Register /
High-Performance AI Storage Solution
For vertical-industry foundation models and related applications, Sugon uses a high-performance distributed storage architecture to deliver AI Infra. It works with mainstream AI frameworks and container deployments, connects to compute clusters, and provides PB-scale elastic capacity with millisecond data response. The platform covers the full AI training and inference data path, raises cluster utilization and model-build efficiency, and improves inference experience — a low-latency, high-throughput storage base for autonomous driving, NLP, and similar workloads.
-
AI Training
Faster data throughput and model iteration.
-
Inference Service
Low-latency response for real-time service.
-
Agent Development
Smarter data management and interaction.
Solution Architecture
Connects seamlessly to compute clusters, with PB-scale elastic capacity and millisecond data response for the full AI training and inference data path
Unique Advantages
-
-
Heterogeneous Fusion
One cluster covers the full multimodal data lifecycle; 100-billion-file storage and intelligent management; a unified data pool and unified resource scheduling.
-
-
Extreme Performance
Intelligent BurstBuffer local cache; XDS acceleration shortens the I/O path via DMA, optimized for GPU Direct Storage.
-
-
Native Security
Independent software and hardware; secure and reliable clustered storage services.
Recommended Products
-
-
-
分布式存储ParaStor S6000
ParaStor S6000以1.44PB/框的超大容量和20%成本降低，通过更高的单位存储密度、更少的空间占用、更优异的性能，显著降低存储系统的构建成本。
Learn More >
-
-
-
-
分布式液冷存储ParaStor
业内首款高密度液冷存储方案 —— 曙光 ParaStor 液冷存储系统，让存储节点的 PUE 值降至 1.2 以下。可搭配曙光液冷计算节点，高效、便捷地形成存算一栈式液冷方案，将数据中心 PUE 值降至 1.1。
Learn More >
-
