The development of AI large models is making great progress, from ChatGPT for text generation to SORA for Wensheng**, and multimodal interaction continues to iterate. In the process of evolution and implementation of large models, data storage systems face performance challenges, stable operation, and cost pressures.
Sugon Storage launched Pastor distributed all-flash storage, with the industry's first technology XDS, with training acceleration, strong stability, and cost-effective unique values, covering the network, computing and platform in all dimensions, providing storage solutions for AI large model developers in thousands of industries. 
All-flash storage XDS accelerates model training by several times
In the training and inference scenarios of large AI models, Dawning Parator distributed all-flash storage embeds the Parabuffer acceleration engine with XDS technology, supports the high-speed data direct access mechanism of AI chips, improves the overall IO performance of the system by several times, and optimizes the storage and computing collaborationReduce training time from tens of days to days.
Full-stack self-developed, stable, secure, and more cost-effective
Based on full-stack self-developed product capabilities, Sugon Storage has built a four-level security system (component-node-system-solution) to reduce the impact of interruptions caused by frequent parameter tuning, server network failures, etc., and ensure the safe and stable operation of the storage system throughout the data life cycle during the development of AI large models and related applications. Based on the infinite scalability of distributed all-flash storage, the overall cost of ownership of Evergon Pastor distributed all-flash storage is more cost-effective, and the advantages of Sugon Pastor distributed all-flash storage have been fully verified in the development of AI applications in industries such as science and education, finance, and autonomous driving. Taking Suzhou Laboratory as an example, as a national laboratory, it has built a large-scale scientific data platform and introduced Pastor distributed all-flash storage to ensure that the laboratory promotes AI for Science through large models and other methods with advanced storage power with extreme performance.