EonStor GSi Family

Gaining Competitive Advantages with AI and Machine Learning

Product Highlights

Use Case

Video Clips

EonStor GSi storage family is an AI storage appliance designed for enterprises that demand integrated system and combines compute (GPU and CPU), storage (SAN and NAS), network and cloud in an all-in-one box. Essentially, GSi is the GS storage with integrated GPUs and Docker platform with GUI that can easily implement machine learning (ML) and deep learning (DL) applications. It also supports popular frameworks such as TensorFlow, Caffe and more.

Deploy AI with GSi

GSi has two series: GSi 3000 and GSi 5000.
GSi 3000 is a cost-effective inference solution best suited for AI-enhanced IoT and edge computing. GSi 5000 is a high performance solution fitting for ML/DL model training.

Implement AI Solution the Easier Way

GSi incorporates Compute (GPU and CPU), storage (SAN and NAS), network, and Cloud in an all-in-one box. With a built-in AI environment called Docker, enterprise can get into AI right away, easy implementation without complicated hardware and software setups.

Support Major AI Frameworks

GSi supports popular frameworks such Caffe, Caffe2, CNTK, MXNet, TensorFlow, and Theano, allowing organizations to implement ML/DL applications such as facial recognition, intelligent video analytics, financial modeling, medical analysis, and others.

Manage AI Applications with User Friendly GUI

GSi comes with EonOne, the central management software, as well as GUI-based Docker interface to easily develop and deploy applications. Furthermore, EonOne can manage multiple systems, monitoring GPU and capacity usage, and complete of all system-related configurations.

App Store on Docker

A Software Platform Allows User to Run and Manage Applications

The App Store features templates for various Docker container images—artificial intelligence (AI), Internet of things (IoT), video surveillance, and much more. A template specifies a Docker container image along with preset parameters of the execution environment. By using a template, you can quickly deploy and run a Docker container with little fuss.

GSi's Inherited Features from GS

Powerful NAS, SAN, and Cloud Storage Capability

GSi inherited GS line's superior storage capability and supports file-level, block-level, and cloud applications with PB-level capacity, high throughput, and data protection features. With this capability, all training dataset are saved, processed, and retrieved within the appliance for maximum performance.

PB-level Capacity

GSi supports 316/436 drives expand so no extra external storage is required

High Throughput

GSi can handle large amounts training data

Data Protection

GSi supports IDR (Intelligent Drive Recovery) and SED (Self-encrypting drives) to ensure that all training data and training models are well-preserved

Connect to the Cloud with EonCloud Gateway

GSi supports major cloud services such as Amazon S3, Microsoft Azure, OpenStack Swift, Google Cloud, Wasabi Cloud, Alibaba Cloud, Baidu Cloud and Tencent Cloud with EonCloud Gateway and can optimize data storage resources between edge and cloud. It also provides cloud backup, cloud cache and cloud tiering functions. For example, you can set all training dataset to have a backup copy in the cloud when unexpected happens.

GSi Use Case

Healthcare

In healthcare industry, AI application provides faster results in data analysis and diagnostic concerns. It helps identify the genetic information of someone with a particular medical condition. In addition, patients who don't have easy access to their healthcare providers can send their medical conditions via photographs taken by their camera phones and let the providers analyze the images with AI image analysis, a type of telemedicine. This system is faster and more effective for the providers to give near accurate diagnosis to their patients. Because of high and powerful demand in medical data process and access, GSi 5016 supports multiple GPUs with high sequential 5.4GB read/ 4.2GB write throughput performance and maximum capacity up to 436 hard drives and store 7.6 billion CT (512*512 units16 resolution) images. (12TB, RAID5)

Configuration

Training Platform: GSi 5016 with two GPUs
Process Time: 150 CT images/second at 512*512 unit16 resolution without compression (base on ResNet-50 benchmark)
Storage Capacity: 7.6 billion CT images (512*512 unit16 resolution)