Predictive CPU Resource Control via Correlation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data center management systems face challenges in efficiently managing calculation resources to meet service level agreements (SLAs) due to time lags in responding to changes in load, leading to potential SLA violations and increased power consumption.
Innovation Solution
A control device that uses just-in-time (JIT) modeling to predict CPU use rates by selecting relevant reference data based on correlation coefficients, allowing for timely adjustments to calculation resources such as CPU allocation, thereby reducing power consumption while ensuring SLA satisfaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional monitoring systems are used to track calculation resources, then resource allocation can be adjusted, but time lags occur in responding to load changes leading to SLA violations
Solution Approach 1:
The system performs preliminary actions by calculating predictive values of reference data using correlation coefficients before actual load changes occur. This allows the monitoring system to anticipate resource needs and adjust allocation proactively, eliminating response time lags and ensuring SLA satisfaction.
Solution Approach 2:
The system dynamically adjusts resource allocation based on real-time correlation analysis between different reference data sets. By continuously updating predictive models with current system state, the monitoring system adapts to changing load patterns without time lags, maintaining optimal resource distribution.
2Reliability
If more calculation resources are allocated to handle peak loads, then SLA satisfaction improves, but power consumption increases
Solution Approach 1:
By calculating predictive values before peak loads occur, the system can pre-allocate resources efficiently only when and where needed. This avoids continuous over-provisioning of resources, reducing power consumption while maintaining SLA satisfaction during actual peak demand periods.
Solution Approach 2:
The system changes resource allocation parameters dynamically based on predictive correlations rather than static over-provisioning. By adjusting resource levels according to calculated predictive values, the system optimizes the balance between SLA satisfaction and power consumption, allocating resources only when predictive analysis indicates actual need.
Data Source
AI summary
An apparatus stores plural pieces of reference data selected based on operation data input from a host OS executed by an information processing device. The apparatus selects one or more pieces of selection reference data from the plural pieces of reference data, based on a plurality of correlation coefficients which are calculated upon receiving a predictive demand and which respectively indicate correlations between predictive target reference data indicating reference data to be predicted and other reference data. The apparatus calculates predictive values of the predictive target reference data, based on the selected one or more pieces of selection reference data, and controls the plurality of calculation resources for the information processing device, based on the predictive values of the predictive target reference data and a predetermined reference value.