Skip to Main Content (Press Enter)

Logo UNITO
  • ×
  • Home
  • Pubblicazioni
  • Progetti
  • Persone
  • Competenze
  • Settori
  • Strutture
  • Terza Missione

UNI-FIND
Logo UNITO

|

UNI-FIND

unito.it
  • ×
  • Home
  • Pubblicazioni
  • Progetti
  • Persone
  • Competenze
  • Settori
  • Strutture
  • Terza Missione
  1. Progetti

Dynamically Managed Self-Cooling HPC Data Centers

Progetto
Data centres and IT equipment consumed an estimated 200-500 TWh of electricity in 2018, driven by the growth of computation-intensive applications like AI. Cooling, a crucial aspect of data centre operation, also contributes significantly to their electricity usage. Traditional cooling methods, relying on water-cooled chillers and cooling towers, have been common. However, with the increasing demand for AI, Big Data, High-Performance Computing, and machine learning, data centre servers require more power and generate more heat. This results in higher electricity consumption for cooling. While some have explored using adsorption chillers with low-temperature waste heat from water-cooled processors for cooling, challenges emerged when scaling this solution to large installations. The inefficiency of standard adsorption chillers and water network complexities hindered achieving the desired energy efficiency. As data centre cooling demand continues to rise, alternative solutions are needed: 1. For Large HPC supercomputers with processors and storages in different racks. In-Row Solution: Water cooled compute racks drive the adsorption chiller, which generates chilled water for the separate storage racks. 2. For Small data centres or individual racks where processors and air-cooled components are within the same rack. Rack integrated solution: The adsorption chillers are integrated into the rear door of the rack and are directly driven by a two-phase flow system cooling the processors and will directly provide air-cooling with an attached fan coil. So as first core concept DYMAN targets the development of a completely new design of adsorption chillers based on the following innovations (1) New low-temperature adsorbents achieving high capacities at very low driving temperatures below 50 °C. (2) New type of adsorption heat exchangers made of 3D printed structures integrating the adsorption material into a porous structure, which reduces the internal thermal resistances and improvement of heat transfer by two-phase flow, enhancing the heat transfer rate and reducing the internal electricity consumption of the unit. Additionally, the project aims to develop as second core concept to further develop an existing two-phase cooling system, for high-performance computing servers to handle thermal loads more efficiently from next-generation processors. Goals include increasing cooling capacities for processors generating high heat fluxes like the Nvidia H100 chip which produces 70 W/cm2. An additional objective is to recover 50% of waste heat from processors to generate additional cooling power through a adsorption chiller. Combining two-phase cooling directly with heat-powered cooling could significantly improve efficiency over conventional air or water-based cooling methods alone. The objective here is to 1) further develop the present two-phase cooling system to workin an efficient way in combination with the adsorption chiller (concept 1). 2) Development of new evaporator with new advanced surfaces for high heat transfer coefficients (3) Development of new condenser integrated with the heat adsorber of adsorption chiller. This is a crucial component that can improve the efficiency of the whole integrated system to recovery up to 50% of rejected heat. Furthermore, the cooling data centre management is a complex engineering system with interactions with different components of the data centres. Moreover, different technological solutions coexist to deal with different device types. So, DYMAN proposes a new way of active management of the data centre integrating the cooling system as part of the optimization of processor management The control of working conditions is very complex, and it has been attacked from a pure engineering point of view, being very complex and difficult to adapt to specific data centre conditions and new technologies. In this project we propose to explore a new a
  • Dati Generali
  • Aree Di Ricerca
  • Pubblicazioni

Dati Generali

Partecipanti (5)

BIRKE Robert Renè Maria   Responsabile scientifico  
ALDINUCCI Marco   Partecipante  
CANTALUPO Barbara   Partecipante  
DI GIUDA Giuseppe Martino   Partecipante  
TAGLIABUE Lavinia Chiara   Partecipante  

Referenti (2)

BARALE Claudia   Amministrativo  
MASINO Andrea   Amministrativo  

Dipartimenti coinvolti

INFORMATICA   Principale  

Tipo

HEU EIC Pathfinder

Finanziatore

EUROPEAN INNOVATION COUNCIL AND SMES EXECUTIVE AGENCY (EISMEA)
Ente Finanziatore

Partner

Università degli Studi di TORINO

Contributo Totale Ottenuto (EURO)

408.500€

Periodo di attività

Luglio 1, 2024 - Giugno 30, 2027

Durata progetto

36 mesi

Aree Di Ricerca

Settori (21)


72.19.09 - Ricerca e sviluppo sperimentale nel campo delle altre scienze naturali e dell'ingegneria

PE6_11 - Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video) - (2022)

PE6_12 - Scientific computing, simulation and modelling tools - (2022)

PE6_2 - Distributed systems, parallel computing, sensor networks, cyber-physical systems - (2022)

PE6_6 - Algorithms and complexity, distributed, parallel and network algorithms, algorithmic game theory - (2022)

PE7_1 - Control engineering - (2022)

Settore INF/01 - Informatica

CIBO, AGRICOLTURA e ALLEVAMENTI - Farmacologia Veterinaria

CULTURA, ARTE e CREATIVITA' - Culture moderne

ECONOMIA, AZIENDE E ORGANIZZAZIONI - Sistemi e metodologie per la Qualità

INFORMATICA, AUTOMAZIONE e INTELLIGENZA ARTIFICIALE - Digitalizzazione della Cultura e della Creatività

INFORMATICA, AUTOMAZIONE e INTELLIGENZA ARTIFICIALE - Digitalizzazione della Società e della Pubblica Amministrazione

INFORMATICA, AUTOMAZIONE e INTELLIGENZA ARTIFICIALE - Industria X.0

INFORMATICA, AUTOMAZIONE e INTELLIGENZA ARTIFICIALE - Salute e Informatica

LINGUE e LETTERATURA - Linguistica

PIANETA TERRA, AMBIENTE, CLIMA, ENERGIA e SOSTENIBILITA' - Diritto dell'Ambiente

PIANETA TERRA, AMBIENTE, CLIMA, ENERGIA e SOSTENIBILITA' - Informatica e Ambiente

SCIENZE DELLA VITA e FARMACOLOGIA - Tecnologie Farmaceutiche e Cosmetiche

SCIENZE MATEMATICHE, CHIMICHE, FISICHE - Fisica delle Particelle e dei Nuclei

SCIENZE MATEMATICHE, CHIMICHE, FISICHE - Laboratori innovativi, strumentazione e modellizzazione fisica

SCIENZE MATEMATICHE, CHIMICHE, FISICHE - Teorie e modelli Matematici

Parole chiave (3)

HPC
data centre cooling
two-phase cooling
No Results Found

Pubblicazioni

Pubblicazioni (2)

HPC4AI@UNITO: A Use Case For Datacenter Digital Twin 
IEEE
2025
Contributo in Atti di convegno
Altmetric disabilitato. Abilitalo su "Utilizzo dei cookie"
TabWak: A Watermark for Tabular Diffusion Models 
OPENREVIEW.NET
2025
Contributo in Atti di convegno
  • Utilizzo dei cookie

Realizzato con VIVO | Designed by Cineca | 25.4.2.0