After cloud computing first clarified its academic definition in 1997, a series of hardware-assisted virtualization technologies were born after 2005 to solve core problems such as large-scale resource virtualization and pooled resource on-demand elastic services. Cloud computing is used in the field of general computing scenarios. Rapid implementation has become the infrastructure of the new generation of information technology. However, cloud computing systems cannot meet the strong demands of low-latency application scenarios in terms of theory, methods and technology. The latency of cloud computing systems mainly comes from three aspects: the native virtualization capabilities of heterogeneous resources; hierarchical superposition of software stacks and resource competition in the internal virtualization environment of cloud nodes; and data partitioning and data migration overhead in cross-node computing environments. After more than ten years of research, the project team has achieved the following technical inventions in terms of low-latency methods and key technologies for cloud computing systems: 1. Heterogeneous hardware resource virtualization method: A method for decoupling resource authorization and resource usage is proposed, which overcomes the lack of native virtualization capabilities of new low-latency heterogeneous hardware such as GPU, NVM, SGX and TrustZone, and innovates a heterogeneous virtualization method that balances sharing, isolation and latency., is the first in the industry to provide virtualization capabilities for these hardware and enable them in cloud computing systems. 2. Low-latency single-node virtualization architecture: Invented a virtualization method that links up and down across levels, solved the problem of resource scheduling and disk write operation amplification caused by the superposition of vertical virtualization levels, and invented a delay-aware multi-core read-write synchronization mechanism to eliminate synchronization overhead during the read process. 3. Low-latency cross-node virtual resource aggregation method: Invented application feature-driven differentiated data partitioning method and storage mode, breaking through the traditional ldquo; one-size-fits-all design idea, overcoming unbalanced data partitioning such as graphs and high network communication overhead, etc. Problems such as problems; Invented a cross-node in-situ computing model based on RDMA ldquo; Mobile data rdquo;, breaking through the traditional thinking paradigm of ldquo; Mobile computing rather than data rdquo;; Invented low-latency cross-node high availability methods such as virtual machine hot standby based on ldquo; output consistency The project has obtained 31 domestic and foreign invention patent authorizations (including 3 U.S. authorized patents), registered 6 sets of software copyrights, and published 58 papers. The above invention supports mainstream cloud computing platforms running Xen, KVM, etc. with open source software;GPU virtualization methods have been integrated into the Linux 4.10 distribution (and became the first of nine Prominent Features in this version) and Intel gVirt open source virtualization solution; the delay-aware multi-core read and write synchronization mechanism has been integrated into mainstream open source virtualization software such as Xen, and the low-latency memory management method has been adopted by the famous Linux distribution ArchLinux; the virtual machine hot backup method COLO is adopted by mainstream virtualization open source software Xen and KVM; Application feature-driven differentiated data partitioning methods have been used by platforms such as WeChat for real-time data analysis of hundreds of millions of users. The project results have also been transformed into Huawei's cloud computing products and Tencent's social platforms through school-enterprise cooperation and other forms, significantly reducing the latency of the cloud computing system. In the past three years, the new output value has been 2.34 billion yuan and the new profit has been 936 million yuan, achieving outstanding economic and social benefits. Project members have won talent programs and science and technology awards such as Changjiang Distinguished Professor of the Ministry of Education, National Outstanding Young Talents, LDquo; Ten Thousand Talents Planrdquo; Young Talents, CCF Young Scientist Award, Outstanding Academic Leaders of Shanghai City, and Young Science and Technology Star of Shanghai City, as well as Best Paper Awards and Nomination Awards at famous international conferences such as EuroSys, APSys, and HPCA. The project has also cultivated talents including CCF Youbo and Electronics Society Youbo.