Name: Final SEA Projects Workshop – A Flexible, Integrated Software Stack for HPC Systems
Start: 2024-01-16 09:30:00
End: 2024-01-16 17:00:00

Final SEA Projects Workshop – A Flexible, Integrated Software Stack for HPC Systems

by Leibniz Supercomputing Centre

Registration Link:
Final SEA Projects Workshop – A Flexible, Integrated Software Stack for HPC Systems – Leibniz Supercomputing Centre (edoobox.com)

The EuroHPC projects DEEP-SEA, IO-SEA and RED-SEA have developed exciting European technology for use in HPC systems, targeting improvements in delivered performance and energy efficiency as well as ease of use of heterogeneous systems.

We are glad to announce an open workshop on key project results, and invite all HPC practitioners to participate. In this event, the SEA projects will present and demonstrate open source software elements and best practices which can be deployed today and will bring benefits to developers of HPC applications and operators of supercomputers alike.

This workshop takes place on-site at LRZ in Garching near Munich

How to register:

This event is open to HCP practitioners, system administrators and application developers interested in the outcomes of the SEA project.

Master and PhD students are particularly invited

For organizational reasons, registration via this web page is mandatory: https://tiny.badw.de/gsgXiM
before January 5ths 2024.

Event Program

Time	Session Description	Session Leaders
9h30 – 10h00	Registration
10h00 – 10h10	Welcome : Intro to the day	HC Hoppe + Sai Narasimhamurthy + Jesús Escudero-Sahuquillo and Pedro J. García
10h10 – 11h00	The SEA project family This talk will give a short introduction to the 3 SEA projects and how they link together	HC Hoppe + Sai Narasimhamurthy + Jesús Escudero-Sahuquillo and Pedro J. García
11h00 – 11h45	DASI: a user centric data access and storage interface DASI, a Data Access and Storage Interface, enables users to manage data using domain specific and scientifically meaningful metadata keys, and separates data management from the underlying backend storage technology through abstraction. Based on the FDB object store developed and in operational use at ECMWF, we will describe its concept and how it can be easily adopted by any domain as a data management solution.	Jenny Wong (ECMWF)
11h45 – 12h30	The VEF traces framework: collecting traffic traces from modern, highly-demanding applications This framework is a set of open-source tools that allows to generate traces from applications demanding high network performance. We will describe the different tools, and how the resulting traces can be used to characterise the traffic in modern systems supported by high-performance networks, as well as how these traces can be used to feed interconnection network simulators.	Jesús Escudero-Sahuquillo and Pedro J. García (UCLM)
12h30 – 13h30	Lunch
13h30 – 14h15	Optimisation Cycles for Modular Supercomputing and Energy Efficiency Optimisation Cycles encapsulate advanced performance analysis and tuning tools to guide users through the steps required to obtain a certain analysis or optimisation result. We introduce two of these cycles: one for determining on which modules (such as CPU or GPU-based) to best run an application, and one to optimise the energy efficiency of CPU workloads at runtime.	Alexander Geiß (TUD) and Mathieu Stoffel (EVIDEN)
14h15 – 15h00	Continuous Integration of an HPC SW Stack Continuous integration is common practice in many fields of computing. We introduce its use for guaranteeing the consistency of the full HPC stack, covering applications, programming models, libraries, and system software, and provide examples on how to use open source SW to achieve this.	Manoel Roemer (FZJ) and Sonja Happ (Partec)
15h00 – 16h00	Coffee break and in // Visit of LRZ Computing Center
16h00 – 16h45	The Smart Burst Buffer: An example of an ephemeral service and its connection to the long term storage SBB, as an example of ephemeral IO service, provides Workflows with isolation from other applications, buffering and cache, improving time to solution. As an ephemeral service, it relies upon a long term storage service for which the Hestia API has been developed as an abstraction of a tiered storage infrastructure. We will introduce in this talk the IO-SEA run time architecture and the benefits workflows can get of it.	Philippe Couvée (Eviden) and James Grogan (ICHEC)
16h45 – 17h00	Joint SEA Demo	Philippe Couvée, HC Hoppe + Grégoire Pichon
17h00 – 17h15	Wrap up and Conclusion	HC Hoppe + Sai Narasimhamurthy + Jesús Escudero-Sahuquillo and Pedro J. García