In this working group, we seek to:
We do NOT aim to do any of the following:
The intrinsic generality of the workflow paradigm makes it a powerful abstraction for designing complex applications and executing them on large-scale distributed infrastructures, such as HPC centres, Grid environments, and Cloud providers. However, such generality becomes an obstacle when evaluating workflow implementations or Workflow Management Systems (WMSs), as no consistent and commonly agreed key performance metrics exist in the state-of-art computer science literature.
Instead, different application domains tend to privilege different aspects of the workflow execution process when designing their ideal workflow system. For example, minimising the control-plane overhead is fundamental when running compute-intensive workflows with billions of fine-grained tasks, while for data-intensive workflows with few giant steps overlapping computation and communication is far more prominent.
Consequently, different workflow systems excel in handling different kinds of workflows. Still, the lack of community consensus on workflow benchmarking suites represents a massive obstacle for domain experts trying to compare WMSs based on their needs. Indeed, a direct and fair comparison is possible only by running multiple state-of-art implementations of the same application on the same execution environment.
Defining several benchmarking suites to evaluate different metrics of interest would represent a crucial improvement for the workflow research community. Still, benchmarks have no value without building community consensus around them. Conversely, history tells us that highly recognised benchmarks can tremendously impact research communities, fostering a positive continuous improvement process for years. For example, think about the role of HPLinpack in the High-Performance Computing community or the ongoing efforts around mastering the training of Deep Neural Networks (DNNs) on the ImageNet dataset.
Potential performance reporting formats:
Articles below are published as Open Access, or with green open access preprints where gold open access is not possible. Please let us know if you are unable to access any of these publications. To add to this list, please suggest a change.
Tainã Coleman, Henri Casanova, Ketan Maheshwari, Loïc Pottier, Sean R. Wilkinson, Justin Wozniak, Frédéric Suter, Mallikarjun Shankar, Rafael Ferreira da Silva (2022): WfBench: Automated Generation of Scientific Workflow Benchmarks 2022 IEEE/ACM International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS), pp. 100-111 https://doi.org/10.1109/PMBS56514.2022.00014 (arXiv:2210.03170)
Elliott Slaughter, Wei Wu, Yuankun Fu, Legend Brandenburg, Nicolai Garcia, Wilhem Kautz, Emily Marx, Kaleb S. Morris, Qinglei Cao, George Bosilca, Seema Mirchandaney, Wonchan Lee, Sean Treichler, Patrick S. McCormick, Alex Aiken (2020): Task bench: a parameterized benchmark for evaluating parallel runtime performance International Conference for High Performance Computing, Networking, Storage and Analysis (SC), 62, pp. 1-15 https://doi.org/10.1109/SC41405.2020.00066 (arXiv:1908.05790)
E. Larsonneur, J. Mercier, N. Wiart, E. L. Floch, O. Delhomme and V. Meyer (2018): Evaluating Workflow Management Systems: A Bioinformatics Use Case 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 2773-2775 https://doi.org/10.1109/BIBM.2018.8621141
Tainã Coleman, Henri Casanova, Rafael Ferreira da Silva (2021): WfChef: Automated Generation of Accurate Scientific Workflow Generators 17th IEEE EScience Conference, pp. 159–168 https://doi.org/10.1109/eScience51609.2021.00026 (arXiv:2105.00129)
Daniel S. Katz, Andre Merzky, Zhao Zhang, Shantenu Jha (2016): Application skeletons: Construction and use in eScience Future Generation Computer Systems, 59, pp. 114-124 https://doi.org/10.1016/j.future.2015.10.001
Daniel Garijo, Pinar Alper, Khalid Belhajjame, Óscar Corcho, Yolanda Gil, Carole A. Goble (2014): Common motifs in scientific workflows: An empirical analysis Future Generation Computer Systems, 36, pp. 338-351 https://doi.org/10.1016/j.future.2013.09.018
Sara Migliorini, Mauro Gambini, Marcello La Rosa, Arthur H.M. ter Hofstede (2011): Pattern-Based Evaluation of Scientific Workflow Management Systems Unpublished https://eprints.qut.edu.au/216123/
L. Versluis, Roland Mathá, Sacheendra Talluri, Tim Hegeman, Radu Prodan, Ewa Deelman, Alexandru Iosup (2020): The Workflow Trace Archive: Open-Access Data From Public and Private Computing Infrastructures IEEE Transactions on Parallel and Distributed Systems, 31:9, pp. 2170-2184 https://doi.org/10.1109/TPDS.2020.2984821 (arXiv:1906.07471)
Tazro Ohta, Tomoya Tanjo, Osamu Ogasawara (2019): Accumulating computational resource usage of genomic data analysis workflow to optimize cloud computing instance selection. GigaScience, 8:4, giz052 https://doi.org/10.1093/gigascience/giz052 (bioRxiv:456756)
Tainã Coleman, Henri Casanova, Loïc Pottier, Manav Kaushik, Ewa Deelman, Rafael Ferreira da Silva (2022): WfCommons: A framework for enabling scientific workflow research and development Future Generation Computer Systems, 128, pp. 16-27 https://doi.org/10.1016/j.future.2021.09.043 (arXiv:2105.14352)
Salvador Capella-Gutierrez, Diana de la Iglesia, Juergen Haas, Analia Lourenco, José María Fernández, Dmitry Repchevsky, Christophe Dessimoz, Torsten Schwede, Cedric Notredame, Josep Ll Gelpi, Alfonso Valencia (2017): Lessons Learned: Recommendations for Establishing Critical Periodic Scientific Benchmarking bioRxiv:181677 https://doi.org/10.1101/181677
The WfBG: Workflow Benchmarking Group working group is composed of 4 members.Join Working Group