.png){kind=icon}
.png){kind=icon}
Return
Engineering AIOps Controllers for High-Performance Software Operations: An Action Design Research Study
.png){kind=icon}
PRE
AI
DOI：10.1109/tcc.2026.3681823.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Efficient memory allocation is critical for controlling and optimizing High-Performance Computing (HPC) infrastructures. This action design research study aims to identify the best principles, practices, and patterns for improving the precision and efficiency of control and optimization mechanisms in large-scale HPC infrastructure. To achieve this, we explore the use of Artificial Intelligence Operations (AIOps), which applies artificial intelligence to manage software infrastructure operations. In this context, we propose Job Prophet (JP), a machine-learning-based application. JP is designed to regulate and continuously optimize HPC resource management, enhancing operational efficiency and performance. The tool integrates continuous training, inference, monitoring, and CI/CD pipelines, ensuring adaptability to dynamic workloads. By leveraging historical execution data from heterogeneous sources, JP enhances scheduling efficiency while maintaining model reliability through continuous retraining and monitoring. We evaluate JP on real-world HPC workloads, demonstrating significant improvements in resource utilization and job performance. Our findings highlight the benefits of AI-driven memory allocation and provide insights into scalable, automated machine-learning integration within HPC environments.
Keywords:
HPC
AIOps software
MLOps
action design research
machine learning software
Journal
I
IF:
5
Papers:
1.8K
Citations:
4.3K