Reinforcement Learning for Load-Balanced Parallel Particle Tracing.
Gespeichert in:
| Titel: | Reinforcement Learning for Load-Balanced Parallel Particle Tracing. |
|---|---|
| Autoren: | Xu J, Guo H, Shen HW, Raj M, Wurster SW, Peterka T |
| Quelle: | IEEE transactions on visualization and computer graphics [IEEE Trans Vis Comput Graph] 2023 Jun; Vol. 29 (6), pp. 3052-3066. Date of Electronic Publication: 2023 May 03. |
| Publikationsart: | Journal Article |
| Sprache: | English |
| Info zur Zeitschrift: | Publisher: IEEE Computer Society Country of Publication: United States NLM ID: 9891704 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1941-0506 (Electronic) Linking ISSN: 10772626 NLM ISO Abbreviation: IEEE Trans Vis Comput Graph Subsets: PubMed not MEDLINE; MEDLINE |
| Imprint Name(s): | Original Publication: New York, NY : IEEE Computer Society, c1995- |
| Abstract: | We explore an online reinforcement learning (RL) paradigm to dynamically optimize parallel particle tracing performance in distributed-memory systems. Our method combines three novel components: (1) a work donation algorithm, (2) a high-order workload estimation model, and (3) a communication cost model. First, we design an RL-based work donation algorithm. Our algorithm monitors workloads of processes and creates RL agents to donate data blocks and particles from high-workload processes to low-workload processes to minimize program execution time. The agents learn the donation strategy on the fly based on reward and cost functions designed to consider processes' workload changes and data transfer costs of donation actions. Second, we propose a workload estimation model, helping RL agents estimate the workload distribution of processes in future computations. Third, we design a communication cost model that considers both block and particle data exchange costs, helping RL agents make effective decisions with minimized communication costs. We demonstrate that our algorithm adapts to different flow behaviors in large-scale fluid dynamics, ocean, and weather simulation data. Our algorithm improves parallel particle tracing performance in terms of parallel efficiency, load balance, and costs of I/O and communication for evaluations with up to 16,384 processors. |
| Entry Date(s): | Date Created: 20220207 Date Completed: 20230504 Latest Revision: 20230504 |
| Update Code: | 20250114 |
| DOI: | 10.1109/TVCG.2022.3148745 |
| PMID: | 35130159 |
| Datenbank: | MEDLINE |
Full Text Finder 
Nájsť tento článok vo Web of Science Schreiben Sie den ersten Kommentar!