GP5000 Series
GP5000 Series
The GP5000 series is a domestically produced high-performance all-flash storage. With a latency as low as 4μs, a bandwidth of 72GB/s, and an excellent performance of 16.2 million IOPS, combined with a new storage-computing separation architecture design, it fills the technical gap in the high-end storage field and becomes the preferred solution for AI training and reasoning scenarios. The GP5000 series has been successfully applied to many cutting-edge fields such as intelligent computing centers, autonomous driving research and development, and biomedical big data analysis, becoming the core engine for promoting the development of the domestic high-end storage ecosystem. Its technical indicators and implementation practices have reached the world’s first-class level, marking my country’s strategic leap from “following” to “leading” in the field of ultra-high performance storage.
As a flagship product carefully crafted by LUISUAN Technology, GP5000 is innovative in many aspects:
(1) The product adopts PCIe Switch technology. This advanced design ensures high-speed data transmission;
(2) Data access adopts RoCE protocol, which greatly reduces latency;
(3) The overall architecture adopts storage and computing separation;
A series of innovations in LUISUAN Technology not only subvert the traditional storage technology system while meeting the storage bandwidth requirements of parallel processing of massive data in the AI training process, but also greatly improves the level of domestic storage technology. It not only breaks the blockade of foreign technology systems, but also provides strong technical support for the development of AI and other high-tech fields in China, and promotes the independent control and innovative development of the domestic storage industry.
| IOPS: | System latency: | Bandwidth output: | Number of controllers: | Number of disk slots: | Capacity management: |
| 16.2 million/s | 4μs | 600Gb/s | 2 | 24 NVMe U.2 | Read 140 GB/s / Write 80 GB/s |
| AI acceleration and intelligent optimization | Multimodal data support | Sustainability and scalability | Green energy-saving technology |
| Extremely fast resource scheduling: RDMA technology is used to maximize the potential of NVMe and improve the ability to load massive data. |
Diversified data types: Support more types of unstructured data (such as video, image, audio, etc.) to improve the versatility of the storage system. Cross-modal fusion: Explore the fusion processing capabilities of different types of data to meet data needs in complex scenarios. |
Modular design: The modular architecture supports flexible expansion of the storage system and adapts to the rapid growth of data volume in the future. | Low power consumption design: Further optimize hardware design, reduce energy consumption of storage systems, and improve energy efficiency. |
