Overview of the High‑Performance Computing Cloud (HPC Cloud) White Paper

On September 23, 2021, the Supercomputing Innovation Alliance held its 2020‑2021 annual meeting in Lanzhou, Gansu, and officially released the "High‑Performance Computing Cloud (HPC Cloud) White Paper" as a key outcome of the working group.

The white paper defines HPC Cloud as a service model that combines high‑performance computing (the core service) with cloud computing technologies (the innovative delivery method), extending capabilities through multi‑cloud interconnection and deep integration with big data and artificial intelligence to provide unified intelligent computing services for industry needs.

According to product form, HPC Cloud is divided into three categories: (1) Super‑computer Cloud – built on super‑computing resources and delivered via cloud service models; (2) Cloud‑Supercomputer – built on general‑purpose cloud resources to offer HPC services to multiple tenants; (3) Multi‑Cloud Interconnect – enabling resource, data, application, and service coordination across different HPC clouds with unified management.

The white paper introduces the first reference architecture for HPC Cloud, a comprehensive framework that incorporates the common capabilities of Super‑computer Cloud and Cloud‑Supercomputer as well as their distinctive features, and outlines future evolutionary paths for each.

The architecture is organized into layers: the Infrastructure layer provides the necessary compute, storage, and network resources; the Platform layer abstracts HPC resources and supplies environments for deploying high‑performance applications; the Service layer delivers unified HPC cloud services tailored to industry‑specific requirements; the Operations & Security layer offers monitoring, alerting, and protection for data and interfaces; and the Multi‑Cloud Management layer enables inter‑cloud connectivity and unified governance across resources, data, applications, and services.

Compared with traditional supercomputers, Cloud‑Supercomputer leverages cloud‑native technologies, offering advantages such as on‑demand provisioning, diverse resource types, elastic scaling, ease of use and management, trusted security, and strong connectivity, representing a new generation of HPC services that blend hardware, software, cloud‑native, and AI technologies.

Multi‑cloud interconnect is presented as a strategy that allows enterprises to link multiple private, public, and hybrid clouds, selecting optimal service combinations while reducing reliance on a single provider; it includes public‑cloud‑to‑public‑cloud, public‑to‑private cloud, and cloud‑edge interconnections to support varied workloads such as AI inference and training.

Market analysis shows that global cloud computing growth slowed to 13.1% (US$208.3 billion) due to the pandemic, with IaaS at US$59.2 billion, PaaS at US$46.3 billion, and SaaS at US$102.8 billion. Forecasts predict the HPC Cloud market will reach US$8.8 billion by 2024, growing 2.5 times faster than the offline HPC market, with a CAGR of 17.6% from 2019 to 2024. By 2024, HPC Cloud is expected to account for 18.7% of the total HPC market (US$47 billion).

The white paper also examines current industry status, cloud adoption, and technology convergence, analyzing application scenarios such as energy exploration, industrial simulation, multimedia, life sciences, weather/environmental protection, and material research, and presents typical case studies for each.

For the full white paper and related resources, readers can download the document via the provided link and explore additional references on GPU‑accelerated computing, deep learning frameworks, and supercomputing research reports.