OceanStor A600

Description

Huawei OceanStor A600 is the next-gen high-performance distributed file storage for artificial intelligence (AI), accelerating the entire process of model training and inference through extreme performance. Each controller enclosure provides up to 160 GB/s of bandwidth and 3.6 million input/output operations per second (IOPS). With innovative data-control plane separation architecture, a single cluster supports flexible expansion to 32 controllers, meeting the smooth evolution requirements of multi-modal models with trillions—or even tens of trillions—of parameters.

OceanStor A600 also supports new data paradigms, such as vectors, tensors, and key-value (KV) cache, as well as large AI model inference acceleration. With multi-level KV cache architecture, it reduces the time to first token (TTFT) by 78% and improves inference throughput by 60%. This enables long-sequence inference—where input sequence length exceeds the context window—while optimizing inference cost-effectiveness.