Storage Configuration Selection Using Predictive Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Selecting an optimal configuration of storage resources in a datacenter to meet agreed-upon service levels for performance and dependability while optimizing cost is challenging due to the large number of possible configurations and the impracticality of testing each directly.
Innovation Solution
Estimating the probability of failure scenarios and system performance in various configurations using models and benchmark data to identify an optimal full configuration that meets service level objectives with reduced testing efforts, focusing on a smaller set of test configurations to estimate performability and cost effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all possible storage resource configurations are tested directly to determine optimal configuration, then measurement precision of service level compliance is improved, but loss of time and productivity deteriorate due to the large number of configurations
Solution Approach 1:
The patent segments the configuration evaluation process into two distinct phases: (1) a benchmarking phase where a subset of configurations is tested to build predictive models, and (2) an estimation phase where the models predict performance of remaining configurations without direct testing. This segmentation allows comprehensive evaluation while minimizing actual testing time.
Solution Approach 2:
The patent performs preliminary benchmarking on a selected subset of configurations to establish predictive models before evaluating the full configuration space. This preliminary action on representative samples enables accurate estimation of remaining configurations without exhaustive testing, significantly reducing total evaluation time.
2Reliability
If all possible storage resource configurations are tested directly, then reliability of service level agreement compliance is improved, but device complexity and cost increase due to extensive testing requirements
Solution Approach 1:
The patent creates predictive models that copy the performance characteristics of actual configurations based on benchmark data from a subset. These models serve as virtual representations that can be evaluated without physical testing, reducing complexity while maintaining reliability through statistically sound prediction methods.
Solution Approach 2:
The patent changes the evaluation approach from direct physical testing to mathematical estimation using predictive models. By transforming the problem from empirical measurement to model-based prediction, the system reduces testing complexity while maintaining reliable service level compliance assessment through validated estimation techniques.
3Productivity
If a smaller set of test configurations is used to estimate performability, then productivity is improved by reducing testing efforts, but measurement precision deteriorates
Solution Approach 1:
The patent implements feedback mechanisms where benchmark results from the test configuration subset continuously refine and validate the predictive models. This feedback loop ensures that even with limited direct testing, the models achieve sufficient precision by learning from actual performance data and adjusting predictions accordingly.
Solution Approach 2:
The predictive models serve multiple functions: they estimate performance of untested configurations, validate service level compliance, and guide configuration selection. This multi-functionality allows the system to achieve high productivity with limited testing while maintaining measurement precision through the models' versatile application across different evaluation scenarios.
Data Source
AI summary
There is provided a computer-implemented method for selecting from a plurality of full configurations of a storage system an operational configuration for executing an application. An exemplary method comprises obtaining application performance data for the application on each of a plurality of test configurations. The exemplary method also comprises obtaining benchmark performance data with respect to execution of a benchmark on the plurality of full configurations, one or more degraded configurations of the full configurations and the plurality of test configurations. The exemplary method additionally comprises estimating a metric for executing the application on each of the plurality of full configurations based on the application performance data and the benchmark performance data. The operational configuration may be selected from among the plurality full configurations based on the metric.