Infortrend Storage enables AI deep learning at top space center
Infortrend® Technology, Inc., a leading enterprise storage manufacturer, has provided one of the top five space research organizations in the world with the EonStor GS enterprise unified storage system to enable AI deep learning in space exploration projects.
To analyze data captured by its exploration equipment, this top-five space research organization employs AI deep learning algorithms, recognizing and classifying objects, scenes, and activities for efficient content verification and tagging, which is essential for risk assessment and mission planning. In this application, high-performance storage is crucial for handling the large volumes of data generated by AI-driven processes. The organization was in search for a suitable storage system because their outdated system couldn’t meet the performance needs of AI workloads.
The organization adopted two units of Infortrend’s high performance EonStor GS 4024U U.2 NVMe hybrid flash storage and two expansion enclosures (JBODs), designated JB 3016. To maximize performance, the GS 4024U appliances connect to the network switch through a 100GbE interface. Overall, the system provides 590 TB to store space exploration data. To optimize storage efficiency, the auto-tiering function allocates data undergoing AI analysis on the high-performance SSD pool on the GS appliance and analyzed data on the cost-saving HDD pool on the JBOD. Each GS 4024U appliance can deliver up to 17 GB/s and 6 GB/s in read and write file throughput performance to meet requirements of the AI deep learning application.
“When selecting high-performance storage, our exclusive consideration was Infortrend’s system. Without it, our AI neural networks for image analysis and interpretation would not be operational,” said the Director of Data Visualization Group of Earth Observation Department at one of the world’s top five space research organizations.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
