The “ready-to-go” GPU appliance
GAIA is a powerful solution that exclusively uses top-of-the range hardware: multi-core CPUs, up to 2TB of RAM, NVMe drives and high speed interconnected multi-GPUs.
The end-user can access a series of containerised workspaces for the development of Data Analytics, Machine Learning and Deep Learning applications while using better underlying hardware. Additionally, GAIA uses leading open-source frameworks (Rapids.AI, Tensorflow, PyTorch, MXnet, HuggingFace…).
Engineered, not assembled
GAIA produces an easy-to-use interactive multiuser interface (based on notebook technology).
GAIA is an open solution, making it easy to integrate third-party containers (Nvidia, DockerHub…). Moreover, with regular additions and updates delivered by E4 Analytics, GAIA will always remain state of the art for Data Science.
Ready for Generative AI
GAIA is an out-of-the-box solution for training, fine-tuning and inferencing popular open-source Large Language Models (LLMs) such as LLAMA2 and FALCON.
Due to the flexibility of its architecture, GAIA can integrate LLM-based services – local or remote (ChatGPT) – into Jupyter Notebooks, for example, to generate new code.
The GAIA approach
GAIA (GPU Artificial Intelligence Appliance) is the solution for enabling scalable and high performance Machine Learning and Deep learning workloads.
GAIA was designed by our professionals at E4 Analytics to help data scientists maximise their productivity in collecting, cleaning up and transforming company data, implementing and testing the most advanced Machine Learning algorithms, and exploiting the results of Big Data analyses for the benefit of their company.
POWERFUL
The foundation of GAIA’s hardware is a super server which can integrate up to a 64 core CPU, up to 2TB of RAM, fast and large storage devices based on NVMe technology and up to 8 interconnected all-to-all GPUs.
CONTAINERIZED
Workspaces are always implemented across container images: the end user has access to numerous workspaces, designed to perform optimally at each stage of the typical data science workflow, without compromising performance.
READY TO USE
GAIA is a “ready-to-use” solution that integrates a multi-user web interface based on Jupyter Notebook, which incorporates code with descriptive text, formulae, graphical visualisation, and multimedia content, interactively.
FUTURE PROOF
The world of Data Science is in a constant state of change. Like all services offered by E4 analytics, GAIA’s stack software is regularly renewed and upgraded with the most innovative updates from open source.
Designed to be the best: always
Advanced Analytics and Artificial Intelligence: decision support systems.
Business Intelligence: understanding what happened in the past.
Artificial Intelligence: making predictions and generating recommendations for the future.
In these technologies, computers are trained to perform specific tasks through the processing of large quantities of data and the identification of models present within these data. This requires both high performance server accelerators (GPU and FPGA), as well as networking and storage components.
Discover the advantages
Solution Layout
Technical features
GAIA consists of two main modules: the infrastructure (Kaptain single node) that provides high-performance container orchestration services, and the platform (E4AI-PLATFORM) that enables services and functionalities to maximize the productivity of data scientists.
Kaptain single node
Kaptain single node is the stand-alone Kubernetes solution designed for computationally intensive workloads, which includes a Web UI.
E4AI-PLATFORM
E4AI-PLATFORM is a software stack which integrates all the necessary components to implement a full AI workflow.
- Identity access management system: user authentication and management with 2FA option
- Object storage: service for high-performance storage accessible with S3 protocol
- Cloud native batch scheduler: batch scheduling system for computationally intensive workloads
- ICE4AI: ambiente di lavoro interattivo multi-utente basato sui Jupyter Notebook
- Data Services: set of services needed to implement and automate AI workflows
ICE4AI
- ICE4AI is an Interactive Computing Environment for development in Python and beyond
- ICE4AI is based on Jupyter Notebook technology
- ICE4AI is designed and configured for Data Analysis, Machine Learning and Deep Learning
- ICE4AI can be configured to offer end-user dedicated computing resources
- ICE4AI is easily integrated with some NVIDIA containers (NGC)
Architectural advantages
HIGH PERFORMANCE APPLIANCE
GAIA is a ready-to-use, high performance appliance, used for the development, testing and deployment of scalable Data Analytics, Machine Learning and Deep Learning applications. It is the ideal solution for companies that want to extract the most value from their data.
VERSATILE
GAIA enables the user to have different integrated versions of each workspace online and allows them to make additional customised versions if required.
OPEN SOURCE
GAIA exclusively uses open source technology developed by the most relevant and active Data Science communities
SCALABLE
GAIA‘s architecture responds to growing demands for better computing resources, enabling the creation of a cluster consisting of multiple GAIAs, functioning as worker nodes.
Why choose this E4 solution
READY-TO-USE
A ready-to-use appliance equipped with a multi-user Interactive Computing Environment made for Data Science
VALIDATED
Performance tests are carried out on all nodes before the solution is released. In addition to the usual firmware, homogeneity, sanity and setup check, we use additional tools to verify whether performance levels correspond to those requested by the customer. Relevant tests include: HPL (High Performance Linpack) to test machine’s computing power, measured in FLOPs; STREAM to test memory’s bandwidth, measured in MB/s; and IOzone to test disk’s access speed, measured in MB/s and IOPS.
TESTED
According to a protocol developed by E4, each component undergoes a burn-in test for up to 120hrs to ensures properly engineered and functioning systems. This procedure reduces DoA (Dead on Arrival), decreases “early failure” rate and highly improves the overall reliability of E4 solutions.
SERVICED
E4 is one of the few companies currently providing services at the highest level in large academic and private infrastructures as well as in national and international research centres of great complexity and relevance.
System support and solution customization
• per day
• in packages of “x” days on a pay-as-you-go basis
• per project
*minimum billable 1/2 day
Data science consulting
• per day
• in packages of “x” days on a pay-as-you-go basis
• per project
*minimum billable 1/2 day
Functional training on the E4AI-Platform environment
Extra | New functionalities coming soon
NEW FUNCTIONALITIES
• GAIA will soon directly integrate several NVIDIA containers (NGC)
• GAIA enterprise based on E4 VSTONE
GPU ARTIFICIAL INTELLIGENCE APPLIANCE
