The TEXTAROSSA Approach to Thermal Control of Future HPC Systems William Fornaciari 1(B ) , Federico Terraneo 1 , Giovanni Agosta 1 , Zummo Giuseppe 2 , Luca Saraceno 2 , Giorgia Lancione 2 , Daniele Gregori 3 , and Massimo Celino 4 1 Politecnico di Milano, Milan, Italy {william.fornaciari,federico.terraneo,giovanni.agosta}@polimi.it 2 In Quattro srl, Roma, Italy {giuseppe.zummo,luca.saraceno,giorgia.lancione}@in-quattro.com 3 E4 Computer Engineering SpA, Scandiano, Italy daniele.gregori@e4company.com 4 Agenzia nazionale per le nuove tecnologie, l’energia e lo sviluppo economico sostenibile (ENEA), Roma, Italy massimo.celino@enea.it https://heaplab.deib.polimi.it, https://www.in-quattro.com, https://www.e4company.com/, https://www.enea.it Abstract. Thermal control is a key aspect of large-scale HPC centers, where a large number of computing elements is employed. Temperature is directly related to both reliability, as excessing heating of components leads to a shorter lifespan and increased fault probability, and power efficiency, since a large fragment of power is used in the cooling system itself. In this paper, we introduce the TEXTAROSSA approach to ther- mal control, which couples innovative two-phase cooling with multi-level thermal control strategies able to address thermal issues at system and node level. Keywords: High Performance Computing · 2-phase cooling · Thermal modeling and control 1 Introduction High Performance Computing is a strategic asset for countries and large com- panies alike. Such infrastructures are of key importance to support a variety of applications in domains such as oil & gas, finance, and weather forecasting. Recently, emerging domains have been gaining traction, such as bioinformat- ics, medicine, security and surveillance. These newer applications tend to fall in This work is supported in part by the EuroHPC JU and the Italian Ministry for Economic Development (MiSE) under GA 956831 “TEXTAROSSA”. c  The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 A. Orailoglu et al. (Eds.): SAMOS 2022, LNCS 13511, pp. 420–433, 2022. https://doi.org/10.1007/978-3-031-15074-6_27