L’EMbeDS provides advanced research computing services for data-intensive analysis, simulation, statistical computing, machine learning, deep learning, and GPU-accelerated workflows. Its infrastructure combines flexible virtual machines, large-memory scheduled computing, and next-generation accelerator technologies, offering different environments for different research needs across the Department of Excellence.

L’EMbeDS virtual HPC

The L’EMbeDS virtual HPC platform provides flexible virtual machines for interactive and project-oriented research computing. Built on VMware vSphere, it supports shared CPU and GPU resources, secure access through SSH or Remote Desktop, and external connectivity through VPN. The platform is designed for machine learning, databases, software development, and exploratory parallel computing in a centrally managed environment.

The infrastructure is based on two Dell PowerEdge R740 servers with Intel Xeon Gold 6252 processors, 40 TB of SAN-backed storage, and NVIDIA Volta V100 GPUs shared through NVIDIA vGPU technology. Standard vCPU and vGPU virtual machines are available, with supported operating systems including Windows 10 Pro and Ubuntu 18.04 LTS.

Number-crunching HPC

The Number-crunching HPC platform is a Slurm-based environment designed for large-scale, batch-oriented scientific computing. It is particularly suitable for statistical analysis, simulations, and computational workloads that require very high memory capacity and structured job execution. The service runs on two Lenovo ThinkSystem servers, each with 128 CPU cores and 2.3 TB of RAM. Users can work through SSH or through a browser-based dashboard that supports job submission, job history, file management, and other core Slurm functions. The platform also integrates RStudio Server, allowing interactive statistical computing on cluster resources within a managed HPC workflow.

Next-generation GPU accelerator

The next-generation GPU platform is the newest step in the evolution of research computing at L’EMbeDS. Based on Dell PowerEdge R7725 hardware, it is designed for advanced AI workloads, deep learning, accelerated scientific computing, and other GPU-intensive applications that require stronger performance, larger memory capacity, and faster local storage than the earlier virtualized environment. The proposed configurations are based on dual AMD EPYC 9355 processors, high-capacity DDR5 memory, four 3.84 TB enterprise NVMe Gen5 SSDs, and high-speed 10/25GbE connectivity. Accelerator options include NVIDIA H200 NVL with 141 GB of HBM3e memory and NVIDIA L40S GPUs with 48 GB of GDDR6 memory. This platform is best presented as under preparation and intended for next-generation AI research, model training, and demanding scientific workflows.

Access and support

Access to L’EMbeDS computing resources is open to affiliated researchers and students through a project-based request procedure. In the established L’EMbeDS model, projects require scientific responsibility and structured access procedures, with security measures and operational practices aligned with institutional ICT requirements. Further details are available in the L’EMbeDS HPC regulation, in the Number-crunching HPC guide, and in the related access documentation.

For technical support and information please contact:
Sima Sarv Ahrabi – sima.sarvahrabi@santannapisa.it
Emanuele Guerrazzi – emanuele.guerrazzi@santannapisa.it