 Hello everyone, we are very glad to share our work on telecom cloud and today the content we bring to you is practice and thinking of telecom cloud management platform based on cloud native and The slides will be shared by two guys Zhang Yan and Li Mingxuan both of them from China Unicom The I am Zhang Yan first I will talk about the telecom cloud management platform And we call it unified cloud management platform Then Li Mingxuan will give the speech of the cloud native And the content about the unified cloud management platform is made up by three parts Now we go to the first part the challenges and problems on telecom cloud management Let me introduce the telecom cloud It is based on OpenStack We each are infrastructure of China Unicom to carry future telecommunication services and support the development of future network service Including 5G service ad computer services and so on With the rapid growth of the scale of cloud resource and the increasing number of servers the following challenges have been brought into the management First of all This data center resource force in China Unicom is all over the country The resource allocation need to be unified and optimized In order to achieve a good multiple Data centers management it is necessary to consider that Providing the capability of accessing to different cloud platforms and unified infrastructure integration second massive physical equipment and cloud resources require unified operation and maintenance and also need unified monitoring and rapid-force location and challenge of the cooperation and maintenance functions are emphasized three Petro-genius resources requires a unified portal for operation and maintenance and to shield the differences and improve the service quality The centralized the management of telecom cloud makes it possible to manage the heterogeneousarity of resources Such as physical machines virtual machines containers and in the physical machines there was 186 a chapter and a RM architecture will also exist at the same time and in the same resource pool And for we need to achieve the Unified data governance of the multiple cloud platform and support for Custom expansion and open-on demand In order to cope this in order to cope with these problems We have made some corresponding designs for telecom cloud We established as a UCMP to achieve the goal of unified cloud resource management So we have formulated Two level data centers architecture in regional and local In regional data center, we use them to deploy regional level platform such as OSS NFVO and the UCMP. In addition, it also includes VNFM and the control plan network elements of regional level In local data center mainly on-demand deployed at the forwarding plan network elements and edge computing Access network layer elements The UCMP cooperates with OSS and NFVO enhances the cloud resource operation and maintenance in open stack and Also improves the system architecture of telecom cloud The UCMP is connecting with PIM and VIM Managing and scheduling of infrastructure unified It mainly realized the monitoring of resource fillers alarms and performance the business of Outlastation is still hand-over to NFVO And we need to complete the connections with NFVO and OSS too We should keep the consistency of communication with NFVO for the allocation of virtual resource to avoid the conflicts about the resource reuse and We also need to improve resource utilization and we will connect it to OSS in order to form a closed loop of cloud infrastructure operation and the maintenance processing. That's the important infrastructure management component UCMP is in line with the overall planning of Telecom cloud in functional design and construction and now we go to the second part practice and exploration of unified cloud management platform The UCMP is oriented to resource management operation and maintenance Unified monitoring and business support in heterogeneous match cloud environment. In this picture We can see the UCMP contains six modulars one portal provides the unified services and unified authentication Resource management take responsible for the unified management of all cloud infrastructure We can control the overall resource allocation by full view at any time And it also provides the optimal cloud resource configuration Unified monitoring to monitor the topology alarm performance capacity and other information of virtual and physical resources It can quickly be limited and locate the faults O and M provide unified operation and the maintenance and centralize the data governance CMDB saves the basic information of the resource in the full left circle and the last is adaption layer Shared with the difference of VM and PIM of different manufacturers at the bottom layer The cloud resource of various equipment access to the UCMP The implementation will be achieved step by step And then providing all functions gradually as the first step We implemented the two Components there are unified monitoring and adoption layer in terms of technology selection we built the preliminary Monitoring system of telecom cloud with a two-level architecture based on the open source promise use We built a promise use secondary level node for each telecom cloud pool Takes the responsible for resource information and data collection And we built the first level promise use master cluster to summarize all the resource pool state Also, the monitoring system provide a large screen display interface and provide the alarm information subscription interface as the entire system adopts cloud based architecture and It is built in China Unicom's shelf built Tian Gong container cloud platform The tasks of the monitoring Functions are all achieved based on open source system and components in the project adopts containerized Development and the promise use bending with two holes to achieve indicators collection query and storage Redis take responsible for log in authentication and data caching My circle achieves storing the detail this of alarms and System management information Kafka is receiving and summarized alarm information and Forwarding it to the work order system And the difficulty of UCMP design is similar to the difficulty of NFV3 layer decoupling and Both of them are facing the problem that the different date model and the interfaces during multi-bender cloud platform access we designed a multi-cloud access solution based on the adoption layer to solve this problem The principle of the adoption layer is as follows First is supporting real-time and offline data collection real-time conversion of these models and then it can be self operated and maintained and It can be upgraded separately and Finally because the sensitive of the data It must be in dependent control and achieve wonder neutrality then based on this design's principle we need to realize three point one is the collection function of heterogeneous that in the adoption layer and two the standardization of the northbound interface three the standardization of the resource data model We can see the architecture of the adoption layer from this picture it designs a set of drivers or each cloud and then adopts the cloud platform in the north direction and in the south direction and in the north direction It converts the collected data into a unified form So the option layer can not only adapt to the cloud system of the FV but also can adapt to the third party cloud system of not an FV and the driver model can also be convert to a no driver model other Management platform can obtain the original date Through the adoption layer and can also obtain the data processed by UCMP through some open interface And this this is greatly expands the application range of the adoption layer and can deliver the data from upper layer services At the same time the adoption layer can provide its own shelf operation and maintenance functions the adoption layer is constantly improving Due to the engineer reason The adoption layer is constructed first After the promissives monitoring technology stack is determined The interface is upgraded to promissives export mode From which the performance data and the resource data is pulled by the promissives masternode The alarm data interface of the adoption layer is pushed to the Kafka component in real time The third-party platform system realized the real-time processing of alarm data by subscribing to the corresponding topic in Kafka and realized the process of sending the work orders Based on the adoption layer the UCMP can get consistent cloud resource access and simple operation and maintenance management and efficient resource integration Each modular of the adoption layer can be electrically scaled according to the amount of the resources and Also support the distributed deployment The UCMP also supports flexible configuration and expansion to the indicators and Unified resource model is designed for the entire domain Open data collect functions other system can connect to the adoption layer directly and obtain the original data The efficiency of the heterogeneous cloud integration can be improved and also the efficiency of the scale expansion needs The UCMP is centrally developed deployed in regional DC to support management with new build and existing cloud resource regarding the expect effect of the cloud management platform and the first is to manage the physical and virtual resources and can allocate the Resource in unified manner and the second is to evaluate and approve the resource applications that on demand and complete the resource Isolation and the recycling The UCMP is also achieved the full process automation as the resource management and also the operation and maintenance concluence Telecom cloud is the key infrastructure for network transformation of China Unicom and It is a large-scale commercial practice of open stack in the telecom cloud and it is now moving towards cloud native gradually the development of the business The patients require a supporting platform based on cloud native and The last the UCMP is the basic platform for telecom cloud management, and it is an important the grantee for implementation of cloud native The UCMP needs to realize the collaboration of heterogeneous clouds such as open stack and the communities in the future so ultimate goal is to build Unified cloud resource operation system based on cloud native So next part my colleague the mission will introduce the related work about the cloud native in telecom cloud We divided the telecom cloud into the three development Stages the first stage is the now we are it is a virtuous machine resource poor scenario three They are kind of architecture and a two-layer deployment Complete a business deployment based on resource in this stage cloud platform and the cloud manager platform Followed up our one form introduced up on up on about combining the cloud resource managers and the system in many Provide the virtual machine for every and other services and achieve to deploy on every on cloud We get the first step from nothing Now we call for 5g use in business and then telecom cloud as the infrastructure carry 5g service need to provide more virtual function, which is the second stage It applied the container cloud based on virtual machine and supports 5g service and a little work slice and Faster deployment Meanwhile, we put the cloud native thinking into telecom cloud and promote the micro service deployment in some business Then in the third stage, we want to build a telecom cloud by containers on physical machines build a new physical container resource pool for abundant business news and It prior to the network slice in business use Which is a can achieve a second the level development of an extra expansion capability We want to do the business night and wait and the fast open Networks nights in some vertical industry and the age of computing scenario If Kcon cloud as the capability it goes to involve in cloud native And that is our one of the important to go and the direction So how to involve from virtual machine to contender and how to solve the problem in the development of a 5g Actually, 5g's SBA architecture determines the that container will become the important in telecom technology of 5g Identity and efficiency is the key advantage of a container. We should have built a application already and to the fresh check The current and research direction is based on Kubernetes and the folks are providing past capability for business application Because of the API of a community is very rich including five classes such as a workload can only service Corporation and storage mental data and the interface number is more than 150 and the closing relay related to these APS Kubernetes technology upgrades the frequency and the changes reckoning quickly and it takes at least a half a year to very fine each a version by multiple manufacturers to according to the needs of operators the Kubernetes and the contender technologies need to be continuous enhanced in the take a Technic come uses such as a native worker multiple plane TPTK Luma large page memory scheduling and so on The biggest problem is the contender complex maintenance Application driven and the contender management is more flexible into great integration A call closer make factual is more difficult operation and the maintenance interface is more different. So Contender cloud management platform is also very important The development direction of on unified the cloud management platform based on cloud native There are so key problems. We need to think about the first supporting of the cloud of the telecom business operation supporting system and the MEC money Money for business scenario the second cloud resource management enable cloud computing and the cloud native collaboration Providing the different capability of industry cloud and the private cloud The third based on the infrastructure of unified the cloud resource Achieving them a management of a resource schedule in data center In this logic infrastructure picture the that on the left a white highlight We think is the future of the telecom cloud logic education We divided the technical telecom cloud into three layers the industry cloud is on the top and The management of this day cloud management platform claims the supporting rule should be the first supporting core Network architecture of SBA achieve 5G and Microsoft is a delivery. The second is a supporting OSS Cloudy and the multiple talented resource sharing The third providing solution and the product based on MEC industry The fourth is the architect orchestration capability of providing cloud native in some special measures The middle layer is a cloud resource layer. You can see the virtual machine cloud platform based on open stack and there is a Industry cloud provide a private cloud and the container platform on the open stack and there is also a physical container platform development of independent of open stack In this is a layer. We also need to build a container platform on the other hand for the bottom layer it could be get a unified management and the unified orchestration capability of a unified cloud resource pool on data center resource In the process of cloud native or transformation of telecom cloud We should not adjust on to consider Containerization of contender elements, but also content consider the change on over company information architecture The application support the platform based on cloud native become particularly important the platform need to support all telecom this in cloud resource departure and application management and also need to consider Consider the micro service developed DevOps and the middleware China Unicom always attracts open source Open source community communities. What cloud is the infrastructure resource system? Independently developed by China Unicom, which is a base of our open stack since from launched on 2009 What cloud has been widely adopted in my in our own data centers and customers in many into the industries Well, what cloud is keeping up with the needs of the new technology and the market development and continuously and Optimizing now it has to release the half which was in the sixth combined with the new technology development needs on cloud computing Cloud native service mesh and the service it has six characterize these ticks such as comprehensive needs open Rapid delivery structure flexibility Efficiency Impensability and so on Meanwhile, we also feedback our research on open-stack community and other communities. We also share our More practice and achievements into community That's all. Thank you. Thank you