bifa必发

Extreme performance

PLStor A9600 new-gen AI storage: DataTurbo engine enables kernel-mode direct storage access/caching, cutting memory usage by 50%, delivering 160GB/s per-chassis bandwidth and over TB-level cluster bandwidth. NDS direct NPU access creates optimal path, bypassing protocols, eliminating copies for 30% performance boost.

Flexible Scalability

PLStor A9600 new-gen AI storage: new hardware architecture, starting with 2 controllers, scaling to 36 per cluster (10PB capacity) with high performance/density.

PLStor A9600 new-gen AI storage meets diverse ecosystems via file, object*, vector protocols, covering full AI workflow (data aggregation, preprocessing, training, inference) with zero data copy to accelerate model iteration.

PLStor A9600 new-gen AI storage features SmartQuota (file system quota tech), letting admins control storage for directories/users/groups, limiting usage to prevent overoccupation and maximize storage value.

Intelligent inference

PLStor A9600 new-gen AI storage: supports new data paradigms (NAS, objects, vectors, graphs) with high-performance multimodal retrieval, excelling in RAG and Unified Cache acceleration. Built-in multimodal knowledge base with leading QPS for efficiency and accuracy.

Unified Cache offers KV Cache multi-level caching, a PB-level shared pool, flexible allocation—cuts first-token latency by 78%, boosts inference throughput by 60%. Sparse attention enables ultra-large model context windows for accurate, fast, scalable central inference.