High-performance computers with ultra-high computing performance are widely used in various fields of the national economy, in response to major challenging problems, to promote the transformation and upgrading of traditional industries, and to improve people’s living standards. High-performance computing cloud has become an innovative computing power service model. It focuses on improving the computing power experience of industrial users, solving the pain point problems faced by traditional supercomputing, such as large asset investment, and providing broader application prospects for general supercomputing and industrial supercomputing.
Under the background of the development of high-performance computing cloud, the deep integration of edge computing and high-performance computing, on the one hand, the cloud-based high-performance computing services from the centre to the edge, which can further satisfy the demand for high-efficiency arithmetic in more business scenarios, and on the other hand, combined with the edge of the cloud synergistic ability, to achieve the expansion of high-performance computing cloud service capacity support.
Application Scenario Example: Multimedia Rendering Scenario
Multimedia rendering refers to the process of generating images from models using software. A model is a description of a three-dimensional object in a strictly defined language or data structure that includes geometry, viewpoints, textures, and lighting information.
Rendering has become more and more widely used in recent years as market demand grows and computer hardware and graphics develop rapidly. Animation rendering consumes a lot of time has become an important factor affecting the efficiency of animation production, rendering in the animation programme production process requires more and more time; and today’s film and television post-production has been inseparable from the support of the rendering of special effects, post-production departments to face the challenges of the audience’s increasingly demanding, the artists not only need to pursue innovation in each project and need to solve the problem of realizing the complexity of the special effects of the fast! Artists not only need to innovate on every project, but they also need tools to solve the problem of realising complex effects quickly; even the most beautiful effects have no market value if they can’t be delivered by the end of the delivery period.
Rendering with traditional PC workstations is simply not enough to cope with the sheer volume of computation. Therefore, the introduction of edge HPC technology into the rendering industry to build a rendering system has become a necessary means to solve the bottleneck of animation and film production and improve work efficiency.
The edge high-performance computing architecture mainly includes four parts: edge infrastructure layer, HPC platform layer, application layer and multi-cloud interconnection:
Edge Infrastructure Layer: The HPC platform provides the basic arithmetic resources required to run industry applications, and can provide virtualisation capabilities upwards. The hardware resources in the infrastructure layer mainly include general-purpose servers, storage devices, network devices, supercomputing servers, high-speed interconnection networks, GPUs, FPGAs, TPUs, and other types of computing, storage, and network hardware, of which supercomputing cloud is dominated by supercomputing resources and cloud supercomputing is dominated by general-purpose resources. Resource virtualisation provides centralised resource management, elastic resource deployment, and low operation and maintenance costs for HPC cloud services.
HPC Platform Layer: Provides job scheduling, high-performance algorithm library, performance monitoring, cluster management and other related functions, and uses a unified task scheduling system to achieve unified scheduling and orchestration of edge computing resources and jobs.
Application Layer: A portal for industry users, providing HPC application services in a variety of industry scenarios in addition to basic platform services, including intelligent analysis, energy survey, design simulation, gene sequencing, image rendering, and so on.
Multi-cloud Interconnection: It mainly includes mutual collaboration among HPC public cloud, HPC private cloud, and HPC edge cloud at four levels of resources, data, applications and services, and on this basis, it realises multi-cloud management functions, such as multi-cloud access, heterogeneous management, and resource management.
Source: Network Convergence Industry and Standards Promotion Committee.