HPC: Key Technologies and Tools

Scenario

The European HPC ecosystem involves three distinct pillars: Infrastructures, Applications, and Key HPC Technologies. Within each one of these three areas, important challenges may be identified, as well as different evolution directions, including classical research topics, extreme scale specific requirements and education&training. Furthermore, as confirmed by the most recent HiPEAC Vision document in addition to the ETP4HPC SRA, the whole HPC scenario is going to be redefined due to the convergence of traditional HPC (i.e., computationally intensive applications such as physical or chemical simulations) with the processing and storage of vast quantities of data. 

 

 

Positioning

In a scenario—both European and Italian—where HPC public investments have been mostly directed to the Infrastructures and Applications pillars, we focus on the importance of the Key HPC Technologies pillar that is being identified in the forthcoming European HPC project calls as the main area to investigate (e.g., in the EuroHPC-JU program) and where several groups in Italy contributed with notable results and research activities. Different areas of interest—with specific expertise in this group—are relevant to the Italian key HPC technology initiatives including: 

  • Energy efficiency, both in terms of hardware (e.g. exploitation of specialized heterogeneous accelerators—most likely reconfigurable—and/or close-to-memory computing) and software (e.g. adaptivity and self-tuning tools and mechanisms supporting the performance/energy tradeoff) 

  • Programmability, including all the software stack evolutions related to programming models and compilers, primarily addressing the performance/energy trade-off, capable of dealing with large data and providing suitable tools to manage reconfigurable hardware. 

 

Target systems

The working group plans to address research subjects in the area of key HPC technologies in the scenario of pre-exascale and exascale systems and also in the context of smaller scale systems that are often in the availability of research institutions, big industries and SME consortiums. This because we are aware of the possibility and convenience to apply different scales of the very same software environment, tools and technologies when targeting different scale of HPC systems, and because smaller scale parallel systems in terms of market share  and of interest for academic and industrial research may be considered the real cornerstone of HPC value chain. The working group does not exclude to consider embedded systems and distributed/cloud architectures when key technologies and tools for parallelism will be involved.  

 

Specific “management” activities

The working group plans to develop specific activities including:  

  • Technology transfer and creation of awareness activities; 

  • Creation of a map of the competences with the goal to enable the proposing of an HPClab as a unique partner in research projects; 

  • Participation and organization of international events and the active presence in the monitoring of the HPC initiative at EU-level, including the brokerage events; 

  • Creation of a web site to be linked to the overall set of CINI Labs and design of other material for the dissemination of our initiatives and competences; 

  • Establishing specific initiatives related to education, targeting all high education levels, including masters and PhD; 

  • Setting up a suitable internal management structure, including a responsible for each one of the research groups participating in the working group and some kind of “steering committee”. 

 

Relation with other HPC pillars

The working group is established in the perspective of being complementary and synergistic with respect to both the groups and the activities more directly related to the Infrastructure and the Application pillars. We foresee the possibility to identify synergies in the co-design or co-tuning of the different scales of HPC infrastructures as well as a very to exploit interesting perspectives in reciprocal contamination among techniques developed to design and deploy domain specific parallel HPC applications and the key HPC technology assessments.  

 

Positioning within CINI

The working group aims at interacting with existing working groups and labs in CINI that attack the HPC scenario from more specific and application dependent perspectives (AI, Big Data, Machine learning, Bioinformatics, Medical applications, Cybersecurity, Smart cities, IoT, etc.) in the perspective of collecting new challenges to be investigated in the key HPC technology scenario and of providing parallel expertise in the specific domain areas. The working group may obviously contribute to improve the quality of the vision and of the impact of CINI in the general HPC scenario.  

 

Share This

S5 Box

Cini Single Sign ON

Questo sito utilizza cookie, anche di terze parti, per migliorare la tua esperienza e offrire servizi in linea con le tue preferenze. Chiudendo questo banner, acconsenti all’uso dei cookie. Se vuoi saperne di più o negare il consenso a tutti o ad alcuni cookie vai alla sezione Cookie Policy