A codesign framework for online data analysis and reduction

Abstract

Science applications preparing for the exascale era are increasingly exploring in situ computations comprising of simulation-analysis-reduction pipelines coupled in-memory. Efficient composition and execution of such complex pipelines for a target platform is a codesign process that evaluates the impact and tradeoffs of various application-and system-specific parameters. In this article, we describe a toolset for automating performance studies of composed HPC applications that perform online data reduction and analysis. We describe Cheetah, a new framework for composing parametric studies on coupled applications, and Savanna, a runtime engine for orchestrating and executing campaigns of codesign experiments. This toolset facilitates understanding the impact of various factors such as process placement, synchronicity of algorithms, and storage versus compute requirements for online analysis of large data. Ultimately, we aim to create a catalog of performance results that can help scientists understand tradeoffs when designing next-generation simulations that make use of online processing techniques. We illustrate the design of Cheetah and Savanna, and present application examples that use this framework to conduct codesign studies on small clusters as well as leadership class supercomputers.

2022 Review of Data-Driven Plasma Science

“…The EFFIS [53] framework, initially designed to loosely couple multiple fusion codes running on HPC resources, is a workflow management system that uses a combination of enabling technologies, including ADIOS [54], Kepler and eSimMon [55], a web-based dashboard. EFFIS is built upon the Cheetah-Savanna [56] suite of workflow tools and provides an API for composing and executing codesign studies for online data analysis on different supercomputers. It supports both the execution of strongly coupled workflows on HPC resources [57] and the execution of data streaming from the fusion KSTAR experimental facility to NERSC [58].…”

“…A plasma treatment by the kINPen disrupts the docking and subsequently enzyme activity, strongly suggesting that the biomedical application of cold plasma may utilize the (in-) activation of proteins to achieve effectivity. Via high-resolution mass spectrometry/bioinformatics and molecular dynamics simulation (GROMACS [56] program package (version 5.0) OPLS-AA/L all-atom force field), the amino acid residue tryptophan 128 was identified to be the target of plasma-derived singlet oxygen dioxidation, yielding a ring-open kynurenine derivative that subsequently distorted the secondary structure of the C-terminal β-sheets of PLA 2 (Fig. 42) [669].…”

Application Results on Early Exascale Hardware

“…It is essential to have tools to manage concurrent runs of two or multiple applications, each of which simulates different spatial regions with varying physics principles. To enable WDMApp's multi-scale multi-physics coupled simulation runs, the CODAR team worked with scientists in the WDMApp project and developed Cheetah/Savanna [36,37] (workflow engine) and EFFIS [38] (workflow composition and execution) and provided methods for efficient data exchanges.…”

“…We reported results from this work multiple publications; physics results are in the Physics of Plasmas journal [39], while the workflow engine and the framework appeared in multiple computer science workshops and journals [36][37][38]40]. The CODAR team also worked with the WDMApp team to evaluate the performance of the XGC plasma physics simulation on Summit.…”