DapuStor Enterprise SCM - Haishen3-XL 

As machine learning, artificial intelligence, and real-time data analysis are widely deployed, real-time data processing becomes more and more important. How to find the best balance between key factors such as performance, price, and capacity remains a challenge for the storage industry. DapuStor introduces a new Storage Class Memory (SCM) product based on the latest XL-Flash, which provides extremely low latency, long life and ultra-high performance for great product experience.



Data Cache and Acceleration

In-memory Database

AI Training

Big Data

  Key Features:

Ultra-low Latency with High Performance

Adopting industry leading XL-Flash and DapuStor patented Smart-IO technology, Haishen3-XL exhibits 20μs ultra-low read latency and up to 3.5GB/3.2GB bandwidth, 830K/350K IOPS at low queue depth workloads.



Industry Leading Endurance

Ultra-high performance scenarios usually require high endurance. Based on the optimized SLC-class Fast NAND, Haishen3-XL delivers very high endurance with up to 30DWPD.

  Product Spec:
Model NoH3900
Capacity (TB1)0.751.6
OutlineU.2 & HHHL
Interface ProtocolPCIe3.0 x 4 NVMe 1.3
Flash Type96L 3D XL-FLASH
Read Bandwidth (128KB) MB/s235003500
Write Bandwidth (128KB) MB/s231003200
Random Read (4KB) KIOPS3830830
Random Write (4KB) KIOPS3350288
Power Consumption4 (Typ./Max) Watt7.0/8.57.0/9.5
Lifespan30 DWPD6
4K Random Latency (Typ. ) R/W μs5   30/177 | 20/098
Uncorrectable Bit Error Rate (UBER)<10-17
Mean Time Between Failure (MTBF)2 million hours
Supported OSRHEL, SLES, CentOS, Ubuntu, Windows Server, VMware ESXi
CertificationFCC, CE, ROHS, REACH, WEEE, PCI express, NVM express


       1. TB = 1,000, 000, 000,000Byte.

       2. MB/s = 1,000,000 Bytes per second. Performance is measured with 1 worker*Queue Depth 256*128KB (131,072 bytes) for sequential read and write bandwidth on LINUX*Cent OS 7.5 by using FIO.

       3. Performance is measured with 4 workers*Queue Depth 64*4KB (4096 bytes) for random read and write IOPS on LINUX*Cent OS 7.5 by using FIO.

       4. Typical power Consumption is measured with 4 workers*Queue Depth 64*4096 bytes for random read/write ratio (7/3); Max power Consumption is measured with 1 worker* Queue Depth 128*131,072 bytes for sequential write.

       5. μs = Microsecond. Performance is measured with 1 worker *Queue Depth 1*4KB (4096 bytes) for random/ sequential read and write average latency on LINUX*Cent OS 7.5 by using FIO.

       6. DWPD: Drive Writes Per Day, last for 5 years.

       7.This latency is measured by Linux Fio libaio io engine.

       8. This latency is measured by Linux SPDK 18.01 .

  Product Pictures