Compute Task Assignment via Weighted Performance Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computing systems often fail to efficiently distribute compute tasks across available computational devices, leading to underutilization of resources and suboptimal performance in terms of throughput, bandwidth, and energy consumption.
Innovation Solution
A method and system for assigning compute tasks to computational devices based on the performance characteristics of both the tasks and the devices, using a weighted graph representation of the computational workload to optimize task distribution and resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If compute tasks are distributed across multiple computational devices, then resource utilization improves, but system complexity increases
Solution Approach 1:
The patent introduces a task assignment manager as an intermediary component that coordinates between compute tasks and computational devices. This manager evaluates task characteristics, device capabilities, and performance metrics to make intelligent assignment decisions, thereby improving resource utilization while managing system complexity through centralized control logic rather than distributed complexity
Solution Approach 2:
The system dynamically adjusts assignment decisions based on changing parameters including task characteristics (compute requirements, data access patterns), device states (current workload, performance metrics), and performance weights. This parameter-based approach allows flexible adaptation to different scenarios without hardcoding complex decision logic for each case
2Productivity
If compute tasks are assigned based on detailed performance evaluation, then task execution efficiency improves, but assignment overhead increases
Solution Approach 1:
The system performs preliminary evaluation of task characteristics and device capabilities before actual task assignment. By pre-assessing compute requirements, data access patterns, and device performance metrics, the system prepares assignment candidates in advance, reducing the time needed for actual assignment decisions and minimizing assignment overhead
Solution Approach 2:
The patent replaces complex mechanical evaluation processes with performance weight-based scoring. Instead of exhaustive analysis of all task and device parameters, the system uses weighted metrics that capture essential performance characteristics, enabling faster assignment decisions while maintaining execution efficiency
3Loss of energy
If compute tasks are consolidated on fewer devices, then data transfer overhead reduces, but resource utilization decreases
Solution Approach 1:
The patent applies local quality by assigning tasks to devices based on their specific capabilities and current state rather than uniform distribution. Tasks with heavy data access requirements are assigned to devices with favorable data locality, while compute-intensive tasks go to devices with superior processing capabilities. This localized optimization reduces data transfer overhead for specific task-device pairs while maintaining high overall resource utilization
Solution Approach 2:
The system dynamically adjusts task assignments based on changing conditions including device workload, performance metrics, and task characteristics. This dynamic approach allows the system to respond to varying data locality requirements and device states, optimizing the balance between data transfer overhead and resource utilization in real-time
Data Source
AI summary
A method may include determining a first performance of a first compute task on one or more computational devices, wherein the first performance may be determined based on a first weight of the first compute task, determining a second performance of a second compute task on the one or more computational devices, and assigning, based on the first performance and the second performance, the first compute task to at least one of the one or more computational devices. The method may further include determining, based on a characteristic of the first compute task, the first weight. The characteristic of the first compute task may include at least one of a type of the first compute task, computational complexity of the first compute task, priority of the first compute task, latency of the first compute task, or amount of data used by the first compute task.