PFASST with dynamic resource management for large-scale applications

Dominik Huber; Hans-Joachim Bungartz; Jan Fecht; Keerthi Gaddameedi; Martin Schreiber; Martin Schulz; Tobias Neckel; Valentina Schüller

PFASST with dynamic resource management for large-scale applications

Dominik Huber, Hans-Joachim Bungartz (Editor), Jan Fecht, Keerthi Gaddameedi, Martin Schreiber (Editor), Martin Schulz (Editor), Tobias Neckel (Editor), Valentina Schüller

Research output: Contribution to conference › Poster

Abstract

Dynamic resource management is a fairly recent development made to improve the flexibility of job scheduling. MPI-based approach [1] makes this possible by treating resources on a set granularity and performs set operations whenever changes are made. Due to the extreme resource requirements, using dynamic resources will likely be compulsory for parallel-in-time methods. Development towards this has been made by prototyping an interface for dynamic resource management and then incorporating it into LibPFASST to provide proof of concept. The intent of this poster is to motivate further investigation and exploration of dynamic resources for large-scale applications such as weather/climate simulations, molecular dynamics, graph design and more.

Original language	English
Publication status	Accepted/In press - 2023 Jul
Event	12th Workshop on Parallel-in-Time Integration - Hamburg, Germany Duration: 2023 Jul 17 → 2023 Jul 21 https://www.mat.tuhh.de/veranstaltungen/pint2023/

Conference

Conference	12th Workshop on Parallel-in-Time Integration
Abbreviated title	PinT 2023
Country/Territory	Germany
City	Hamburg
Period	2023/07/17 → 2023/07/21
Internet address	https://www.mat.tuhh.de/veranstaltungen/pint2023/

Subject classification (UKÄ)

Computer Systems

Access to Document

https://mediatum.ub.tum.de/doc/1717900/6vaynue56l2cyneih2xj5c26n.pdf

Cite this

Huber, D., Bungartz, H.-J. (Ed.), Fecht, J., Gaddameedi, K., Schreiber, M. (Ed.), Schulz, M. (Ed.), Neckel, T. (Ed.), & Schüller, V. (Accepted/In press). PFASST with dynamic resource management for large-scale applications. Poster session presented at 12th Workshop on Parallel-in-Time Integration, Hamburg, Hamburg, Germany. https://mediatum.ub.tum.de/doc/1717900/6vaynue56l2cyneih2xj5c26n.pdf

@conference{a65468fdff674b78a63444062bbe6144,

title = "PFASST with dynamic resource management for large-scale applications",

abstract = "Dynamic resource management is a fairly recent development made to improve the flexibility of job scheduling. MPI-based approach [1] makes this possible by treating resources on a set granularity and performs set operations whenever changes are made. Due to the extreme resource requirements, using dynamic resources will likely be compulsory for parallel-in-time methods. Development towards this has been made by prototyping an interface for dynamic resource management and then incorporating it into LibPFASST to provide proof of concept. The intent of this poster is to motivate further investigation and exploration of dynamic resources for large-scale applications such as weather/climate simulations, molecular dynamics, graph design and more.",

author = "Dominik Huber and Hans-Joachim Bungartz and Jan Fecht and Keerthi Gaddameedi and Martin Schreiber and Martin Schulz and Tobias Neckel and Valentina Sch{\"u}ller",

year = "2023",

month = jul,

language = "English",

note = "12th Workshop on Parallel-in-Time Integration ; Conference date: 17-07-2023 Through 21-07-2023",

url = "https://www.mat.tuhh.de/veranstaltungen/pint2023/",

}

TY - CONF

T1 - PFASST with dynamic resource management for large-scale applications

AU - Huber, Dominik

AU - Fecht, Jan

AU - Gaddameedi, Keerthi

AU - Schüller, Valentina

A2 - Bungartz, Hans-Joachim

A2 - Schreiber, Martin

A2 - Schulz, Martin

A2 - Neckel, Tobias

PY - 2023/7

Y1 - 2023/7

N2 - Dynamic resource management is a fairly recent development made to improve the flexibility of job scheduling. MPI-based approach [1] makes this possible by treating resources on a set granularity and performs set operations whenever changes are made. Due to the extreme resource requirements, using dynamic resources will likely be compulsory for parallel-in-time methods. Development towards this has been made by prototyping an interface for dynamic resource management and then incorporating it into LibPFASST to provide proof of concept. The intent of this poster is to motivate further investigation and exploration of dynamic resources for large-scale applications such as weather/climate simulations, molecular dynamics, graph design and more.

AB - Dynamic resource management is a fairly recent development made to improve the flexibility of job scheduling. MPI-based approach [1] makes this possible by treating resources on a set granularity and performs set operations whenever changes are made. Due to the extreme resource requirements, using dynamic resources will likely be compulsory for parallel-in-time methods. Development towards this has been made by prototyping an interface for dynamic resource management and then incorporating it into LibPFASST to provide proof of concept. The intent of this poster is to motivate further investigation and exploration of dynamic resources for large-scale applications such as weather/climate simulations, molecular dynamics, graph design and more.

M3 - Poster

T2 - 12th Workshop on Parallel-in-Time Integration

Y2 - 17 July 2023 through 21 July 2023

ER -

PFASST with dynamic resource management for large-scale applications

Abstract

Conference

Subject classification (UKÄ)

Access to Document

Fingerprint

Cite this