Many-Core Trainer for Adaptive Resource Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In many core systems, accurately determining system utilization and regulating resource allocation becomes increasingly complex as the number of cores grows, leading to erroneous frequency and resource management, which can result in inefficient power usage and failure to meet real-time requirements.
Innovation Solution
A trainer system for many core systems that incorporates a synthetic variable load and probe mechanism to measure real-time requirements, allowing for the calculation of a real-time slack measurement constant and system load constant, which are used to adjust system parameters such as frequency and resource pools, ensuring accurate resource utilization and efficient power management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of cores is increased to improve system processing capacity, then productivity is improved, but device complexity increases making accurate system utilization measurement difficult
Solution Approach 1:
The patent introduces a trainer system as an intermediary component that mediates between the complex many-core system and the resource management functions. The trainer provides standardized interfaces and measurement mechanisms, allowing the system to accurately track utilization without requiring complex direct monitoring of each core's state. This intermediary layer simplifies the management complexity while maintaining high productivity.
2Ease of operation
If hysteresis-based regulation is used to control system frequency and resources, then ease of operation is improved, but measurement precision deteriorates leading to erroneous resource management
Solution Approach 1:
The patent implements a feedback mechanism where the trainer continuously monitors actual system utilization and provides this information back to the resource manager. This feedback loop enables precise measurement of core utilization, task execution patterns, and system load, allowing the hysteresis-based regulation to operate on accurate data rather than estimates, thereby maintaining both ease of operation and measurement precision.
3Reliability
If system frequency is increased to meet real-time requirements, then reliability is improved, but energy consumption increases
Solution Approach 1:
The patent employs dynamic frequency adjustment where the system frequency is continuously adapted based on real-time utilization measurements from the trainer. When system load is low, frequency is reduced to save energy; when load increases and real-time requirements approach limits, frequency is increased to maintain reliability. This dynamic approach allows the system to operate at optimal frequency points, balancing energy consumption with real-time performance requirements.
Data Source
AI summary
Disclosed herein is a trainer of a many core system comprising a plurality of cores for controlling resource utilization within the many core system, wherein the trainer comprises a controller configured to cause a task scheduler to add a first synthetic variable load (202, 302) to at least one task thread comprising at least one task and to schedule the at least one task thread; cause a generic probe element (205, 305) to set a plurality of probes configured to measure a real time requirement at a respective plurality of points within an execution of the at least one task thread; cause a training element (204, 304) to calculate a real time slack measurement constant (RS) value based on the worst case timing for each of the plurality of probes and to select at least one of the plurality of probes, wherein the selected at least one probe has a worst case RS value, wherein the training element is further configured to calculate a system load constant (SL) value based on the worst case RS value; and to cause a resource manager to adjust at least one system parameter based on the at least one probe having the worst case RS value. Also disclosed is a method of a trainer and a computer program product.