Shulou(Shulou.com)06/01 Report--

The basic steps of building a deep learning environment are very tedious. The server with GPU first needs to install OS and the application of deep learning framework, and the switch and storage system also need to be taken into account when building. In order to simplify the tedious construction process and achieve rapid deployment of deep learning environment, Infortrend Pu'an Technology launched the EonStor GSi deep learning AI storage integrated machine.

Figure 12-EonStor GSi system architecture diagram

EonStor GSi supports NVIDIA's latest Turing platform and Pascal computing platform, and uses GPU to speed up DNN computing efficiency. GSi itself is a unified storage system that supports SAN and NAS services and can reach PB-level storage space. GSi has its own Docker platform of Linux, so it is very convenient to build a deep learning environment for AI. As long as the framework application of deep learning is selected, it can be installed on the Docker platform through a user-friendly interface. Because of the lightweight nature of Docker, it is also very suitable to perform multiple deep learning applications at the same time.

EonStor GSi mounts the logic volume to the Docker platform of NAS, and the whole system is designed with modularization in hardware. The application of Docker's deep learning framework can directly access the space of NAS, eliminating the bottleneck of the intermediate switching layer, not worrying about the waiting time of the network, and reducing the difficulty of deployment and maintenance. For example, in the intelligent security application scenario, the real-time monitoring data of the IP camera can be written to the NAS space through the NAS service interface of GSi, and the deep learning framework can directly call the data to calculate and output the results to improve the timeliness. The results of edge computing can also be uploaded to the cloud in real time through the embedded cloud gateway service for further calculation and sharing.

Figure 12-Edge Computing Application scenario of Infortrend EonStor GSi Modular Design and GPU Card with EonCloud Gateway

EonStor GSi is equipped with the advanced data functions of a unified storage system, through which the results of deep learning training can be synchronized to other environments, even in the cloud. For example, the native model and historical data are copied to EonStor GSi remotely, and after the training is completed, multiple EonSotr GSi training results are synchronized to the cloud through built-in EonCloud Gateway software to do further model training or data analysis to realize the DevOps automation process of AIoT model. In terms of performance, EonStor GSi not only allows users to quickly build AI deep learning environment, but also can estimate the efficiency of production through test reports.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.