 Welcome back, day two, session two of Open Networking and Edge Executive Forum, ONIF as we call it. We really appreciate you all joining and this is a very exciting section today. We have some great speakers and representing pretty much the top companies that we can think of. And what I would like to highlight again for those of you who may have just joined, you can ask question and answers. These are live sessions, not pre-recorded. There's a Q&A button on the top. Just send them the question. If you have been with us in the morning, as you know, I was able to at least get 15, 20 questions quickly into the speakers. So, you know, they will answer it in real time. And we really appreciate you all joining. So, let's begin the second section with a company that also needs no introduction by do. And we have the chief architect of IoT, letting Lee, who has been a very active contributor to several projects in open source in the Linux Foundation Edge as well as other umbrellas. And, you know, we want to, you know, understand what they are thinking from an open source and an Edge perspective. So without any further delay, please welcome Lede. Hello, everyone. I'm very glad to be able to participate in this exclusive forum today. I'm Lede Lee, the chief architect of Baidu Cloud IoT. As a member of the Arab Edge Technologies Steering Committee, I'm very fortunate to be able to participate in the technological innovation of global edge computing. Edge computing has grown rapidly in the past few years, from the early as a data agent service for the IoT to a general-purpose computing service covering various applications, granaries, and facilities. We can see that with the population of various new computing resources, such as the mobile edge computing, the new kind of content-delivering network, the boundary between the Edge and the Cloud is consistently blurred. By moving towards an era where computing is ubiquitous and workloads from the Edge are going fusion with the Cloud. We can see in many real cases, companies collect data in the Cloud and the train models and then use these models to perform a real-time data processing at the Edge. This means that the Edge is becoming combined with the Cloud to become a single integrated service. What I want to share with you today is how the Beetle product of the LF Edge connects the Edge and the Cloud together. Next page, please. First, let me briefly introduce the Beetle product. Beetle, organized from the in-challenge Edge product, incubated by Baidu in 2017. And then in the next year, we decided to open-sort it and successfully joined the LF Edge community in 2019. Today, Beetle has grown into an open and a vendor-neutral product. As one of the earliest open-sort products in exploring Edge computing, Beetle has already got a bunch of features. Beetle can now provide a microservice orchestration in various IoT environments and can simply extend the Cloud applications and data to users' private networks and private network equipment and provide a continuous application operating environment for all type of equipment. You can visit the project homepage, beetle.io for more technical information. And next page, please. The Beetle project has set three core development goals for its value, corresponding to the applications, the management, and the developers. In terms of applications, Beetle's goal is to bring a standard and open-cloud native environment for all Edge screeners so that applications can be freely deployed and run according to the actual needs without adoption and modification. The next goal is to allow equipment to operate automatically and in unattended conditions through real-time and through the remote management and configuration. And even can be able to make independent decisions and collaborate with each other when the environment and the data change. The final goal is to provide developers with end-to-end application development tools and services to promote the further innovation development of Edge computing. And next page, please. In the past years, Beetle has made great progress. We have released an important 2.0 version, which enables us to move further towards Beetle's ultimate goal. In Beetle 2.0, we have introduced a standard Kubernetes environment for Edge devices. Applications can be defined in the form of resources such as pause, service, and deployment, just like what you do in the cloud. Beetle will accept these definitions from the cloud, save it to the local storage, and manipulate local Kubernetes in the local cluster. In fact, all services, including the Beetle system itself, are cloud-native. Beetle runs as a Kubernetes service, and we have implemented the self-aggrading with the Kubernetes way in Beetle 2.1. At a state time, we continue to support the Docker and the non-container bare process models, this for the low-end commits and introduce an experimental windows environment support. Beetle is working hard to continuously expand the range of device support and provide developers with more different choices. Another important feature is that we released the Beetle cloud. This is a completely open-source remote management system. You can use Beetle Cloud's open API to realize the configuration, software upgrade, application deployment, and data collection and synchronization for various edge devices. Beetle Cloud can directly connect with your Kubernetes control plane to obtain resource definitions and send them to Beetle with a secure network channel. With the cooperation between Beetle and Beetle Cloud, the edge and the cloud are connected together. Next page, please. So here is the architecture of Beetle. Beetle Cloud is located in the upper part. You can find that the whole Beetle Cloud itself runs as a service in Kubernetes and the Beetle Cloud connect to the Kubernetes API server. So by this design, the Beetle Cloud can receive all your Kubernetes configurations and try to synchronize it to your local edge instances. And the Beetle itself is located in the lower part of the picture. And Beetle and Beetle maintaining the connection with both the Beetle Cloud and its own Kubernetes cluster. So in logical, the local Kubernetes cluster and the cloud Kubernetes cluster can be configured as one. They are integrated by the operation between Beetle and Beetle Cloud. And then in this picture, we found that all the original Beetle services are moving inside the Kubernetes environment. In the previous version, the 1.0 of Beetle, we run Beetle as an independent separated process out of the container system. This process will operate with the Docker system and run all the applications as the container in Docker. But the Beetle itself is controlled out of the container system. So the administrators need to lock into your local console and update Beetle itself manually. We found that this is the first step to support the edge system, but this is not our ultimate goal. We want to make this component unattended. So in the 2.0, we move all the Beetle systems inside the Kubernetes. We define all the Beetle services as pods, as services, and we give the privilege to connect with the local Kubernetes API server with the standard Kubernetes way. And finally, we can upgrade and manipulate all your workloads, both the system services and user applications in the same way. So this is what we do in Beetle 2.0. And we consider this is our most important move to our goals. So next page, please. And next, please. Maybe it's a later. Yes, this way. Thank you. So to achieve the integration of the local Kubernetes cluster and the remote cloud Kubernetes cluster, there is still an issue to solve. The most critical issue is how to maintain data consistency between the cloud and the edge and unstable network conditions. Because most of your edge equipment is deployed around the world. There won't be a very stable and high quality cable network connected to a data center. We use the unreliable network, maybe the 4G or even the more lower bandwidth network connections. So there will be data consistency issues. When you're configuring your cloud, the data is not synchronized to the edge side successfully. So they will run in different view of your configurations. That will make data correction. So we have to solve this issue. This is the shadow CRD technology proposed by Beetle. In this technology, we combine the user-defined resource of Kubernetes with the digital twin of IoT together. The former one, the CRD guarantees the integration with Kubernetes environment. So all your operations will go to the Kubernetes way and the later one will get the exactly once QoS requirement through the MQTT protocol and the desire and the report semantics. So let me show how this works. In the picture on the right, there's a typical edge cloud synchronization process. First, the Beetle cloud connects to the Kubernetes API server and the cloud, obtains the deployment resource, and the pack them into a shadow object. We call that a desire object. And then sends the desired shadow object to the edge through the MQTT protocol. The MQTT protocol itself includes the capabilities such as the error checking and the network retransmissions and fulfills the quality of service. By using the MQTT protocol, the desire shadow object will successfully be transmitted to the edge side. The edge side, by receiving the shadow object at the edge side, the Beetle will unpack it and send the original CRD definitions to the local Kubernetes cluster for the deployment. After the deployment is achieved, the new state which reflects all your configurations of your applications and the data will be collected by Beetle from the local Kubernetes API server and the packed again as a shadow object we call this shadow object as the report. And we send this report shadow object by the MQTT protocol again to the cloud. And in the cloud, the Beetle cloud system will unpack the report shadow object and reflect all the changes from the edge to the cloud. This will solve the data consistency issue through an unstable network. This is the shadow CRD technology of Beetle. And next page, please. So finally, let me introduce the Beetle's roadmap in 2021. The first and most important is we will continue to work hard to achieve a deeper integration of edge and the cloud together. And ultimately, we'll use a single control plan to manipulate all your workloads. Our image, our think, finally, in the future, all your workloads can be controlled and be declared, controlled, and observated in a single Kubernetes dashboard. So this is our final goal. And we will make it better in 2021. And how to make this happen? The first important step is to strengthen the observation and the measurement of edge status. We will send the edge health data and application runtime metrics data to the cloud to provide a better capability for the operations. In addition, we will work with the agrino community to explore how to use the mobile edge computing by 5G network to implement a better low latency services for new kind of applications like the other driving or the VRs or all the cloud gaming. And we will try to research how to serve a moving application by your mobile phone or by other mobile devices that will move across different base stations of 5G networks. The application should go to keep low latency with the applications. So this means the application maybe will be roaming between base stations or roaming with the MEC data centers of 5G networks. This may be, I think, these are very interesting things. And we will try to explore how to achieve this in this year. At the same time, we'll strengthen the construction of documents and to attract more contributors, interesting beta products. And we are going to collaborate with more chipset providers and more infrastructure providers to support not only the X68 chipset, the arms chipset, maybe the RISC-5 chipset in the future. And we are going to collaborate with more Arab-edged communities such as address foundry and agrino to make more cooperation. Next page, please. This route will be reflected in our specific detail works. The most important thing is the cooperation with communities while working with the AgX foundry to bring an open source IoT solution to beta. So users can define your IoT system in the cloud and test it and deploy it to all your Edge devices. And by trying to work with the Flash community to introduce more industry screeners to accept more industry protocols and the data to the Edge system and connect them with the new technology like the data processing or the artificial intelligence. I believe that with the joint efforts of the community, Edge computing will play a more and more important role in the future. So this is what I want to share today. Thank you. Excellent. Thank you, Leading. Thank you. Yeah. And really, we have seen a lot of great progress in just one year since the time we chatted. And I think the Beetle has not only gotten a great roadmap but a lot of collaboration with LF Edge. LF Edge, in general, has a lot of momentum going into from a unification of frameworks perspective. So thank you very much for your leadership. And thank you for communicating today. Thank you. Thank you, Arbit. All right. So we're going to move to the next presentation from the director of open source we who at ZTE. Again, company needs no introduction, obviously. And I'm just going to hand it over and hear what you have to say. Take it on. Hello, everyone. I'm Jeff from LTE. And in this time, I want to try to introduce my companies working on the open world. Please move to the next slide. In this picture, I want to try to demonstrate our footprint in the open world, not only the open source community but also the open standard. As you know, ZTE is a networking equipment provider. Our customer is the main network operator and also consumers. And now we expand our business to a lot of vertical industry like the manufacturer and also to cooperate with our partners and working with partners to expand our business. In this picture, we can see ZTE mainly working on 3GPP, ITU, ATC, and IEEE. It seems like that in the open standard organizations. And also from the 10 years before, ZTE began to move into the open source areas like OpenStack and also Linux Foundation Networking and also Linux Foundation AI and also a lot of others. Next slide, please. First, I want to introduce. We have worked and contributed a lot of open source area communities like OpenStack. Since 2013, ZTE began to contribute to the world of cloud in the first structure. In the OpenStack, we continue working until now. And in the train release, we ranked the seventh in the world and also we have eight project leaders and ranked second place in the election of the OpenStack Victoria Release project leaders. And we have 13 core members working in the communities. Next slide. Since 2014, we worked from the ACI NFV standard and began to cooperate with the carriers. On the OP NFV, we began to the 19 members and also working contributed on several key projects like Qtube, DayC, Farrows, FungTest, and Relent. And also we have a board member, TSC member, and two cores. And we get the OVP certifications like the infrastructure and also open labs. We continue to cooperate with the OP NFV in the CNTT last year and we try to work on the advocate in the future. Next slide. In the SDM field, we are the member of the pending member of the Urban Daylight. We ranked fourth in the history. I think the Urban Daylight is more stable now. And we have six PTAs and 21 core members, one TSC member in the communities. We have initiated six new projects like Bias, OFF Configure, P4 Plugin, Telemetry, and Datanet, and Outer Datestores with our partners in the ODL communities. Next slide. Begin with 2016, we joined with China Mobile to initiate the network automation project, OPEN-O, and also be a team member after OPEN-O joined with ECAMP. We ranked third place by 7.3% contribution ratio. And also we have two PTAs on micro service bus and home bus. We have 21 core members and TSC members. And also we worked with operators on VEPC and VIMS, which is verified in China Mobile's lab and also with China Telecom Lab on VCP. Next slide, please. Since 1907, we also contribute to the cloud-native world. In Kubernetes, we ranked first on the world and also second in the China area and also cooperated with Docker and also in the computer-cognitive. We worked on the container D and ranked second place in China. We are the CNCF Gold member and also we have several key members in Kubernetes. Also we get the Kubernetes certification for our product, OPEN-Palates. Next slide, please. Also we focused on the massive storage areas in CERF project. Now CERF became a CERF Confoundation. We are the founding planning member in CERF Foundation and we have one TSC member and three core developers in CERF project. And in the release K2M, we ranked third in the world and the first place in China. We also published some two books on CERF to promote the CERF technology in China. Next slide. Also we initiated a project on AI in the AI plus data project, plus data foundations. Adelaide is an end-to-end optimizing framework for deep learning. It is accelerating deep learning inference process for both cloud and embedded environments. The main components of the Adelaide is a model compiler which can use different model types. And also we have the inference engine which can increase the speed of the inference process. And also we have a branch mark test module to branch mark help the application developer to test their module. Also we cooperate with ITU on AI and ML in 5G challenge in 2020s. Next slide please. So what did you focus on? I think in the open world we will still continue contributing on the 5G-related specifications and also in the open world we work on contributing on the general efficient flexible platform and frameworks. And also in 2020 we have joined the RISC 5 and also we support China Mobile to initiate the XGVILA project. Which we think is very important to our customer and the end user. In addition we will continue support and try to get the batch to certify our products in the next year. Please go to the next page. In this page I want to introduce our product on the one stop 5G private network which can enable in all industries. We have two or three different kinds of configurations from left to right is first one is to do like high configuration. It consists of two cabinets, one for the network, one for the cloud. In the cloud cabinet we provide data middleware and AI middleware and also video and age. It is suitable for big enterprise and headquarter of the end users. In the middle is an integrated cabinet which integrate the cloud with the network. It's more efficient and it's suitable for the enterprise branch and some campus. And also we provide the simple configuration which is a single shelf and it's the most cost less. All of this configuration provide you cloud and network convention and also the fixed and mobile network. And with full access to different kinds of connections. And also we support multiple computing conventions with different type of computing. Next slide. In this slide we show in ZTE common age for Synology Mac which provide you the Mac talents with self servicing provisioning. And provided the operator the ability to unify cloud management. This is a one-stop deployment and we use the embedded Mac, pre-integrated Mac and also Codifirm for Mac. And also this one solutions has got the layer one, two, three award in last year. At the end, please go to the next slide. At the end, I think we can discuss some challenges and opportunities like before us. Since we know most vertical industries has recognized the power of ICT especially on 5G, H, cloud and AI. This can help them increase their productivity and efficiency. At the same time, they also fill the trade for the industry disruptions. This can be a challenge and also opportunities. And also for the COVID-19 pandemic, we think it's already changed our life and work, maybe forever. Working from home has been verified, both in our company and also in most other industries. And also traveling has been limited until vaccines were used worldwide. I think this will take a long time. I'm looking forward that we can meet in person in the future, in a very near future. But in my own works, I found a lot of product people don't think an open source could make much profit for them. How do you deal with them on this? How do you persuade them to trust open source and make them believe the open source? Thank you. Excellent. Very good. First of all, I have to congratulate you that all the various projects and various umbrellas that you have been participating, collaborating and leading in a lot of instances with your team of technical people and yourself. So thank you very much for doing that. I think the overall business model on open source is fairly mature now. I think you heard from Rakuten and others, as well as you can see from the partnership that the carriers have provided in terms of going through the vendors versus just doing it themselves, right? So I think that the things have settled down quite well. So very good. Thank you very much. Thank you for doing this. Thank you, Rakuten. Very good. Excellent. Then now for the final presentation, a very important one, because I do believe that what China Mobile says, the world has to listen, it's the largest carrier, lots of research going on, lots of very, very smart people and we have one of them to speak to us today. And so without any further delay, you know, I do want to welcome Dr. Fang, you know, Chief Scientist and General Manager of China Mobile Research, and I would, you know, we will love to understand where things are, where you see things going, what we should all be looking for. So Dr. Fang, take it away. Hello. This will be interesting. I don't see you, but you can see me. We'll see how this will work. I was very happy and to see our bit, to know we have about over a thousand engineers and the practitioners join us. Today I will not only talk about China Mobile, I just have one slide. So pretty much is how I see how we see the entire industry, how that goes. So move on to, so continue to talk about the topic here is network intelligence, and I see not China Mobile, probably across the entire industry, not just our industry, other industries are increasing their investment intelligence, how to intelligent by the internal management's internal operation across various industry. So today I will cover a few aspects so I'll leave it myself to 15 minutes. What is the motivation talk about what what are the main drivers, what are typical approaches companies are taking to making the business making the operation intelligence. Second is how our observation. And then I will move on to share a survey run conducted by UAG in the community in this community, see how the community says about the stage of their network intelligence, and I will give a few suggestions. Next. So, this one is the motivation. Since they have lots of small funds, so open up for my own version. So there's a couple motivators I'm thinking are driving us to transform towards intelligent network. And the, and the lower layer is the network itself. It's before we have the wireless so we have the core and the transmission network. Now this the whole structure is changed. You have a basic infrastructure and that layer. You have the VNF, you have CNF, then you have the operation layer, then we have the business and service layer. For for us, and how we see the industry I'm thinking that operation intelligence now is the core. If you, if we read all the reports from the agencies and a couple SDOs, we can see this trend where the and this current stage where the intelligence efforts goes to, it's a pretty much focuses on the middle layer, the operation layer. So it's the promotes the transformation to further improve the efficiency and the lower the cost. And also, you know, we have to, they scale up the network, and the data volume network is increasing dramatically how we can in, you know, operate that network and the same time maintain lower costs. And why now we here we see the trend is taking that layer as the core based on layer, then synchronize the efforts with the lower layer they the network itself become more intelligent to also the, the service and then business layer. So where we see a trend is for the typical applications of a 5G people are proposing are, are deploying, and also the typical applications of 5G, the AI and larger scale, over 85% are overlap. So basically, before we plan, we build, we optimize our network for, for voice of IP for data for talking over phone. Now it's the, I'm thinking in the applications with intelligence as a key feature are the, the probably one of the biggest services we have to support well. So, as a key feature of intelligent applications they have to be flexible has to be close loop maintenance. So that will require network to have the same feature. So our current network is not as flexible as the applications hope we are. So that's the, the service layer so we can hope one day we can users give their intent, we can, we can control our network to serve that intent to improve the customer experience for the second of the operation layer. I can see the trend in the standard organizations in open source our open source community and the other open source community and also in the research literature. Lots of great work are coming out on this layer. How we can further control their costs and prove the efficiency. And in terms of network the lower layer. There are a couple of things one is for the network elements, each network element has their key functions. And some of those functions have been changed because of the intelligence technologies that we have as of today. And then the other one is the interface between the network elements, since we have this technology, the way they interfaces between the network elements are changing. So next slide. So this one probably looks quite complicated on the left side is the, the campus forum, they show how we need a few close loop in order to keep our operation intelligent. So, so here you see their three loops that we need to have once we have a business intent, then translate to a service intent and transition network intent and transition to resource intent and the data. And then come back, we need to keep promise so we can provide that service so it's the high quality that's the KPI, and, and, and this SLA feedback, and then the customer experience assurance. So that's the big loop. And in, in this whole circle. You see there is a horizontal circle, which is the business automatic close loop and the service automatic close loop and the resource to control close loop. So that's pretty much that I want the key feature of artificial intelligence applications is you cannot have a product and without touch it for long. So it has to be in a loop, and then the model has to be upgraded. And the data has to be continually processed. So it's, I call it's something itself has it's a life being versus a, a product, the traditional products which you buy a cup, the couple stays on the table. So on the right side, just open up how different parts are connected. So now we can see for the SDOs. They have laid up lots of efforts, almost every single are standard organization has put in lots of work in various forms to push the network to be intelligent. So in this graph you see you have the pink one is the local AI. So for the while, while assuming the run transmission network and the core network, each domain, there is specific efforts in that domain. But we here I show is that's a local AI. Well, I'm proposing here is the if we hope we can provide the service which as intelligent as the business hope we can support. Now we need also a general and intelligence layer that's the right side. Which can cross this from the top all the way to the bottom so we need to have that view. So in order to have the overall optimization optimization targets versus we optimize in each specific domain. And about, we don't have the overall picture that's will make the whole network fall into a local optimum versus a end to end intelligence operation. So next slide. So one please. This next slide. Just talk about. Oh, did I go back to go back to the open source. Yeah. Yeah. Thank you. So this is a quite you can see the pictures are quite similar to the one before just showing where our community that the software the projects that we're putting where we see them on the overall picture. So you can see the on up is it's especially just take on up example. It's a very key components for to assure that business intelligence and the service intelligence layer. So we have acumen so have on up we have the NDA, I hope we can increase that efforts in that layer, because the operation I'm thinking at current stage will be the driver for the top layer and the lower their intelligence. And so we see now I trained is on up also closely work with the open run. Work with the Rick to help configure and orchestrate intelligently in that specific domain. So the infrastructure layer. We see there are a couple of standards come out for fed AI, which is quite hot. It's a, I think it's a standard was passed in 1912. I, I, I last year 2019 to basically redefine how machine learning can learn a model without the risk of data security. This is next to one. Go to the survey. EU AG. Next slides please. Okay. Sorry, I have a cold today. So this is a initiated proposed by China mobile was conducted by our FN UAG. So so far we got 65 feedbacks from 20 telecom operators and more than 10 digital service providers and over 30 vendors. And then quickly go through the data. So overall, looks like we're asking that questions where you see we are in terms of intelligence level. We can see the most companies think they're at our lines out or our two. So it's a level one or level two so so we are at the preliminary stage. And to see where the applications and in the research. So that's also after the pilot stage, which is the work has to hand. So see that topic as network intelligence has become quite hot topic now in the research community, but I'm thinking as the industry we're still we're starting doing that. So R&D can see how, how the intelligence, the a development and the deployment, how who should be the one to do in the work, how they, the operators to see this picture. So we can, for the proof of concept and pilots, we can see the percentile for the on the right side, and the first picture, and the second pie graph, it shows the percentage of self development and rely on partners rely on bigger partners and the small partners. So here, I have a couple in conclusions. And for now, from the survey so some more than 50 operators think they are doing the entitles of platform on their own to the South developed, and the more than two, two, three, two thirds of the operators to think they are also doing their own applications to just to see where well this data we have run the analysis. I'm thinking some of the conclusions in line with our observations some are not. So we will continue to have a new ways to know where overall we are in terms of the whole industry. So I'm going to start almost I use up all my 50 minutes I will quickly go through the suggestions. So I suggest for the three slides. The first one is, we hope we can carry out a top level design for platform function and the manager architect. And build the overall in the general overall general intelligence across the layers on the right side, and build an intelligence and the open industry cooperation platform. We also need to work closely with SDOs, and also other cross organizations also work with the AI companies, and for LFM work well inside our community, we hope we can take operation intelligence as the core, and then we take it as as a driver to have the intelligence transformation and the service layer and also the, the, the network element layer. So we draw those two graphs, and then one is the service intelligence and the other ones the lower layer. Hopefully, that's our just suggestion, maybe, maybe it's right maybe it's not to see the path. The next one is, I'm thinking is in order to make our network intelligent it has to be flexible to be flexible it's not enough. We have a certain amount of network operation knowledge. We have a certain set of data, we also need an environment, an environment which can, which can be a resembled or experimental place in developers and algorithm researchers they all can put in their model see how that work. A simulation environment I'm thinking is pretty hard for me the intelligence to come from what the intelligence of computer come from three resources, one is data, one is knowledge, and the other one is the environment. So the model, the AI algorithm and the model, the product has to explore the real environment, and to improve the self over the on the fly, and to truly get intelligent versus we have the offline intelligence. But once we put in the real network, it's always a bigger question mark. So they are free of put that into network and make it a larger scale. So that's, I will finish my slides orbit. All right. Excellent. Thank you, Dr. Fang. So first of all, you know, I was just so engrossed in what you were saying that I just didn't realize that, you know, we're there. A couple of observations and there's obviously a couple of questions that I want to I want you to address first so the one of the questions is saying China Mobile was largely responsible for the CRAN technology which triggered the entire 5G movement. What are we going to do and how is China Mobile, you know, thinking about once again taking a leadership position and continuing in the future. Yeah. Thank you for the question. Well, CRAN basically is, it's a great invention, it changes the overall architecture of the grant. So I'm thinking for now probably for talking about the intelligence there will continue our overall efforts in CRAN. I'm thinking is that we want, we hope with in China Mobile is invest a lot in the networking intelligence and the same thing further for the grant. So we hope that the CRAN give us more opportunities so we can put the intelligence in. For example, you have to use your separate to DU and CU and between the DU and CU and the AAU. So how we can match the resources. So you need something in the middle, which is the intelligence enough, we can sense the network, sense the dynamic and automatically intelligently adjusted in mapping. So that's the CRAN truly give the architecture, give a way we can put in patches in. Okay. Excellent. I think in general the, your overview of how standards and open source and network intelligence and all the layers couple, they all look very cool and I think with your leadership and with your entire company behind the open source and standards. I really appreciate the support that you all have given. I also want to publicly thank you for being the chair of the governing board of LF Networking, all of last year. And, you know, we're looking forward to some even more cool things from China Mobile and your influence globally. So really appreciate you taking the time. And thank you very much. I'm so happy to work with the community. We have to work together, change the world, make it better. There you go. Thank you. Excellent. This ends our second day of ONIF. We will see you tomorrow morning at 7am Pacific. It'll be a little late in Asia, but do join in. We have very exciting presenters on the EU front and, you know, look forward to all the recordings that come out. But thank you again for joining. Good night. Good morning. Good afternoon.