3BEARS: Broad Bundle of BEnchmARks for Scheduling in HPC, Big Data, and ML
University of Basel (Switzerland), CNRS and University of Bordeaux (France)
Funding agencies: Swiss Academy of Engineering Sciences (Switzerland) via the Germaine de Staël programme
The goal of the project is to develop ways to co-design parallel applications and scheduling algorithms in order to achieve high performance and optimize resource utilization. Parallel applications nowadays are a mix of High-Performance Computing (HPC), Big Data, and Machine Learning (ML) software. They show varied computational profiles, being compute-, data-, I/O-intensive, or a combination thereof. Because of the varied nature of their parallelism, their performance can degrade due to factors such as synchronization, management of parallelism, communication, and load imbalance. In this situation, scheduling has to be done with care to avoid causing new performance problems (e.g., fixing load imbalance may degrade communication performance). In this work, we concentrate explicitly on scheduling algorithms that minimize load imbalance and/or minimize communication costs. Our focus is the characterization of workloads represented by the mix of HPC, Big Data, and ML applications, in order to use them to test existing scheduling techniques and to enable the development of novel and more suitable scheduling techniques.
- Survey the state of the art in resource allocation and scheduling in order to identify the individual benchmarks and mini-apps used by their communities.
- Characterize the existing (but often not scheduling-oriented) benchmarks and mini-apps regarding their features and limitations from scheduling experiments.
- Design the specification of the 3BEARS Benchmark Suite.
- Release the first set of application benchmarks’ code, parameters, analysis, and guidelines that can be used as a starting point for the design and development of the 3BEARS Benchmark Suite.
- Dedicate the Germaine de Staël seed funding to prepare for a bi-national or European funding proposal.