Processor Load Balancing via Volunteer Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing methods in computer systems, such as Round-Robin and Lightest-Loaded, lead to excessive thread migration and poor scalability due to reliance on thread launch policies that amplify load imbalances and require frequent processor queue scanning, resulting in decreased system performance and throughput.
Innovation Solution
A method where processors 'volunteer' by comparing their loads to determine the volunteer processor for thread launch, updating the volunteer information if their load is lower, allowing new threads to be launched on the processor with the lightest load, thereby reducing load imbalances and improving scalability without scanning other processors' queues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Round-Robin load balancing is used, then thread distribution is simple to implement, but load imbalances are exaggerated leading to excessive thread migration
Solution Approach 1:
Each processor independently determines its own thread launch status by comparing its run queue length against a threshold, eliminating the need for centralized load balancing control and inter-processor communication
Solution Approach 2:
The system changes the thread launch decision parameter from fixed round-robin selection to dynamic threshold-based comparison, where processors with run queue length below threshold become thread launch processors
2Reliability
If Lightest-Loaded load balancing is used, then load imbalances are reduced, but system performance decreases due to frequent cache disturbances from scanning all processors
Solution Approach 1:
Processors independently evaluate their own load status using local threshold comparison rather than requiring centralized scanning of all processor queues, eliminating cache disturbances from inter-processor memory access
Solution Approach 2:
The invention extracts the load balancing decision-making process from the traditional centralized model and places it locally at each processor, removing the need for processors to scan and compare against all other processors
3Ease of manufacture
If Run-With-Parent load balancing is used, then implementation is simple and fast, but large load imbalances occur when many new threads are created
Solution Approach 1:
The system changes the thread launch decision from always using parent processor to dynamic selection based on run queue length threshold, allowing new threads to be launched on any processor meeting the threshold criterion
Data Source
AI summary
A method of and computer system for selecting a processor of a computer system on which to launch a processing thread is described. Each processor load is compared with a volunteer load of a volunteer information. If the processor load is lower than the volunteer load, the volunteer information is updated with the compared processor information. If the compared processor is the current volunteer and the compared processor load is higher than the volunteer load, the volunteer information is updated with the compared processor information.