the area that we need to continue developing. This remote is not very swift. Okay, so this data is from Gartner. As you can see, emerging Asia PAC, great China, Latin America, and Africa, these regions are high growth regions for public cloud, and so for those telcos in those regions, they actually have a great opportunity, and it's because, first of all, besides all those reasons I mentioned, they don't have the baggage of existing IT infrastructure, so for them moving to the cloud is a lot easier, and SMEs pretty much don't own IT, and then enterprises, they definitely see the advantage of using a public cloud, so public cloud is actually a very good opportunity for telcos in those regions, and telcos are investing in telco cloud pretty heavily, and these are the reasons because of data sovereignty issue, and also they are already selling city services, selling IT, cloud IT services, it's just another, you know, it's a very easier upsell, and then they can say we are telco-grade, and comparing to the existing OTT guys, they offer more SLA, and the key to success is they need to properly position and then also package their services, and so telcos are actually investing a lot. We are seeing, you know, this data actually comes from the informal telcom and media, actually we are seeing hundreds and hundreds of telcos at least offering one cloud service, and telcos generally they specifically build data centers just to run their cloud service public cloud, and so we see a lot of data center kind of projects which is about building public cloud, and however telcos, they are not as experienced as IT vendors in the sense that they don't know how to build up this ecosystem, because, you know, public cloud is pretty much driven by ITs, so that's something they need to understand and go to market how to sell cloud services, how to service cloud services, so this is where vendors can come in and help, and so who is making money? This is a million dollar question, so, you know, actually there are published information that some of the telcos outside of US are actually making money, like, you know, earlier today you heard some of the telcos, and also you have NTT, China Telecom, and, you know, Deutsche Telecom, I know these telcos, they are actually having, they actually have a certain amount of success, and the catch is a lot of telcos they expect to make money right away on day one, which is impossible because for any kind of public cloud service they need to go through this, you know, subscriber acquisition and all that and then get through certain maturity of their cloud services, and then, you know, they can start getting revenue. So what we are seeing is, oh, these first tier telcos are, have been deploying some sort of public cloud already using VMware, now they are going through second wave of public cloud architecture, and using OpenStack, and so having this conversation is very easy because they know that in order to compete against, you know, Amazon, these guys, they need to find a way to increase their margin, and that means they want to have options in choosing different kind of hardware they want, or hypervisors, and then they can tier the services based on, you know, with the open architecture they can tier the services a lot faster. So we are seeing the second generation, you know, public cloud telco happening. And business case is this one telco in Asia-Pac, and over this particular telco they are more than 51% owned by the government, so like I said, having a government backing, they have a lot more advantage being the de facto public cloud service provided for their region, and again this is, you know, a large telco with over 2 million broadband access users, and they have deployed a public cloud already using VCE two years ago, and back then they can only reach to 99% SLA. And so now they are building the second generation public cloud using OpenStack, hoping that with OpenStack they can increase their margin, decrease their cost, and also they can achieve more differentiators. And this is the solution that we offer them, and it's using Huawei Fusion Sphere, and as a cloud platform, for public cloud generally technology is not the tough part, it's the business model, it's the go-to market. And so with Fusion Sphere, which is OpenStack based, they have interoperability, they have options in hardware and hypervisors and containers, but the complicated area that needs a lot more development is the service, portal service catalog, how you onboard a service, how you, you know, automate this whole service ordering process due billing, and the billing has to be connected with telcos, you know, BSS OSS system, which is very complex. So the area on top, as you can see, that's where most of the professional services come in, and they need to work on those developments, but the cloud platform is pretty straightforward to them. So here this is how we, Huawei, we help this telco win. Instead of, originally they can only achieve two nights, we help them get to three nights, so they can say their differentiator competing against Amazon is, you know, they can offer it more SLA, and they actually have a goal of migrating 300 SMEs from Amazon to their cloud, to their telco cloud, and they also decrease their cost, so that means they increase their margin, and, but when they build their public cloud, one requirement is they want to make sure that the user experience is the same as AWS, so their customers can easily migrate their customers from AWS to their public cloud, and also they need to offer regionalized services, you know, kind of services in US might not apply in their region in terms of pricing, in terms of capabilities, so it's very key for them to develop their regionalized services, and building of that ecosystem could be very challenging to them, but it's a must for them to differentiate. And so in summary, again, you know, when telcos got great opportunities, and then Openset gave them the capability to use different kinds of hardware, so they can offer different tierings, in that case they can monetize better, they can have more margin, and however what's lacking Openset is, you know, telco great, Huawei, we had to add a lot more performance reliability availability to help telco differentiate, and also building this partner ecosystem, and right now pretty much is very limited, and in order for telcos to offer a lot more regionalized services, they need to work with vendors to help them build up this ecosystem, but I have seen success stories, they have apps specifically for law firms, dental offices, and stuff like that, and the regional businesses can fully utilize, and once that takes off, then the public cloud will get a lot more adoption. Okay, the third use case is NFV, you heard a lot about NFV this morning from today's and NFV is very hot, however it is the very early days of OpenStack adoption, and so I don't want to repeat too much, but everybody knows the benefit of NFV with virtualization, telcos can transform the way they develop and deploy network services, and they can cut down their outpacks and cat packs, and increase their innovation in delivering telco services instead of using silo proprietary hardware now, the services can be virtualized so they can deploy a lot faster with a lot more efficiency, and this particular telco is an AMIA based telco broadband and communication providers, they actually run business pretty much globally everywhere, Europe, Asia, North America, and South America, and this is one of the largest telcos in the world, and they really need to innovate their network services, and looking at NFV and OpenStack, and hopefully they can reach their more network service innovations, and so I just want to show this one slide to show you that this is, the current data center is a mess, they have more than 90 data centers, they have more than 50,000 servers, more than 1,000 applications, more than 20 OSs, lots of cat packs, outpacks, and time market is very slow, this is a typical first tier telcos environment, and for them in order to move from this environment to a highly centralized, highly, you know, cloud based kind of environment, they go through this NFV process, and today I have our architect for mobile, Pagash, who is going to, who is very intimately involved in NFV development at Huawei, and he's going to come out and talk to you about how Huawei help this particular customer. Thank you Anay, and thank you audience. So, first thing I want to go with is what do we do in NFV, okay, hopefully. So, let me bring you to the open platform for NFV. Of course, every vendor wants to make money, no comms about it, but at the same time you need to have some standards so that one can build on it. So, we started our journey way back when 2013, when we joined OpenStack as a Gold member and we did some proof of concepts, et cetera, internally, before we came to OpenStack and looked at what do we do, and then we realized that OpenStack has great things, but at the same time it lacks a lot of things, and especially Neutron was really messy. It has evolved over a period essentially over the last couple of years, starting from Folsom to Grizzly to Havan Advice House, Juno to Kilo and finally we are there at Liberty at this time, and it has definitely stabilized a lot. As you saw in the morning, the percentage has grown quite a bit up to 89, 90 percent now. People use it for production. Still, that doesn't still do the carrier grade requirements that are requested by telcos like AT&T, CMCC and various telcos. So put it this way, we went through the open platform for NFE around last September, and then by January we were able to pull together some kind of a requirement. There are currently in OpenFE open platform for NFE, around 38 projects in that group, out of which only four were really released in the first phase called Arno, and that was just to set up the OpenStack plus ODL. So the issue is, when you look at the HC reference, you have got 3 tier, the WIM, the VNF manager and the NFE orchestrator. On the left side, you will find all of the whatever platform that is NFE, it pops what we call it. To get those pops, you need to have first mechanism and that was the first target. So we decided to go with Arno in that first release, what we did was we took OpenStack plus ODL, of course it's not me, it's the whole industry. We have all the competitors, everybody came together with the service provider providing the requirement, and we clipped together the OpenStack plus ODL. At that time there was only SDN available, and then by combining them, we focused on the continuous integration under October's project, what we call then, and that provided the first release somewhere in the June timeframe. It slipped by some time, but you do expect the growing pain, but later on now we decided that it should go into second release. So Arno Nix was followed by L3VPN, and in that we decided to have some of the key features, because for anybody to use if you have multiple sites site to site VPN, L3VPN those are the key telco requirements. Today also we had Bird of Feather and then we discussed with Verizon and AT&T we found that there is what we call L3VPN requirement. So we integrated ODL, open source and open control and that's the goal in the B release. In the C release we are looking at dynamic service training and VNF forwarding graph, and going forward we are also looking at supporting the 4.5G which requires the IoT and MTC which is massive number of people and that's the way we are going. So next I have here what did we do in open source, that is one thing. What do we do inside? Inside we are trying to provide products that needs the telco challenges. So we have some of the challenges which went through, we started with cloud OS and then that's the one which is the more virtualization part, the VIM part the open stack part of it and then at the top we have the VNFs which are VEPC, VMSC and applications. So this is what it is, I will go to the next slide which actually shows the gut of it and the gut of it is this that you need to have something at the bottom which is our fusion sphere product line and then we added to that the product lines which include fusion sphere is one VEPC that is the packet core which is virtual packet core VIMS which is the GI LAN side of it which is the mobility part of it where we need to provide services, VMSC is the multiservice engine then value edge services we have to provide on top of VMSC platform so we have from radio to the core to the what you call the GI LAN or the service side on the radio side of course we have the cloud edge which is the edge services. So these are the product lines which we have put together and then that's what we have given as products to the industry and the whole in this the only thing which carry the message is that we did try lot of stuff but we found that Onos SA SDN controller being key for our product line and so we are focused on Onos even though we started with ODL and we do support both of them. We are agnostic however our product line we want to go with the Onos and that's the key message in this I think this is what I have and I think I'll handle it. Thank you Prakash. So we know in summary that you know telcos definitely see the benefit of OpenStack for their NVV project however the biggest challenge is you know telcos come from world of standards right so you have the this is not going you know I'm just this is not forwarding hold on I'm so sorry is it okay good I got some technical guys helping me out so the biggest challenge is telcos come from world of standards and there's Onos open NVV and you know at these standards all of these standards they all have something to do with NVV and then OpenStack obviously we have something to do with NVV the key thing is to have all of them aligned so we are all moving towards the same direction and this is where as a body together OpenStack community we can you know we're together with the open standards out there and so we can push this NVV progress forward and so in summary I just want to say you know it doesn't matter whether it's a hybrid cloud private cloud, public cloud, telco cloud or NVV and telcos definitely see the benefit of OpenStack they came from VMware world Amazon world now they see and they really the CIOs really believe the promise of interoperability they do understand that OpenStack might not be telco ready today but the fact that the interoperability is going to allow them to have vendor choices they don't have vendor login and then also there's a community innovation it's worth it enough for them to try OpenStack today so as a community that's all work together to move the telco to move OpenStack maturity forward so we can make OpenStack truly telco ready okay with that do I have time for question I'll be happy to answer a couple questions and I think that will be it here microphone okay can somebody pass the microphone Jim hi thanks for the presentation so your first use case was about a carrier looking at internal IT transformation the second was a carrier looking at becoming using OpenStack as a public cloud are any of them thinking about well possibly leveraging both because otherwise it's a very good question even NFV too so from technology standpoint it's totally possible in fact that's what Huawei we offered you know we have this reference architecture called SDDC square cloud distributed architecture however from a political standpoint it's really tough because generally NFV is driven by network leaders and then you have private clouds it's generally that projects led by IT and then the public cloud is led by the enterprise group or CMOs so a lot of times that you know they want to have a control of the platform they all want to become the cloud owner for the company for the telco so a lot of times when we work with them we need to we generally talk to all of them so from a technology standpoint it's definitely doable you can specify VDC according to your requirements but it's the processes generally they are lacking but it's a very good question but I wouldn't be surprised that five years down the road there will be one big cloud platform that services all the business units within telco okay any other question yes microphone please alright thanks for the presentation I appreciate that on your third use case is it my understanding correct that's going to be a brownfield deployment on your NFV you want to answer that question I would say that brownfield is a little farfetched greenfield is the correct approach to go then bridge the gap between the brownfield and greenfield by having a migration plan that's my question right so for that particular use case was that a greenfield or was that a brownfield it's a greenfield brownfield is little so what's your take on how to bridge the gaps so bridging the gap occurs through because network is computing so it is through the STN control programmability that's the key so VC going forward that STN will be the key infrastructure with MANO and all that but key is STN which is gluing all this so to glue the underlay the overlay the old way of doing things versus new wave all has to go through a STN control that's the key message from a business point of view if your physical infrastructure is already completely appreciated why would I want to pay to build on the virtual environment I understand going forward probably makes sense but I wouldn't want to do away with my nobody is asking you to do away it's a question of when you do a park you end up doing a selected segment or a slice and then eventually when it establishes now you say hey this doesn't go with my infrastructure how do I put together so we do see that if you do a slice which is physical underlay you will program through STN control directly and so it could be existing or it could be new it doesn't matter to us right so STN is the key programmability is the key for all these things how you go about it automation is another thank you thank you I think that's all we have for you to come to us