Research on Key Technologies of Computing Power Networks for Scientific Computing

The increasing demand for large-scale deployment of extensive data, driven by technologies like the internet of Things, digital twins, and LLMs, alongside the development of data centers, challenges the conventional “best effort” network model, necessitating a more sophisticated approach. This article explores the impediments encountered in scientific computing scenarios, including computing and storage transmission barriers, uncertain resource status, and coordination and scheduling complexities. It delves into pivotal technologies such as integrated converged networks, resource prediction and scheduling, and multi-cloud unified management to enhance the computing network architecture, offering insights into the applications of computing power networks.