Storage System Performance Evaluation Using Z-Score Deviation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating system performance focus on averages, neglecting deviations that can lead to unacceptable latency spikes, resulting in an incomplete picture of system behavior.
Innovation Solution
A method that samples data points over a period, calculates deviation values by subtracting system specifications, averages these deviations, calculates standard deviation, and divides by the standard deviation to produce a modified performance value accounting for operational characteristics, providing a more comprehensive assessment of system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If average performance is used to evaluate system performance, then overall system performance is captured, but deviations and latency spikes are ignored
Solution Approach 1:
The patent transforms the performance evaluation from using raw performance values to using standardized z-scores. By converting performance measurements into standardized units that account for both mean and standard deviation, the system can identify deviations from expected performance behavior. This parameter transformation enables detection of latency spikes and inconsistencies that would be hidden in average-only metrics.
2Productivity
If only average performance metrics are collected, then data collection is simple, but performance consistency and latency variations are not detected
Solution Approach 1:
The patent implements a feedback mechanism where performance data is continuously collected, standardized, and compared against expected performance characteristics. The z-score calculation provides immediate feedback about whether performance deviations are statistically significant, enabling the system to identify consistency issues without requiring complex analysis of every individual measurement.
3Reliability
If detailed performance sampling and deviation calculation is performed, then performance consistency is evaluated, but computational complexity increases
Solution Approach 1:
The patent simplifies the evaluation process by transforming raw performance data into standardized z-scores using well-established statistical formulas. This parameter transformation reduces complex performance evaluation to straightforward calculations involving mean and standard deviation, making the process computationally efficient while maintaining the ability to detect performance inconsistencies.
Data Source
AI summary
Described embodiments provide storage system evaluation and comparison processes. An aspect includes sampling data points for a workload running on system over a sampling period. The data points indicate a performance metric with respect to operational characteristics of the system. An aspect further includes subtracting a system specification value from each of the averaged sampled data points, thereby producing deviation values reflecting a deviation of the sampled data points from the system specification value. An aspect also includes averaging the sampled data points, calculating a standard deviation of the averaged sampled data points, and dividing the variance value by the standard deviation, thereby producing a modified performance value that accounts for a deviation in the operational characteristics of the system over the sampling period.