Hiroyuki Takizawa

Muhammad Alfian Amrizal, Raka Satya Prasasta, Santana Yuda Pradata et al.

High-performance computing (HPC) systems consume enormous amounts of energy, with idle nodes as a major source of energy waste. Powering down idle nodes can mitigate this problem, but long boot/shutdown delays can introduce significant queueing penalties if transitions are poorly timed. To address this trade-off, we present SPARS, a reinforcement learning-enabled simulator for power management in HPC job scheduling. SPARS integrates job scheduling and node power-state management within a discrete-event simulation framework. It supports traditional scheduling policies such as First Come First Serve and EASY Backfilling, along with enhanced variants that employ reinforcement learning agents to dynamically decide when nodes should be powered on or off. Users can configure workloads and platforms in JSON format, specifying job arrivals, execution times, node power models, and transition delays. The simulator records comprehensive metrics-including energy usage, wasted power, job waiting times, and node utilization-and provides Gantt chart visualizations to analyze scheduling dynamics and power transitions. Unlike widely used Batsim-based frameworks that rely on heavy inter-process communication, SPARS provides lightweight event handling and consistent simulation results, making experiments easier to reproduce and extend. Its modular design allows new scheduling heuristics or learning algorithms to be integrated with minimal effort. By providing a flexible, reproducible, and extensible platform, SPARS enables researchers and practitioners to systematically evaluate power-aware scheduling strategies, explore the trade-offs between energy efficiency and performance, and accelerate the development of sustainable HPC operations.

Toyotaro Suzumura, Akiyoshi Sugiki, Hiroyuki Takizawa et al.

The growing amount of data and advances in data science have created a need for a new kind of cloud platform that provides users with flexibility, strong security, and the ability to couple with supercomputers and edge devices through high-performance networks. We have built such a nation-wide cloud platform, called "mdx" to meet this need. The mdx platform's virtualization service, jointly operated by 9 national universities and 2 national research institutes in Japan, launched in 2021, and more features are in development. Currently mdx is used by researchers in a wide variety of domains, including materials informatics, geo-spatial information science, life science, astronomical science, economics, social science, and computer science. This paper provides an the overview of the mdx platform, details the motivation for its development, reports its current status, and outlines its future plans.

Thomas Budiarjo, Santana Yuda Pradata, Kadek Gemilang Santiyuda et al.

High energy consumption remains a key challenge in high-performance computing (HPC) systems, which often feature hundreds or thousands of nodes drawing substantial power even in idle or standby modes. Although powering down unused nodes can improve energy efficiency, choosing the wrong time to do so can degrade quality of service by delaying job execution. Machine learning, in particular reinforcement learning (RL), has shown promise in determining optimal times to switch nodes on or off. In this study, we enhance the performance of a deep reinforcement learning (DRL) agent for HPC power management by integrating curriculum learning (CL), a training approach that introduces tasks with gradually increasing difficulty. Using the Batsim-py simulation framework, we compare the proposed CL-based agent to both a baseline DRL method (without CL) and the conventional fixed-time timeout strategy. Experimental results confirm that an easy-to-hard curriculum outperforms other training orders in terms of reducing wasted energy usage. The best agent achieves a 3.73% energy reduction over the baseline DRL method and a 4.66% improvement compared to the best timeout configuration (shutdown every 15 minutes of idle time). In addition, it reduces average job waiting time by 9.24% and maintains a higher job-filling rate, indicating more effective resource utilization. Sensitivity tests across various switch-on durations, power levels, and cluster sizes further reveal the agent's adaptability to changing system parameters without retraining. These findings demonstrate that curriculum learning can significantly improve DRL-based power management in HPC, balancing energy savings, quality of service, and robustness to diverse configurations.