HPC Node Scheduling via Spatial Compactness Criteria

The patent changes the parameter of node selection by introducing spatial and compactness criteria. Instead of arbitrary node selection, the system evaluates nodes based on their spatial relationships and compactness metrics, transforming the scheduling approach to achieve better scalability and reliability without proportionally increasing management complexity.

The patent applies asymmetry by treating different node configurations differently based on their spatial and compactness properties. Rather than using a uniform scheduling approach, the system asymmetrically evaluates and selects node combinations based on specific geometric and topological characteristics, leading to improved system performance.

The patent applies preliminary action by pre-calculating and storing spatial relationships and compactness metrics for node combinations. This preparatory work allows the scheduling system to quickly evaluate and select optimal node assignments without performing complex computations during the actual scheduling moment, thus improving efficiency while managing computational load.

The patent segments the scheduling problem into distinct evaluation criteria: spatial relationships, compactness metrics, and availability checks. By dividing the complex scheduling decision into separable components, the system can evaluate each aspect independently and combine results, improving overall scheduling efficiency while keeping individual computational requirements manageable.