 Hello, I'm Jack Lee from INSPRA. I would like to introduce the new migration of machine from WMRA to Upstate. The content of my speech consists of two parts. The first part is the demand from cross-cloud migration. Describing market demand and technology trend. The second part includes the current execution of INSPRA in cloud operation system and the new migration project properties of WMRA to Upstate. Let's start with the virtualization market. WMRA is the anticipated leader in the virtualization market both in terms of functionality, performance, and stability. Many industries in China use WMRA virtual machines to host their application, such as the finance. Back to Upstate. According to IDC, the global Upstate services markets is projected to reach 8 billion by 2023 with a compound growth rate of 29% to 2017 to 2023. At present, the vast majority of mainstream cloud platform in China, such as INSPRA content, Huawei are built by Upstate. Upstate's feature, such as open source, standardized node, and the South API has also become the gathering standard of communication, finance, and other industries. It can be said that Upstate has become the DEE perspective standard for virtualized cloud platforms. In addition, with the development of Upstate platform and the QM virtualized technology and the gradual optimization and the functions and the performance, Upstate can completely match WMRA in terms of function and performance. As a result, many industries in China are gradually switching virtual machines from WMRA to Upstate. In the second part, let me introduce the online migration function of INSPRA in cloud operating system, virtual machines from WMRA to Upstate. At present, many vendors on the market provide offline migration from WMRA to Upstate, but they need to shut down for migration. Many operations will be allowed, resulting in a long business interruptions, and many operations are proven to miss operation. Therefore, we provide online one-click migration function. On the basis of shorting the business interruption process as much as possible. Provide automatic under what we saw migration. First, I will introduce INSPRA in cloud operating system. We launched the cloud strategy in 2009. We started with management of WMRA virtual machines. In 2015, we established OpenStack as a cloud. It has become a comprehensive platform with OpenStack containers, big data, cloud management, etc. In cloud OpenStack, products introduce and enhance the 24 modules in the community. And self-develop the 12 modules. At present, it has the characteristics of three high, three integration and two intelligent. In terms of high performance, we have enhanced from the underlying underlying authorization layer like QM, IO path, etc., to the platform layer, like database, message, and web access layer. After our custom performance test, we found that the performance is not weaker than WMRA. In terms of our high efficiency, last year we conducted the world's largest single class test based on the Rocky version, which was optimized in terms of large scale, high concurrency, and high density. And completed the development of 500 nodes in 1D, while managing 20,000 virtual machines and concurrent 2,000 virtual machines. No pressure on the platform. In terms of high availability, it is currently possible to achieve high availability on all platforms. The control plane is based on HEProxy and QPyli to achieve high availability of all components. And the database is based on Galera to achieve database activities. The computing plane is based on masquerade and a console to achieve multi-plane sensing. And computing load fields. Load of virtual machines can be quickly migrated to other available nodes. Within five minutes, it will compute 180 virtual machines. When key process, such as no computer field, it can be auto-medically sensed and restart. The network layer can realize the multi-plane sensing. Based on metrics judgment to determine whether the virtual machines is evacuated. Management network failures and virtual machines when business storage network is normal can operate normally. In addition, in order to ensure the large scale failure, large scale failure and the virtual machines with the business storage network is normal, can operate normally. In addition, in order to ensure the large scale failure of data center, the few metrics is self-developed and the evacuation stops. When the number of virtual machines is evacuated, failure reaches the threshold. In terms of integration, the platform can provide the virtual machines bare mental and containers with source at the same time and can manage open stack, wind-bra and public cloud environment at the same time to achieve unified management and unified scheduling. It can provide a big data cluster and AI training inference cluster with the one click. In terms of intelligency, in cloud operation system provides an AI training and a reading platform support testing flow by touch, iMac net, coffee and other AI frameworks. In addition, it introduces AI ops based on time-serial model. It can improve the time when the cluster resource will have a city which the bottlenecks. It also provides the root cause analysis when key health check and other functions. First analysis, the difficulty is our new migration. The first is a disk data reading. Some migration tour needed to install proxy software in the virtual machines and obtain this data through the proxy software. But this will cause intrusion to the user's business of virtual machines and does not met the security requirement of some customers. This is an agentless way to read the virtual machines data without marking any changes to the source of the machine. The second is the dynamic of quiz vision and disk data, disk change data because we are migration online. The original virtual machine is still running during the migration process. When you disk data may generate data, how to obtain this part of data? The third is the quiz position. Our memory catch data. During migration process, there will be data in memory catch of the original virtual machine that has not been written to the disk. How about the consistency of memory and disk data? The fourth is the consistency of IP and MAC address before and after migration. Some applications will bind the IP or MAC addresses. So the majority of the virtual machines need to be constantly with the source of virtual machines, IP and MAC address. The fifth is the device driver. Due to device compatibility issues, some operation system will not start normally due to driver types after migration. This problem must be resolved. Let's briefly understand the architecture of in-cloud manager. The in-sprout cloud manager platform. In-cloud manager is a multi-cloud management platform that can manage when we are open stack and a multi-public cloud platform through STK to achieve a unified management and a unified scheduling of resources. The first is to read the virtual machine's disk information without agent, including disk partition size and the other metadata and the disk content. When we are provided the virtual disk developer kit, SDK package, which is part of reading without an agent, the contents of the virtual disk after in-cloud manager managed in resphere, it can read the virtual machine disk information by calling the really big key interface. The second is the dynamic perception of changing data during the migration process. It also uses Wun-Wraps existing technology. Wun-Wraps provide the change block tracking technology, which provide a basis of virtual machine in create mental backup, except for the first backup, a complete backup and transmission of the entire. In addition to VM data, subsequent backup only needed to query CBT records to obtain which can change blocks. When they only backup, this changes the blocks. We will see with the time of scanning the disk and this meaning that changed the blocks and the amount of data to be transferred also greatly reduced. Therefore, we need to perform a full backup and source virtual machine disk for the first transfer. After the migration, the subsequent dynamic data is dynamically tracked through the CBT technology and transferred through the incremental backup situation. The third is the problem of data retention in the memory cache. During the migration process, the source of virtual machine is still running and a part of the data is in the memory cache and that has not been flushed to the disk. We use Wun-Wraps' silent step-short technology to make the final system of source virtual machine temporarily suspended reading and writing to the disk. During this momentary pause, the data in the operation system cache is coded to the hard disk and then combined with VDDK to read the backup data, migrate to a new cluster to realize that the migration of memory cache data and ensure the consistency of memory and the disk data. Next is the managing of the virtual machine's IP and MAC address. You should be more aware of this issue. The Wun-Wraps site opens the network card information from the FDDK including MAC and IP. The OpenStack site realize that the injection of IP and MAC addresses through cloud in it and realizes the migration of IP and MAC addresses. In cloud OpenStack, in CAP solution, this function supported the creation of virtual machines by spacing IP and MAC address and provide the UI and API interface. Of course, the premise is that the OpenStack in the network has been planted in other ones. The next step is to maintain device driver. For example, what I need to install drivers in guest or some Windows operation system are not built-in. So OpenStack need to handle disk and network card drivers. When creating a migration task, we can choose the disk interface type and network card driver type to provide the best compatibility. The default is IDE and E-1000 so you can choose the what I will type. How to migrate the data from the disk to destination virtual machines? We need to create a migration plus the virtual machine in the OpenStack cluster which needed to communicate with VMware and the OpenStack management network and can obtain the resource of the OpenStack disk for backup and incremental backup through the VDDK API interface. After creating the migration task, you need to create a disk of the same number and the size as the resource, the source of virtual machines in the OpenStack cluster and amount of the disk to migrate agent virtual machine. The proceed virtual machine reads the source of virtual machines data and write it to the code response in OpenStack volume to realize the writing of data. Ensure that the business start automatically after migrating to OpenStack can be achieved through cloud-init script, et cetera. Finally, let's sort out the entire migration process for cloud management platform in cloud management. Manage is with fear and OpenStack through information such as IP username and password. Create a migration agent virtual machines in OpenStack for data migration. Create a migration task select the virtual machines to be migrated and select the information related to the OpenStack cluster. Including information such as configuration, storage, and network. After the task is created, it will be automatically executed called the API to create a cloud disk machine machine. Create a cloud disk matches the source virtual machine in the OpenStack cluster. Mod it to the proceed virtual machines and continually copy the disk data to the OpenStack volume through full and incremental backup. While loop copying after the incremental backup size is less than a content threshold. The copying is stopped. The source of virtual machine is suspended. The OpenStack lawyer creates the virtual machines through the volume and mod the data volume then start a virtual machine and the business is started automatically complete automatically then complete the migration. That's all, thank you.