Integrated Data Analysis Pipelines for Large-Scale Data Management, HPC, and Machine Learning
Coordinator: Know-Center GmbH, Austria
Partners: 13 partners from 7 European countries
Know-Center GmbH (Austria), AVL List GmbH (Austria), Deutsches Zentrum für Luft- und Raumfahrt EV (Germany), ETHZ (Switzerland), Hasso Pattner Institute (Germany), ICCS (Greece), Infineon Technologies (Austria), Intel (Poland), IT University Copenhagen (Denmark), KAI GmbH (Austria), TU Dresden (Germany), University of Maribor (Slovenia), University of Basel (Switzerland).
Central project webpage: daphne-eu.eu
The DAPHNE project aims to define and build an open and extensible system infrastructure for integrated data analysis pipelines, including data management and processing, high-performance computing (HPC), and machine learning (ML) training and scoring. Key observations are that:
(1) systems of these areas share many compilation and runtime techniques,
(2) there is a trend towards complex data analysis pipelines that combine these systems, and
(3) the used, increasingly heterogeneous, hardware infrastructure converges as well.
Yet, the programming paradigms, cluster resource management, as well as data formats and representations differ substantially. Therefore, this project aims – with a joint consortium of experts from the data management, ML systems, and HPC communities – to systematically investigating the necessary system infrastructure, language abstractions, compilation, and runtime techniques, as well as systems and tools necessary to increase the productivity when building such data analysis pipelines, and eliminating unnecessary performance bottlenecks.
At the University of Basel
Throughout the hierarchy from integrated pipelines in distributed environments down to specialized storage and accelerator devices, scheduling of tasks (which may represent operation- and data-bundles) is crucial for achieving high system utilization, throughput, and latency.
N. Ihde, P. Marten, A. Eleliemy, G. Poerwawinata, P. Silva, I. Tolovski, F. M. Ciorba, T. Rabl. “A Survey of Big Data, HPC and Machine Learning Benchmarks”. In Proceedings of the 13th Transaction Processing Council Technology Conference on Performance Evaluation & Benchmarking (TPCTC 2021) of the 47th International Conference on Very Large Data Bases, Copenhagen, Denmark, August 2021 online [C62.bib]
J.H. Müller Korndörfer, A. Eleliemy, A. Mohammed, F.M. Ciorba. “LB4OMP: A Dynamic Load Balancing Library for Multithreaded Applications”. Transactions on Parallel and Distributed Systems (TPDS), Tier A*, 2021. [to appear] Videos about LB4OMP (OpenMP booth video, OpenMP Users Developer Conference video)
P. Damme, et al. “DAPHNE: An Open and Extensible System Infrastructure for Integrated Data Analysis Pipelines”. In Proceedings of the 12th Annual Conference on Innovative Data Systems Research (CIDR ’22), Chaminade, USA., January 2022 online [C65.bib]
Multilevel Scheduling Reflection on DAPHNE
Speaker: A. Eleliemy
Invited talk at EVEREST+DAPHNE: Workshop on Design and Programming High-performance, distributed, reconfigurable and heterogeneous platforms for extreme-scale analytics, held at the HiPEAC 2023 Conference, January 18. 2023.