hpc

CINECA releases innovative services for interactive computing via two-factor authentication, leveraging the ICE4HPC application by E4 Computer Engineering.

CINECA Consorzio Interuniversitario, an organisation that has been supporting the research activities of the academic scientific community and offering computing services to universities in Italy for more than 50 years, recently announced the release of additional functionality for the more than 5000 users of supercomputing systems and applications

The new interactive computing service, which went into pre-production at the end of June on CINECA’s Galileo100 supercomputer, will provide users with a completely new way of accessing supercomputing resources with a capacity of over 2 Pflops and over 20 PB of data storage. Access to the computing systems will be completely via a web browser, an innovative and flexible approach in the field of supercomputing. It will thus be possible for each user to create a working environment that best suits their research needs, combining data analysis tools, visualisation, workflow creation, and other functions that will gradually be implemented in the new interactive computing service. Once logged in with the two-factor authentication system, in order to increase the level of security, resources will be accessible almost immediately, eliminating the waiting time usually required to start an interactive processing session.

“When it comes to applied research, the constant availability of applications is an important value, as is the ability to switch seamlessly from one session to another,” explains Sanzio Bassini, Director of CINECA’s SuperComputing Applications and Innovation Department (SCAI). “With the support of E4 Analytics, we have been able to provide our data scientists with a specific module, created specifically for the HPC world, which allows them to optimise their activities in terms of simplicity, functionality and effectiveness, exploiting a mode of cloud access to CINECA’s HPC architecture.“

The interface, based on the new ICE4HPC application developed by E4 Computer Engineering and E4 Analytics, provides web-based access to computing resources, while introducing Jupyter technology into the HPC environment, which enables the creation of interactive notebooks and allows users to explore data, run simulations and display results in real time. In addition, the new interactive environment integrates directly with the resource management system currently in production at CINECA (the SLURM scheduler), and enables efficient management of all requests sent to supercomputing environments. The services are designed to improve the user experience for both interactive and command-line users, prioritising usability and efficiency.

“The collaboration with CINECA is a historic one for E4, and stems from the deep harmony of views that exists between the two organisations, both focused on technological excellence,” adds Cosimo Gianfreda, CEO of E4 Computer Engineering. “CINECA researchers contribute to shaping the technological developments that will define the future, and it is an honour for us to be able to support this research activity with our application solutions that can help them operate with maximum effectiveness.“

The working environments available to scientists and researchers will include from the outset:

• Python for specific tasks (e.g. Dask, Pytorch, Tensorflow…) with the possibility of creating customised ones;
• C/C++ kernel (Xeus);
• SLURM Queue Manager (a web interface to submit and monitor jobs on CINECA-enabled systems);
• Integration with NVIDIA monitoring tools;
• VSCode interface;
• Kernel R.

“By its nature, HPC research requires a mix of high-level application resources, which researchers and data scientists are called upon to use every day, seamlessly,” adds Mario Rosati, Board Member of E4 Computer Engineering and CEO of E4 Analytics. “With ICE4HPC, we provide them with a multi-user interactive computing service for HPC clusters based on Jupyter Notebook technologies, which implements an interactive workbench capable of handling GPU computing for data analysis and visualisation, big data analytics, machine learning and deep learning.“

The revolutionary liquid cooling system for more efficient HPC servers was presented at ISC 2023 in Hamburg as part of the European project Textarossa.

At ISC 2023, E4 Computer Engineering unveiled the world premiere of the TEXTAROSSA liquid-cooled server prototype, featuring an innovative two-phase cooling system that enables HPC systems to reach new levels of performance and energy efficiency, solving a historical dilemma, that of thermal management.
Temperature is a critical issue in HPC centres housing numerous computing elements because – since a significant portion of the energy is dedicated to the cooling system – it has a direct impact on the system’s reliability and energy efficiency. The approach used in the TEXTAROSSA project combines an advanced two-phase cooling device with multi-level thermal control strategies to effectively address thermal challenges at both system and node levels. The objective of the European TEXTAROSSA project is to contribute to the goals outlined in the EuroHPC Strategic Research and Innovation Agenda and the ETP4HPC Strategic Research Agenda.

“E4 has always played a role of absolute excellence in the HPC world, with projects of international importance and high-level collaborations with research institutions and companies,” explains Cosimo Gianfreda, CEO of E4 Computer Engineering. “In TEXTAROSSA, we have pooled our advanced skills and the technological partnerships we have forged over the years to create a system that is unrivalled on the market, and which opens up unprecedented prospects for the world of research and industry thanks to its ability to maximise performance, reliability and process efficiency.”

High-performance computing, perspectives and challenges

High-Performance Computing (HPC) plays a crucial role for Countries and large companies, supporting a variety of applications in areas such as finance, weather forecasting, renewable energy research and many others. In addition, emerging sectors such as security, surveillance, bioinformatics and medicine are increasingly relying on HPC, particularly in the areas of high-performance data analysis (HPDA) and high-performance computing for artificial intelligence (HPC-AI).
To meet the needs of large-scale HPC centres, achieving high efficiency while staying within power and energy limits is a significant challenge. Researchers need to consider various factors within the HPC hardware/software combination, such as the use of specialised and highly efficient hardware accelerators, effective management of software resources and the use of advanced cooling systems to achieve optimal performance.

The TEXTAROSSA project strives to address these challenges by focusing on temperature control, energy efficiency, performance and optimal integration of new accelerators based on reconfigurability principles. The development of an innovative two-phase cooling system aims to provide an efficient and cost-effective solution to meet the needs of the current generation and prepare the appropriate technologies for the next generation of HPC infrastructure. According to the Dell’Oro Group, the use of liquid cooling in data centre thermal management represents an extremely significant growth opportunity in the physical evolution of data centre infrastructures, with forecasts of a strong increase to USD 32 billion by the end of 2026.

The TEXTAROSSA server and its cooling system

The solution outlined by E4 Computer Engineering utilises the Ampere® Mt. Collins 2U system with the Ampere® Altra® Max processor, which offers outstanding performance while maintaining impressive energy efficiency per core. This system offers several advantages, including:

• Use of ARM architecture, which aligns with the compatibility goals of the EPI project.
• Large physical space available for cooling system integration.
• Support for a significant number of PCIe slots, allowing FPGA cards or other expansion cards to be added if required.
• Optimal compatibility between cooling system design point and heat dissipation requirements.

In terms of connectivity, this versatile platform boasts 160 PCIe Gen4 lanes, enabling flexible I/O connectivity across PCIe slots, as well as 16 additional PCIe Gen4 lanes for OCP 3.0 networking. The Mt. Collins system supports up to thirty-two 3200 MT/s DDR4 DIMMs, with a maximum memory capacity of 8 TB.

ISC High Performance 2023

This liquid-cooled server prototype was presented during ISC High Performance 2023, an event dedicated to the HPC world held in Hamburg from 21 to 25 May, in which E4 participated both as an exhibitor and by contributing its contribution and know-how to numerous panels and workshops.
The event was also an opportunity to provide an update on the innovative solutions developed by the company in the fields of High Performance Computing (Medooza), High Performance Container Solution (Kaptain), GPU Artificial Intelligence Appliance (GAIA), Virtualisation (Stones), High Performance Cloud (Fluctus), and to anticipate the birth of Urania, the Cloud-Native Data Science solution, which will be presented shortly.