Multicore CPU Load Balancing via Event-Driven Threshold Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional load-balancing methods in multi-core portable terminals lead to unnecessary task migration, increasing cache invalidation, scheduling latency, and power consumption due to frequent checks for CPU load balance, which are not effectively managed by existing Linux multicore schedulers.
Innovation Solution
A method that calculates CPU utilization averages and performs load-balancing only when the average exceeds a predetermined threshold, with task migration occurring between CPUs to balance utilization, using a load-balancing policy that minimizes unnecessary migrations and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the conventional load balancer periodically checks CPU load balance and performs task migration when loads are unbalanced, then CPU utilization rates are equalized, but cache invalidation, scheduling latency, and power consumption increase
Solution Approach 1:
The invention changes the parameter for triggering load balance from periodic checking to event-driven checking (when task state changes). This parameter change eliminates unnecessary periodic checks and triggers load balance only when actually needed, reducing power consumption while maintaining CPU utilization balance
Solution Approach 2:
The invention replaces periodic load balance checking with event-driven checking. Instead of checking at fixed intervals, the system checks only when a task state changes, eliminating unnecessary periodic operations and reducing power consumption while maintaining system reliability
2Reliability
If the conventional load balancer performs task migration to equalize CPU loads, then load balance is achieved, but scheduling latency increases
Solution Approach 1:
The invention performs preliminary checks to determine if load balance is actually needed before performing task migration. By checking task state changes and evaluating whether migration is necessary, the system avoids unnecessary migration operations that would increase scheduling latency
3Reliability
If the conventional load balancer performs frequent task migration to maintain CPU load balance, then utilization rates are equalized, but cache invalidation increases
Solution Approach 1:
The invention changes the triggering parameter from periodic checks to event-driven checks based on task state changes. This reduces the frequency of load balance operations and associated cache invalidation events while maintaining CPU utilization balance
Data Source
AI summary
Methods and apparatus are provided for load-balancing in a portable terminal having a plurality of Central Processing Units (CPUs). A utilization is calculated for each of the plurality of CPUs, when a state of a task is changed. An average of the utilizations of the plurality of CPUs is calculated. It is determined whether the average exceeds a predetermined threshold. Load-balancing is performed when the average exceeds the predetermined threshold.