Home / Products / GP5000 series all-flash storage

data center storage

GP5000 SeriesAll-Flash Storage

A domestic high-performance all-flash storage platform with a compute-storage separation architecture, combining high throughput and ultra-low latency to reliably support core workloads in large-scale data centers.

GP5000 series all-flash storage
16 million High IOPS
72GB/s High throughput
4μs Ultra-low latency
450W Low energy consumption
Core Features

corecharacteristic

Storage and calculation separation architecture

Diskless architecture + full-stack RDMA technology reduces remote access latency to the local level.

High performance indicators

A single system delivers 16 million IOPS, 72GB/s high throughput, and 4μs ultra-low latency, with total power consumption of about 450W.

GPU direct-attached storage

Supports GPUDirect Storage (GDS), EB-level capacity can be flexibly expanded, and Checkpoint can be saved in minutes.

End-to-end NVMe

Compatible with NVMe SSD and SCM media, the controller is active-active + dual-port NVMe, safe and reliable.

Product details

Product overview

ForinnBase GroundPool 5000 (GP5000) series is a domestic high-performance all-flash storage platform launched by LUISUAN Technology. It is specially designed for intelligent computing centers, AI training inference, autonomous driving and key enterprise business scenarios.

GP5000 adopts storage and calculation separation (Diskless) architecture and full-stack RDMA technology, realizing zero copy of the network layer and kernel bypass through the RoCE protocol, reducing remote access latency to the local level. Single system provided 16 million IOPS、72GB/s throughput and 4μs ultra-low latency, the performance is dozens of times higher than that of traditional storage, and the power consumption of the whole machine is only about 450W, significantly reducing data center operating costs.

support GPUDirect Storage(GDS)and Exabyte capacityFlexible expansion, seamless integration with kilo/ten thousand card GPU clusters, enabling minute-level storage of checkpoints and efficient processing of massive unstructured data. It has been successfully used in intelligent computing centers, autonomous driving research and development, biomedical big data and other fields. With the innovative concept of "storing power into computing power", it helps enterprises break through performance bottlenecks and realize the leap from "following" to "leading" in domestic high-end storage.

Product structure

GP5000 adopts a modular architecture design, with clearly partitioned front panel disks, controllers, power supplies, and cooling systems, taking into account high-density deployment and rapid maintenance, providing a hardware foundation for long-term stable operation of the data center.

GP5000 product structure exploded view

Product features

End-to-end NVMe

It is specially designed for NVMe technology and is only compatible with NVMe SSD and NVMe SCM storage media. It has extremely low IO latency characteristics and is designed to meet users' planning and construction needs for data centers lasting 5 to 10 years.

Safe and reliable

The controller adopts an Active-Active design, and the controllers are configured with each other. Dual-port NVMe SSD is used, and combined with the storage software, the multi-pass function of data can be realized.

Management functions

It supports IPMI management and has a dedicated management LAN, which facilitates users to implement remote monitoring and management operations on the platform.

Technical Specifications

technologySpecification

parameter GP5000-2401 GP5000-2402
technology type Storage and calculation separation architecture
Maximum number of hard drives 24 pieces
Hard disk interface type Single-port NVMe U.2 SSD Dual-port NVMe U.2 SSD
Hard drive capacity 960GB / 1.92TB / 3.84TB / 7.68TB / 15.36TB / 30.72TB / 61.44TB / 128TB / 250TB
Front-end business interface Up to 6x 100GbE NVMe over RoCE v2
maximum bandwidth 600Gbps
highest throughput 72GB/s
Maximum IOPS 16 million
management port 2 x 1Gbps RJ45
power supply 2 x 800W power modules (100–240V AC, 50–60Hz)
Heat dissipation 6 high-performance fan modules
BMC Management Visual Web UI, remote management and automatic operation and maintenance, supporting IPMI / SNMP / REDFISH
health monitoring Real-time monitoring of overall machine performance, SSD health status, fan speed, temperature, and power status
System compatibility Kirin, Tongxin, Linux, Ubuntu, CentOS, SUSE, Red Hat, Windows Server
Dimensions (length × width × height) 800mm × 446mm × 89mm
Working environment temperature 5°C ~ 40°C (elevation < 1000m)
Working environment humidity 5% ~ 95% RH
Storage ambient temperature -40°C ~ 55°C