Integrated Organization and Sharing Method and Platform for AI Computing Resources Based on Digital Object

This paper proposes an integrated organization and sharing method for AI computing resources based on digital object to address the lack of unified integration and management capabilities of computing power, data, algorithm, and AI framework required for AI computing. It also tackles the difficulty of sharing, discovery, and flexible scheduling of distributed computing resources under the condition of storage-computing separation architecture. This method encapsulates AI computing resources, such as computing power, data, algorithm, and AI framework into a digital object, which comes from the Digital Object Architecture. Leveraging the interconnectivity and interoperability of digital object resources under the data network, this method enables widespread discovery and shared use. Furthermore, an AI computing and sharing platform is designed to implement and verify this method. The platform’s operation mode and application scenarios are also studied and analyzed in this paper.