Reinforcement Learning for Load-balanced Parallel Particle Tracing
This work addresses load balancing challenges in high-performance computing for fluid dynamics, ocean, and weather simulations, representing an incremental improvement through novel method integration.
The paper tackled the problem of optimizing parallel particle tracing performance in distributed-memory systems by developing an online reinforcement learning method with work donation, workload estimation, and communication cost models, resulting in improved parallel efficiency, load balance, and reduced I/O and communication costs for evaluations with up to 16,384 processors.
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.