Proactive Load Prediction for Computing Device Software Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load distribution methods in computing-device systems become overly complex due to reliance on application software state information, making real-time load balancing difficult.
Innovation Solution
A computing-device management device predicts future loads based on CPU load, memory usage, and data communication volume, allowing for proactive allocation of software components to prevent overload and ensure redundancy, using a mathematical model like Lagrange interpolation polynomials for precise predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If load distribution is carried out on the basis of application software information (process time, process start time, process due), then load distribution can be performed, but the process becomes too complex to carry out in real-time
Solution Approach 1:
The patent extracts the essential load characteristics from complex application software information by focusing only on three key parameters: CPU load, memory used volume, and data communication volume. This extraction simplifies the monitoring and prediction process while retaining the critical information needed for effective load distribution, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent changes the parameters used for load monitoring from detailed application software metrics (process time, start time, due time) to fundamental system resource metrics (CPU load, memory usage, data communication volume). This parameter transformation simplifies the data collection and processing required for real-time load prediction while maintaining the ability to make accurate load distribution decisions.
2Measurement precision
If load monitoring is performed on individual application software, then detailed load information can be obtained, but real-time load distribution becomes difficult to implement
Solution Approach 1:
The patent segments the load monitoring task by dividing it into three independent resource monitoring components (CPU load, memory usage, data communication volume) rather than monitoring each application software individually. This segmentation reduces the overall complexity and enables real-time processing while still providing comprehensive load information through the aggregation of these three key metrics.
3Ease of operation
If software components are allocated based on current load state, then simple allocation can be performed, but proactive prevention of overload cannot be achieved
Solution Approach 1:
The patent implements preliminary action by predicting future load states based on current trends in CPU load, memory usage, and data communication volume. This prediction capability allows the system to proactively allocate software components before overload occurs, rather than simply reacting to current conditions. The maintaining unit continuously updates allocations based on predicted future states, preventing overload while maintaining operational simplicity.
Solution Approach 2:
The patent establishes a feedback mechanism where the maintaining unit continuously monitors resource usage trends, predicts future load states, and adjusts software component allocations accordingly. This closed-loop feedback system enables proactive overload prevention while keeping the allocation process automated and simple to operate, as the system self-adjusts based on predicted conditions.
Data Source
AI summary
A computing-device management device includes a future load prediction unit that calculates a load prediction value based on load information of the computing device informed by the computing device and determines whether the load prediction value exceeds a predetermined threshold value. A software allocation unit detects, as a target computing device to be in an overloaded state in future, the computing device determined by the future load prediction unit to have the load prediction value exceeding the predetermined threshold value and determines which computing device is to be an allocation destination of at least one software component operating in the target computing device based on a CPU load, a used memory volume, and a data communication volume of the computing device. An informing unit informs the target computing device and the allocation-destination computing device of information of the allocated software component.