Dynamic Sample Size Adjustment for Storage Capacity Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predicting overutilization conditions in limited-capacity entities, such as storage devices, is challenging due to abrupt and unpredictable changes in available capacity, making it difficult to prevent performance degradation or catastrophic failures.
Innovation Solution
A method for a resource-utilization monitor that adjusts sample size based on statistical analysis to forecast the likelihood of reaching a critical capacity level, using a processor to determine the validity of a sample set and continuously adjust the sample size for accurate and timely predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistical analysis with fixed sample size is used to predict storage capacity trends, then prediction capability is provided, but accuracy deteriorates when capacity changes abruptly or randomly
Solution Approach 1:
The patent implements dynamic sample size adjustment where the monitor modifies the number of samples collected based on observed capacity change patterns. When abrupt or random changes are detected, the sample size increases to capture sufficient data for accurate statistical analysis, while during stable periods, the sample size decreases to reduce overhead. This dynamic adaptation resolves the contradiction by making the prediction system reliable across varying capacity change behaviors.
Solution Approach 2:
The system incorporates feedback mechanisms where prediction accuracy and capacity change patterns are continuously monitored. The statistical analysis results feed back into adjusting future sample sizes - when predictions show high variance or capacity changes are erratic, the system increases sampling intensity. This feedback loop ensures prediction accuracy is maintained despite unpredictable capacity changes by adaptively responding to observed patterns.
2Measurement precision
If larger sample size is used for statistical analysis, then prediction accuracy improves, but resource consumption increases
Solution Approach 1:
The patent employs dynamic sample size adjustment that balances accuracy requirements with resource constraints. During periods of stable capacity utilization, the system reduces sample size to minimize computational overhead while maintaining adequate prediction capability. When capacity changes become erratic or critical thresholds are approached, the system automatically increases sampling intensity. This dynamic approach ensures prediction accuracy is optimized without unnecessarily consuming computational resources during stable periods.
3Speed
If frequent monitoring is performed to detect capacity changes, then timely detection capability improves, but system overhead increases
Solution Approach 1:
The patent implements periodic monitoring with variable intervals rather than continuous monitoring. The system adjusts the frequency of capacity checks based on observed utilization patterns and prediction results. During stable periods with low risk of capacity exhaustion, monitoring intervals are extended to reduce overhead. When capacity approaches critical levels or erratic changes are detected, the monitoring frequency increases automatically. This periodic action with adaptive intervals achieves timely detection while minimizing system overhead.
Data Source
AI summary
A method and associated systems for a resource-utilization monitor with self-adjusting sample size. A processor monitors availability of a resource-limited entity by analyzing a set of samples that each identify, at the time the sample was recorded, an unused amount of resource available to the entity. The processor computes a Chi-square statistic of the sample set. If the statistic reveals that the number of samples in the sample set is too small to produce valid results, the processor adjusts the sample size to specify a larger number of samples and repeats these steps. If the number of samples is large enough to produce statistically valid results, the processor analyzes the sample set to determine whether the current amount of remaining resource is undesirably low, is likely to become undesirably low during a short-term period of time, or is likely to become undesirably low during a longer-term period of time.