Power-Aware Work Stealing Scheduler for Multi-Core Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Processing systems with multiple processor cores face load imbalances and inefficiencies due to unpredictable execution times and increased power consumption during task migration in work stealing, leading to idle cores and potential thermal emergencies.
Innovation Solution
Implementing a scheduler that dynamically determines whether to steal tasks based on estimated execution time and power overhead, considering the differences in power consumption rates and thermal characteristics of processor cores, to optimize workload distribution and reduce energy usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If work stealing is used to balance load across processor cores, then execution time is reduced, but power consumption increases due to task migration
Solution Approach 1:
The patent changes the parameters of work stealing by introducing power consumption estimates and thermal state considerations. The scheduler evaluates not just execution time benefits but also power overhead and thermal impacts before deciding whether to steal work, thereby optimizing the balance between productivity and energy usage.
Solution Approach 2:
The patent implements dynamic work stealing decisions based on real-time system state. The scheduler continuously monitors power consumption rates, thermal states, and execution time estimates to adaptively determine whether work stealing is beneficial, making the system responsive to changing conditions rather than using a fixed policy.
2Productivity
If tasks are migrated between processor cores, then load balance is improved, but thermal emergencies may occur
Solution Approach 1:
The patent applies preliminary anti-action by estimating thermal state changes before executing work stealing. The scheduler predicts whether migrating a task will cause the destination core to exceed thermal thresholds and prevents the migration if it would lead to thermal emergencies, thereby proactively avoiding harmful thermal conditions.
Solution Approach 2:
The patent converts the potential harm of thermal emergencies into a benefit by using thermal state information to guide work stealing decisions. Instead of simply avoiding thermal issues, the system uses thermal state as a factor to optimize load distribution, preventing thermal problems while maintaining productivity.
3Productivity
If work stealing is performed without considering power overhead, then execution time decreases, but energy efficiency worsens
Solution Approach 1:
The patent introduces feedback mechanisms where the scheduler estimates power consumption overhead associated with work stealing and uses this information to make informed decisions. By incorporating power overhead estimates into the decision-making process, the system provides feedback on energy efficiency implications before executing work stealing operations.
Solution Approach 2:
The patent changes the parameters of work stealing evaluation by adding power consumption and energy efficiency metrics. The scheduler considers multiple parameters including execution time benefits, power overhead costs, and energy efficiency implications to determine whether work stealing should be performed, moving beyond simple execution time optimization.
Data Source
AI summary
First and second processor cores are configured to concurrently execute tasks. A scheduler is configured to schedule tasks for execution by the first and second processor cores. The first processor core is configured to selectively steal a task that was previously scheduled for execution by the second processor core based on additional power consumption incurred by migrating the task from the second processor core to the first processor core.